Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
8
6.12M
function [ bs, block ] = Score( s, DNA, n) % s - odkud začíná konsenzus % DNA - typ cell, ve které jsou zadané sekvence % n - jak bude konsenzus dlouhý (jak daleko od začátku je konec konsenzusu) % bs .. to skóre součet maxim (frekvence) % block .. výpis o jaké konsenzy se jednalo for i = 1: length(DNA) block(i,:) = DNA{1, i}(s(i):s(i)+n-1); end for j = 1:length(block) sum_A(j) = length(find(block(:,j)=='A')); sum_C(j) = length(find(block(:,j)=='C')); sum_T(j) = length(find(block(:,j)=='T')); sum_G(j) = length(find(block(:,j)=='G')); end ACTG =[sum_A;sum_C;sum_T;sum_G]; for q=1:length(ACTG) pom(q) = max(ACTG(:,q)); end bs = sum(pom); end
clc clear close all labeltsize=25; fw = 'normal'; %%是否加粗斜体之类 fn='Times New Roman'; linewide1=5; mkft = 15; load 2N.mat figure(1); hold on plot(SNRout,Pd_SGLRT_mc,'k-o','linewidth',linewide1,'MarkerSize',5) plot(SNRout,Pd_SRAO_mc,'r-x','linewidth',linewide1,'MarkerSize',mkft) plot(SNRout,Pd_SWALD_mc,'b-x','linewidth',linewide1,'MarkerSize',mkft) h_leg=legend('GLRT','RAO','WALD'); xlabel('SNR(dB)','FontSize',labeltsize,'FontWeight',fw,'FontName',fn) ylabel('PD','FontSize',labeltsize,'FontWeight',fw,'FontName',fn) set(gca,'FontSize',labeltsize) set(gcf,'Position',[700 0 1300 1000]) set(h_leg,'Location','NorthWest') grid on box on str=['SNR2N.eps']; print(gcf,'-depsc',str) %保存为png格式的图片到当前路径 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% load 10N.mat figure(2); hold on plot(SNRout,Pd_SGLRT_mc,'k-o','linewidth',linewide1,'MarkerSize',5) plot(SNRout,Pd_SRAO_mc,'r-x','linewidth',linewide1,'MarkerSize',mkft) plot(SNRout,Pd_SWALD_mc,'b-x','linewidth',linewide1,'MarkerSize',mkft) h_leg=legend('GLRT','RAO','WALD'); xlabel('SNR(dB)','FontSize',labeltsize,'FontWeight',fw,'FontName',fn) ylabel('PD','FontSize',labeltsize,'FontWeight',fw,'FontName',fn) set(gca,'FontSize',labeltsize) set(gcf,'Position',[700 0 1300 1000]) set(h_leg,'Location','NorthWest') grid on box on str=['SNR10N.eps']; print(gcf,'-depsc',str) %保存为png格式的图片到当前路径 pause(1) close all
% determine the 3dB mainlobe width function res = res(af, delay_delta) s = size(af); res = zeros(s(1),1); % loop across the Doppler axis for jj=1:s(1) % figure out the mainlobe boundary for this slice [~, mid] = max(af(jj,:)); the_max = 0; the_min = 0; for ii=mid:s(2) if af(jj,ii) < 0.5 the_max = ii-1; break; end end for ii=mid-1:-1:1 if af(jj,ii) < 0.5 the_min = ii+1; break; end end res(jj) = (the_max-the_min)*delay_delta; end end
function analyzesim(varargin) % function analyzesim(filename,opts) % or analyzesim() % Runs the standard set of analysis procedures on a file, or in the base workspace. % Calculates center of mass, swimming direction, kinematic parameters, % muscle work, fluid stress and energy. Saves to an output file % ('outfile',filename). opt.curvesmoothcutoff = 0.025; % filter cutoff for curvature % opt.nboundarypts = 30; opt.steadythresh = 0.05; % threshold velocity fluctuation. Less than this is "steady" opt.steadydX = 0.03; % threshold velocity fluctuation. Less than this is "steady" opt.leftsidecurvature = -1; % definition of "left" in terms of curvature opt.leftsidepeak = 1; % definition of "left" in terms of peak direction % peak finding procedure % Look for local max/min. Find points > pkthresh*max. Peak is center % of all neighboring points > pkthresh*max. opt.pkthresh = 0.9; opt.outfile = ''; % output file name opt.samraibasedir = ''; % base directory for all the samrai data files opt.rho = 1; % density in CGS units opt.savepressure = false; opt.savevorticity = false; if ((nargin >= 1) && ischar(varargin{1}) && ... exist(varargin{1},'file')), datafilename = varargin{1}; fprintf('**** File %s\n', varargin{1}); filenameopt = {'-file',varargin{1}}; p = 2; else fprintf('Base workspace\n'); datafilename = ''; filenameopt = {}; p = 1; end; opt = parsevarargin(opt, varargin(p:end), p); % necessary variables disp(' Loading variables...'); if (~getvar(filenameopt{:}, 'actl','actr','fmusPl','fmusPr',... 'fxlmus','fxltot','fxmtot','fxntot','fxrmus','fxrtot',... 'fylmus','fyltot','fymtot','fyntot','fyrmus','fyrtot',... 't','xl','yl','xm','ym','xn','yn','xr','yr',... 'ul','vl','um','vm','un','vn','ur','vr')) error('Could not find necessary variables'); end; if (getvar(filenameopt{:}, 'good')), actl = actl(:,good); actr = actr(:,good); fmusPl = fmusPl(:,good); fmusPr = fmusPr(:,good); fxlmus = fxlmus(:,good); fxrmus = fxrmus(:,good); fylmus = fylmus(:,good); fyrmus = fyrmus(:,good); fxmtot = fxmtot(:,good); fymtot = fymtot(:,good); fxntot = fxntot(:,good); fyntot = fyntot(:,good); fxltot = fxltot(:,good); fyltot = fyltot(:,good); fxrtot = fxrtot(:,good); fyrtot = fyrtot(:,good); t = t(good); xl = xl(:,good); yl = yl(:,good); xm = xm(:,good); ym = ym(:,good); xn = xn(:,good); yn = yn(:,good); xr = xr(:,good); yr = yr(:,good); ur = ur(:,good); vr = vr(:,good); um = um(:,good); vm = vm(:,good); un = un(:,good); vn = vn(:,good); ul = ul(:,good); vl = vl(:,good); end; %optional variables %default values: disp(' Getting the simulation parameters...'); freq = 1; viscosity = 0.01; sfo = 2.56e6; sfo2 = ''; ps = 3; gridres = 32; isforcetaper = 'true'; if (~getvar(filenameopt{:}, 'freq','viscosity','sfo','sfo2','ps','gridres','isforcetaper',... '-keepundef')), prompt = {'freq','viscosity','sfo','sfo2','ps','gridres','isforcetaper'}; def = cellfun(@num2str,{freq,viscosity,sfo,sfo2,ps,gridres,isforcetaper},'UniformOutput',false); if (isempty(sfo2)), def{4} = '0.2*2.56e6'; end; if (isforcetaper), def{7} = 'true'; else def{7} = 'false'; end; vals = inputdlg(prompt,'Simulation constants',1,def); vals = cellfun(@eval,vals,'UniformOutput',false); [freq,viscosity,sfo,sfo2,ps,gridres,isforcetaper] = vals{:}; putvar(filenameopt{:}, 'freq','viscosity','sfo','sfo2','ps','gridres','isforcetaper'); end; nfr = size(xm,2); npt = size(xm,1); s0 = [zeros(1,nfr); cumsum(sqrt(diff(xm).^2 + diff(ym).^2))]; s = nanmean(s0,2); istether = false; disp(' Estimating com speed...'); if (all(abs(xm(1,:) - xm(1,1)) < 1e-3)) disp(' -> appears to be a tethered head simulation'); istether = true; comspeed = zeros(size(t)); if (isempty(opt.samraibasedir)) disp(' ... cannot get flow speed without samrai dirs'); flowspeed = 0; else samraidirs = getdirectorynames(fullfile(opt.samraibasedir,'visit*'))'; [x,y,V] = importsamrai(samraidirs{2},'vars',{'U_0','U_1'}, ... 'interpolaten',[100 50]); flowspeed = mean(V.U_0(x < xm(1,1)-1)); end; end; [width,area,sn, comx,comy,comvelx,comvely] = ... estcomspeed(t,xm,ym,xn,yn,xl,xr,yl,yr); if (isforcetaper), forcetaper = width(:,1) ./ max(width(:,1)); else forcetaper = ones(npt,1); end; if (~istether) comspeed = sqrt(comvelx.^2 + comvely.^2); %unit vector pointing in the direction of the swimming speed, as a cubic %spline in 3 parts, which should get rid of any fluctuation at the level of %the tail beat frequency dt = t(2) - t(1); dur = round(2/dt); %smooth over 2 sec comxsm = runavg(comx, dur); comysm = runavg(comy, dur); swimvecx = deriv(t,comxsm); swimvecy = deriv(t,comysm); mag = sqrt(swimvecx.^2 + swimvecy.^2); swimvecx = swimvecx ./ mag; swimvecy = swimvecy ./ mag; else swimvecx = -1 * ones(size(t)); swimvecy = zeros(size(t)); end; %now average speeds over cycles to look for "steady" %define cycles cycle = t * freq; cycle = cycle - min(floor(cycle)); cyclenum = floor(cycle)+1; cyclenum2 = floor(cycle*2)+1; disp(' Looking for steady speed...'); if (istether) ncycle = max(cyclenum); dt = t(2) - t(1); cycleindlen = round(1/freq/dt); k = 1:length(t)-cycleindlen; dX = NaN(size(t)); dX(k+cycleindlen) = mean(sqrt((xm(:,k+cycleindlen) - xm(:,k)).^2 + ... (ym(:,k+cycleindlen) - ym(:,k)).^2)); dXbycycle = accumarray(cyclenum(:), dX(:), [], @nanmean); steadycycle = length(dXbycycle) - 1; while ((steadycycle >= 1) && all(dXbycycle(steadycycle:end) < opt.steadydX)) steadycycle = steadycycle-1; end; else %average comspeed in each cycle comspeedbycycle = accumarray(cyclenum(:), comspeed(:), [], @nanmean); %and the amount it changes, relative to the current value dcomspeed = diff(comspeedbycycle) ./ comspeedbycycle(1:end-1); %look for those cycles that change less than opt.steadythresh steadycycle = length(comspeedbycycle)-1; while ((steadycycle >= 1) && all(dcomspeed(steadycycle:end) < opt.steadythresh)), steadycycle = steadycycle-1; end; end; if (steadycycle == 1) steadycycle = 2; end issteady = (cyclenum >= steadycycle); %amplitude = displacement relative to the center of mass, perpendicular to the swimming %vector ampcont = -(xm - comx(ones(npt,1),:)) .* swimvecy(ones(npt,1),:) + ... (ym - comy(ones(npt,1),:)) .* swimvecx(ones(npt,1),:); ampsteady = nanmean(range(ampcont(:,issteady))); goodamp = issteady | (range(ampcont) < 2*ampsteady); ampcont(:,~goodamp) = NaN; %track curvature waves %no smoothing in curvature, because we've already smoothed disp(' Estimating curvature...'); curve = curvature(xm,ym, 'discrete','smooth',opt.curvesmoothcutoff); %track zero crossings in curvature disp(' Tracking curvature zero crossings...'); [indzero0,sgnzero0] = ... analyzeKinematics(s,t,xm,ym, 'curve',curve, 'nsmoothcurve',0, ... 'zeros', 'backwnd',0.5, 'returnpeaksign'); %track peaks in amplitude envelope disp(' Tracking amplitude peaks...'); [indpeak0,sgnpeak0,~,~,~,~,~,wavespeed0,wavelen0,~,waven0] = ... analyzeKinematics(s,t,xm,ym, 'curve',ampcont, 'nsmoothcurve',10, ... 'peaks','backwnd',0.5, 'returnpeaksign'); %also track the activation wave disp(' Finding the activation wave...'); [indact,indactoff,isactleft] = findactivationwave(actl,actr); actlen = NaN(size(indact)); for j = 1:size(indact,2)-1, for i = 1:npt, if (isfinite(indact(i,j))), next = last(indact(:,j+1) <= indact(i,j)); if (~isempty(next)), actlen(i,j) = 2*(s(i)-s(next)); end; end; end; end; tact = NaN(size(indact)); good = isfinite(indact); tact(good) = t(indact(good)); actspeed = NaN(1,size(tact,2)); for i = 1:length(actspeed), good = isfinite(tact(:,i)); if (sum(good) > 100), p = polyfit(tact(good,i),s(good), 1); actspeed(i) = p(1); end; end; %early cycles are those with more than three activation starts on frame 1 early = sum(indact == 1) > 3; firstright = first(~isactleft & ~early); t0 = t(first(isfinite(indact(:,firstright)))); %match up the curvature to the activation disp(' Aligning curvature and amplitude peaks to activation wave...'); cyclematchzero = zeros(1,size(indact,2)); cyclematchpeak = zeros(1,size(indact,2)); numincyclezero = zeros(1,size(indact,2)); numincyclepeak = zeros(1,size(indact,2)); for i = 1:size(indact,2), indact1 = first(indact(:,i),isfinite(indact(:,i))); indact2 = last(indactoff(:,i),isfinite(indactoff(:,i))); if (isactleft(i)), sgn1 = opt.leftsidecurvature; else sgn1 = -opt.leftsidecurvature; end; numincycle1 = sum((indzero0 >= indact1) & (indzero0 <= indact2)); numincycle1(sgnzero0 ~= sgn1) = 0; [n1,a] = max(numincycle1); cyclematchzero(i) = a; numincyclezero(i) = n1; if (isactleft(i)), %NB: we want left side activation to correspond to peak right side %excursion, because that's appx when the left side muscle is %*shortest* sgn1 = -opt.leftsidepeak; else sgn1 = opt.leftsidepeak; end; numincycle1 = sum((indpeak0 >= indact1) & (indpeak0 <= indact2)); numincycle1(sgnpeak0 ~= sgn1) = 0; [n1,a] = max(numincycle1); cyclematchpeak(i) = a; numincyclepeak(i) = n1; end; for i = 1:size(indzero0,2), k = find(cyclematchzero == i); if (length(k) > 1), [~,a] = max(numincyclezero(k)); cyclematchzero(k) = 0; cyclematchzero(k(a)) = i; end; end; for i = 1:size(indpeak0,2), k = find(cyclematchpeak == i); if (length(k) > 1), [~,a] = max(numincyclepeak(k)); cyclematchpeak(k) = 0; cyclematchpeak(k(a)) = i; end; end; %rearrange data to match the activation waves indzero = NaN(size(indact)); good = cyclematchzero ~= 0; indzero(:,good) = indzero0(:,cyclematchzero(good)); sgnzero = NaN(1,size(indact,2)); sgnzero(:,good) = sgnzero0(:,cyclematchzero(good)); indpeak = NaN(size(indact)); good = cyclematchpeak ~= 0; indpeak(:,good) = indpeak0(:,cyclematchpeak(good)); sgnpeak = NaN(1,size(indact,2)); sgnpeak(good) = sgnpeak0(cyclematchpeak(good)); %find the peaks of curvature disp(' Looking for curvature peaks...'); indpeakcurve = NaN(size(indzero)); sgnpeakcurve = NaN(size(indzero)); for j = 1:size(indzero,2)-1, for i = 1:size(indzero,1), if (~isnan(indzero(i,j)) && ~isnan(indzero(i,j+1)) && ... (indzero(i,j+1) > indzero(i,j))), k = indzero(i,j):indzero(i,j+1); curve1 = curve(i,k); mx = max(abs(curve1)); ishigh = abs(curve1) >= opt.pkthresh*mx; sgn1 = sign(first(curve1,ishigh)); if (any(sign(curve1(ishigh)) ~= sgn1)), ctr = NaN; else weight = abs(curve1(ishigh)); if (length(weight) > 1), weight = (weight - min(weight))/range(weight); else weight = 1; end; ctr = sum(k(ishigh) .* weight) ./ ... sum(weight); end; indpeakcurve(i,j) = round(ctr); sgnpeakcurve(i,j) = sgn1; end; end; end; good = cyclematchpeak ~= 0; wavespeed = NaN(1,size(indzero,2)); wavespeed(good) = wavespeed0(cyclematchpeak(good)); waven = zeros(1,size(indzero,2)); waven(good) = waven0(cyclematchpeak(good)); wavelen = NaN(size(indzero)); wavelen(:,good) = wavelen0(:,cyclematchpeak(good)); amp = NaN(size(indact)); for c = 1:size(indact,2)-2, if any((cyclenum2 == c) | (cyclenum2 == c+1)) amp(:,c) = range(ampcont(:,(cyclenum2 == c) | (cyclenum2 == c+1)),2)/2; end end; disp(' Calculating work loops...'); [lennorm,worktot,workpos,workneg,workposact,worknegact] = workloop(t, xl,yl,fmusPl,actl, ... xr,yr,fmusPr,actr, ... 'per',1/freq, 'plot',false, 'forcetaper',forcetaper); disp(' Calculating fluid forces...'); Force = fluidforces(t,freq, swimvecx,swimvecy, xm,ym, fxmtot,fymtot, ... xl,yl,fxltot,fyltot, xr,yr,fxrtot,fyrtot); out.freq = freq; out.viscosity = viscosity; out.sfo = sfo; out.sfo2 = sfo2; out.ps = ps; out.gridres = gridres; out.isforcetaper = isforcetaper; out.t = t; out.s = s; out.s0 = s0; out.sn = sn; out.nfr = nfr; out.npt = npt; out.comx = comx; out.comy = comy; out.comvelx = comvelx; out.comvely = comvely; out.swimvecx = swimvecx; out.swimvecy = swimvecy; out.comspeed = comspeed; out.width = width; out.area = area; out.curve = curve; out.ampcont = ampcont; out.amp = amp; out.wavespeed = wavespeed; out.wavelen = wavelen; out.indact = indact; out.indactoff = indactoff; out.isactleft = isactleft; out.actspeed = actspeed; out.actlen = actlen; out.indzero = indzero; out.sgnzero = sgnzero; out.indpeak = indpeak; out.sgnpeak = sgnpeak; out.indpeakcurve = indpeakcurve; out.sgnpeakcurve = sgnpeakcurve; out.issteady = issteady; out.steadycycle = steadycycle; out.cyclenum = cyclenum; out.lennorm = lennorm; out.worktot = worktot; out.workpos = workpos; out.workneg = workneg; out.workposact = workposact; out.worknegact = worknegact; out.Force = Force; out.HGREV = savehgrev([], 'datafile',datafilename); if (~isempty(opt.samraibasedir) && inputyn('Load fluid data? ')) disp(' Calculating fluid quantities...'); samraidirs = getdirectorynames(fullfile(opt.samraibasedir,'visit*'))'; if (isempty(samraidirs)) saveextra = {}; else if (~isempty(opt.outfile)) [pn,fn,ext] = fileparts(opt.outfile); cont = fullfile(pn,[fn 'Continue.mat']); else cont = ''; end; if (~isempty(cont) && exist(cont,'file')) load(cont, 'Stress','samraidirs','fr','Energy'); fr0 = fr; fprintf('Continuing from frame %d...\n', fr0); else fr0 = 1; end; boundx0 = [xm(1,:); xl(2:end-1,:); xm(end,:); xr(end-1:-1:2,:); xm(1,:)]; boundy0 = [ym(1,:); yl(2:end-1,:); ym(end,:); yr(end-1:-1:2,:); ym(1,:)]; boundu0 = [um(1,:); ul(2:end-1,:); um(end,:); ur(end-1:-1:2,:); um(1,:)]; boundv0 = [vm(1,:); vl(2:end-1,:); vm(end,:); vr(end-1:-1:2,:); vm(1,:)]; Energy.totalke = NaN(1,nfr); Energy.dissip = NaN(1,nfr); Energy.boundflux = NaN(4,nfr); opt1.rho = 1; % in g/cm^3 opt1.mu = viscosity; opt1.savepressure = opt.savepressure; opt1.savevorticity = opt.savevorticity; N = min(length(samraidirs)-1, nfr); for fr = fr0:N, fprintf('Importing %s (%d%%)...\n', samraidirs{fr+1}, round((fr+1)/nfr*100)); V = importsamrai(samraidirs{fr+1},'vars',{'U_0','U_1','P'}); V = getsamraipatchedges(V); V = velderivsamrai(V); [totalke1,dissip1,boundflux1] = energybalance1(V,opt1); Energy.totalke(fr) = totalke1; Energy.dissip(fr) = dissip1; Energy.boundflux(:,fr) = boundflux1; S1 = fluidstressnearboundary1(V, boundx0(:,fr),boundy0(:,fr),... boundu0(:,fr),boundv0(:,fr), opt1); Stress.tanx{fr} = S1.tanx; Stress.tany{fr} = S1.tany; Stress.s{fr} = S1.s; Stress.n{fr} = S1.n; Stress.nearboundx{fr} = S1.nearboundx; Stress.nearboundy{fr} = S1.nearboundy; Stress.normstress{fr} = S1.normstress; Stress.tanstress{fr} = S1.tanstress; if (opt.savepressure) Stress.nearboundp{fr} = S1.nearboundp; end; if (opt.savevorticity) Stress.nearboundut{fr} = S1.nearboundut; Stress.nearboundun{fr} = S1.nearboundun; Stress.vorticity{fr} = S1.vorticity; end; if (~isempty(cont) && (mod(fr,10) == 0)) save(cont,'Stress','samraidirs','fr','Energy'); end; end; out.Energy = Energy; out.Stress = Stress; end; end; if (istether) out.flowspeed = flowspeed; end; disp(' Saving data...'); if (~isempty(opt.outfile)), save(opt.outfile, '-struct','out'); else putvar(filenameopt{:}, '-fromstruct','out', '-all'); end;
classdef neighbourhood < handle properties label m % row n % column end properties (Transient) n_selfneighs end methods function obj = neighbourhood(m, n, varargin) if nargin<1 obj.m = 1; obj.n = 1; else obj.m = m; obj.n = n; end end function obj = uplus(obj1) obj = utils.uplus(obj1); end function set_n_selfneighs(obj) obj.n_selfneighs = ones(obj.m*obj.n,1); % preset so as to normalize to one in the next function % count size of neighbourhood of each node obj.n_selfneighs = obj.neigh_avg(ones(obj.m*obj.n,1)); end function neigh_avg(obj, state) % if m or n were changed in the meantime, update n_selfneighs if ~isequal(size(state),size(obj.n_selfneighs)) obj.set_n_selfneighs(); end end end methods(Static) function [m, n] = get_m_n(k, m_seed, n_seed) if nargin<2; m_seed = 1; n_seed = 1; end if m_seed<n_seed m = m_seed*2^ceil((k-1)/2); n = n_seed*2^floor((k-1)/2); else m = m_seed*2^floor((k-1)/2); n = n_seed*2^ceil((k-1)/2); end end function [m, n] = get_m_n_N(N) for m=floor(sqrt(N)):-1:1 if mod(N,m)==0; break; end end n = N/m; end end end
%Nhom 1 %53 %Pham Ba Tung %B15DCDT221 I0=imread('chuoi.jpg'); size(I0) subplot(2,2,1 ) imshow(I0) I=rgb2gray(I0); I1=im2bw(I0); subplot(2,2,2) imshow(I) subplot(2,2,3) imshow(I1) [n m] = size(I); T = 128; for i=1:n for j=1:m if I(i,j) < T I(i,j) = 0; if j<m I(i,j+1) = I(i,j); end else e = 255 - I(i,j); I(i,j) = 255; if j<m I(i,j+1) = e; end end end end subplot(2,2,4) imshow(I)
function tokenList = split(str, delimiter) % SPLIT Split a string based on a given delimiter % Usage: % tokenList = split(str, delimiter) % Roger Jang, 20010324 tokenList = {}; remain = str; i = 1; while ~isempty(remain), [token, remain] = strtok(remain, delimiter); tokenList{i} = token; i = i+1; end
% CLASSIFYRBFN Classify data with radial basis function network. % OUT = CLASSIFYRBFN(X,MODEL) classifies the data in X (N samples-by-D % features) by using a radial basis function network (RBFN), where MODEL is % a structure with the RBFN parameters: centroids, spreads, and weights. % OUT is a structure with both the scores obtained from output units and the % class labels assigned to each sample in X. If OUT.Labels=1 is a benign lesion, % whereas if OUT.Labels=2 is a malignant lesion % % Example: % ------- % load('bcwd.mat'); % ho = crossvalind('HoldOut',Y,0.2); % Hold-out 80-20% % [Xtr,m,s] = softmaxnorm(X(ho,:)); % Training data normalization % Xtt = softmaxnorm(X(~ho,:),[m;s]); % Test data normalization % Ytr = Y(ho,:); % Training targets % Ytt = Y(~ho,:); % Test targets % Model = trainRBFN(Xtr,Ytr); % Train RBFN % Out = classifyRBFN(Xtt,Model); % Test RBFN % perf = classperf(Out.Labels,Ytt); % Evaluate classification performance % % See also AUC CLASSPERF CLASSIFYLDA CLASSIFYSVM CLASSIFYLSVM CLASSIFYRF % % % References: % ---------- % Theodoridis S, Koutroumbas K. Pattern recognition. 4th edition. % Burlington, MA: Academic Press 2009. % % K. L. Priddy, P. E. Keller, Artificial Neural Networks: An Introduction. % Bellingham, WA: SPIE-The Int. Soc. Optical Eng., 2005. % ------------------------------------------------------------------------ % Cinvestav-IPN (Mexico) - LUS/PEB/COPPE/UFRJ (Brazil) % CLASSIFYRBFN Version 1.0 (Matlab R2014a Unix) % November 2016 % Copyright (c) 2016, Wilfrido Gomez Flores % ------------------------------------------------------------------------ function Out = classifyRBFN(Xtt,Model) % Load model Ci = Model.Centroids; Si = Model.Sigmas; W = Model.Weights; % Classify Ntt = size(Xtt,1); DV = eucdist(Xtt,Ci); S = Si(ones(Ntt,1),:); aux2 = exp(-DV./(2*S.^2)); phi2 = cat(2,ones(Ntt,1),aux2); Out.Scores = phi2*W; [~,Out.Labels] = max(Out.Scores,[],2); %******************************************************************* function D = eucdist(A,B) nA = sum(A.*A,2); nB = sum(B.*B,2); D = abs(bsxfun(@plus,nA,nB') - 2*(A*B'));
% this function is to calcaluate and draw 2 dimensional histogram. function samples = twoDbin load PTsampleing.mat; X = chain(:,1); Y = chain(:,2); xbins = -1.5:1:1.5; % xbins(find(xbins==0))=[]; ybins = -1.5:1:1.5; % ybins(find(ybins==0))=[]; xNumBins = numel(xbins); yNumBins = numel(ybins); Xi = round(interp1(xbins,1:xNumBins,X,'linear','extrap')); Yi = round(interp1(ybins,1:yNumBins,Y,'linear','extrap')); Xi = max( min(Xi,xNumBins), 1); Yi = max( min(Yi,yNumBins), 1); H = accumarray([Yi(:) Xi(:)],1,[yNumBins xNumBins]); %figure %imagesc(xbins, ybins, H), axis on %# axis image %colormap hot; colorbar %hold on, plot(X, Y,'b.','MarkerSize',1), hold off H(2,:) = H(2,:)+H(3,:); H(3,:) = []; H(:,2) = H(:,2)+H(:,3); H(:,3) =[] samples = reshape(H,1,[]); samples = samples/sum(samples); return; clear
function [Vstart,Vstop,Xstart,Xstop] = create_goal(fig) cmap = lines(5); F = [1 2 3 4; 5 6 7 8; 1 2 6 5; 2 3 7 6; 3 4 8 7; 1 4 8 5]; % Start x = 130; y = 300; z = 20; w = 20; h = 20; Vx_start = [x x+w x+w x x x+w x+w x]; Vy_start = [y y y+h y+h y y y+h y+h]; Vz_start = [0 0 0 0 z z z z]; Vstart = [Vx_start' Vy_start' Vz_start']; Xstart = [x+w/2 y+h/2 z/2 0 0 0]'; x = 250; y = 100; w = 20; h = 20; z = 20; Vx_stop = [x x+w x+w x x x+w x+w x]; Vy_stop = [y y y+h y+h y y y+h y+h]; Vz_stop = [0 0 0 0 z z z z]; Vstop = [Vx_stop' Vy_stop' Vz_stop']; Xstop = [x+w/2 y+h/2 z/2 0 0 0]; if ~isempty(fig) figure(fig); hold on; patch('Faces',F,'Vertices',Vstart,'FaceColor','None','EdgeColor',cmap(5,:)); patch('Faces',F(1,:),'Vertices',Vstart(1:4,:),'FaceColor',cmap(5,:),'FaceAlpha',0.5); plot3(Xstart(1),Xstart(2),Xstart(3),'*g','MarkerSize',15,'Linewidth',2); patch('Faces',F,'Vertices',Vstop,'FaceColor','None','EdgeColor',cmap(2,:)); patch('Faces',F(1,:),'Vertices',Vstop(1:4,:),'FaceColor',cmap(2,:),'FaceAlpha',0.5); end end
% Enhances the contrast of the image % Adapted from http://imageprocessingblog.com/histogram-adjustments-in-matlab-part-i/ function [out] = enhance_contrast(im) hsvimg = rgb2hsv(im); for ch=2:3 hsvimg(:,:,ch) = imadjust(hsvimg(:,:,ch), ... stretchlim(hsvimg(:,:,ch), 0.01)); end out = hsv2rgb(hsvimg); end
clc clear all close all % % a=input('Enter the Choice') disp('This is a MASK of Median Filter') q1=input('Enter the size of Mask ') g1=[1 1 1;1 1 1;1 1 1] subplot(2,2,3) imshow(g1,[]) title('This is Mask') a1=imread('cameraman.tif'); subplot(2,2,1) a11=double(a1); imshow(a1) title('Original image') [r,c] = size(a11); a2= imnoise(a1,'salt & pepper',0.02); subplot(2,2,2) imshow(a2) title('Input Image with Salt and Pepper Noise') NIm=zeros(r,c); t=(q1-1)/2; xc=t+1; q2=((q1+1)/2)-1 for i=xc:r-1 for j=xc:c-1 temp=0; for k=-t:t for l=-t:t temp(i,j)=(a2(k+i,l+j).*g1(k+xc,l+xc)); temp2(k+2,l+2)=(temp(i,j)); temp3=temp2(:); temp4=sort(temp3'); temp5=temp4(ceil(end/2)); NIm(i+q2-1,j+q2-1)=temp5; end end end end subplot(2,2,4) imshow(NIm,[]) title('Output Image')
%% Params verbose = 0; use_interval = true; solvers = {'mosek' 'mosek_INTPNT_ONLY' 'sedumi' 'sdpt3'}; basis = 'CG_red'; slack = 'slack(z=x+t-1)'; slackcode = 3; %slack = 'slack(z=x-t)'; slackcode = 1; nout = 5; dist_A = low2high_dist('W-type', nout, use_interval, 0);%infea dist_B = low2high_dist('uniform', nout, use_interval, 0);%fea %dist_A = [];dist_B = []; %start_point = 0.035; %stop_point = 0.045; %start_point = 0.05276; %stop_point = 0.05277; %start_point = 0.3333333; %stop_point = 0.33333336; start_point = 0.065;stop_point = 0.075; %start_point = 0.0573;stop_point = 0.05732;%nout4 %start_point = 0.057352;stop_point = 0.057353;%nout3 steps = 10; %% I/O dt = datestr(datetime('now'), 'mmm-dd-yyyy_HH.MM'); results_path = sprintf('results/slack%d_%s_%s.mat', slackcode, basis, dt); data = struct(); data.start_point = start_point; data.stop_point = stop_point; data.steps = steps; data.nout = nout; data.basis = basis; data.slack = slack; %% Trials for solver = solvers lbs =calc_lbound(nout, dist_A, dist_B, solver{1}, basis, slack, start_point, stop_point, steps, verbose, use_interval); if slackcode == 3 data.([solver{1} '_lbs']) = lbs - 1; else data.([solver{1} '_lbs']) = lbs; end end save(results_path, 'data'); %% Call plotting function lb_plot(results_path);
function [ numfluxSolver, surfluxSolver, volumefluxSolver ] = initFluxSolver( obj ) %INITNUMFLUXSOLVER Summary of this function goes here % Detailed explanation goes here if obj.option.isKey('NumFluxType') if obj.option.isKey('nonhydrostaticType') if obj.getOption('NumFluxType') == enumSWENumFlux.HLL &&... obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Hydrostatic % Numerical flux type pointed and non-hydrostatic type pointed numfluxSolver = SWEHLLNumFluxSolver1d( ); surfluxSolver = SWEFaceFluxSolver1d( ); % LF and Roe flux to be added elseif obj.getOption('NumFluxType') == enumSWENumFlux.HLL &&... obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Nonhydrostatic numfluxSolver = SWENonhydroHLLNumFluxSolver1d( ); surfluxSolver = SWENonhydroFaceFluxSolver1d( ); % LF and Roe flux to be added end else % No nonhydrostatic type pointed, and hydrostatic by default if obj.getOption('NumFluxType') == enumSWENumFlux.HLL numfluxSolver = SWEHLLNumFluxSolver1d( ); surfluxSolver = SWEFaceFluxSolver1d( ); % LF and Roe flux to be added end end else % No numerical flux type pointed if obj.option.isKey('nonhydrostaticType') if obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Hydrostatic numfluxSolver = SWEHLLNumFluxSolver1d( ); surfluxSolver = SWEFaceFluxSolver1d( ); elseif obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Nonhydrostatic numfluxSolver = SWENonhydroHLLNumFluxSolver1d( ); surfluxSolver = SWENonhydroFaceFluxSolver1d( ); end else % No nonhydrostatic type pointed, and hydrostatic by default numfluxSolver = SWEHLLNumFluxSolver1d( ); surfluxSolver = SWEFaceFluxSolver1d( ); end end parent = metaclass(obj); if strcmp(parent.SuperclassList.Name, 'SWEConventional1d') if obj.option.isKey('nonhydrostaticType') if obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Hydrostatic volumefluxSolver = SWEVolumeFluxSolver1d(); else volumefluxSolver = SWENonhydroVolumeFluxSolver1d(); end else volumefluxSolver = SWEVolumeFluxSolver1d(); end elseif strcmp(parent.SuperclassList.Name, 'SWEPreBlanaced1d') % SWEPreBlanaced1d if obj.option.isKey('nonhydrostaticType') if obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Hydrostatic volumefluxSolver = SWEPrebalanceVolumeFlux1d(); else volumefluxSolver = SWENonhydroPrebalanceVolumeFluxSolver1d(); end else volumefluxSolver = SWEPrebalanceVolumeFlux1d(); end elseif strcmp(parent.SuperclassList.Name, 'SWEWD1d') if obj.option.isKey('nonhydrostaticType') if obj.getOption('nonhydrostaticType') == enumNonhydrostaticType.Hydrostatic volumefluxSolver = SWEWDVolumeFlux1d(); else volumefluxSolver = SWENonhydroWDVolumeFluxSolver1d(); end else volumefluxSolver = SWEWDVolumeFlux1d(); end end end
function [ vwillr ] = percentr( vtime,vhigh,vlow,vclose,len ) %PERCENTR 计算willr,根据mc % version 1.0 , luhuaibao, 2013.9.27 [~,id] = sort( -vtime ); % 最近的排在上面 vhigh = vhigh(id); vlow = vlow(id); vclose = vclose(id); var0 = max( vhigh(1:len ) ); var1 = var0 - min( vlow(1:len) ) ; if var1 ~= 0 vwillr = 100 - ( ( var0 - vclose(1) ) / var1 ) * 100 ; else vwillr = 0 ; end end
clear, clc xrng = [100 1000]; yrng = [-350 350]; xdiv = 40; ydiv = 40; % number of x/y divisions hbins = linspace(xrng(1), xrng(2), xdiv); vbins = linspace(yrng(1), yrng(2), xdiv); frmat = zeros(xdiv, ydiv); frtrls = zeros(xdiv, ydiv); targetdir = 'C:\Users\Hrishikesh\Data\krPTBData\'; [filename pathname] = uigetfile([targetdir 'S3*.mat'], 'Load Exp Session File (not sp2)', 'MultiSelect', 'on'); fullpathname = strcat(pathname, filename); % all the files in pathname %% Because I want to combine files and build up the firing rate plots if iscell(fullpathname) numfiles = length(fullpathname); else numfiles = 1; end hasPrintedOnce = false; allwindow = 0.02:0.01:0.2; for wi = 1:length(allwindow) % time back window = allwindow(wi); for dt = 1:numfiles if iscell(fullpathname) thisfilename = fullpathname{dt}; rawname = filename{dt}; else thisfilename = fullpathname; rawname = filename; end % first load the session file load(thisfilename) % this has the locs and successes load(strcat(thisfilename(1:end-4), '_sp2.mat')) eval(['eyeh = ' rawname(1:end-4) '_Ch3.values;']) eval(['eyev = ' rawname(1:end-4) '_Ch4.values;']) eval(['trig = ' rawname(1:end-4) '_Ch5.values;']) eval(['trigTS = ' rawname(1:end-4) '_Ch5.times;']) eval(['photo = ' rawname(1:end-4) '_Ch6.values;']) eval(['photoTS = ' rawname(1:end-4) '_Ch6.times;']) eval(['Allspktimes = ' rawname(1:end-4) '_Ch7.times;']) eval(['spkcodes = ' rawname(1:end-4) '_Ch7.codes;']) eval(['eyeSamplingRate = ' rawname(1:end-4) '_Ch3.interval;']) numIdx1sec = round(1/eyeSamplingRate); %numIdxLittlePost = round(0.5/eyeSamplingRate); clus = 1; spktimes = Allspktimes(spkcodes(:,1) == clus); %seconds if ~hasPrintedOnce, fprintf('Num Clusters: %i, Cluster Plotted: %i \n', length(unique(spkcodes(:,1))), clus), end %% Get data (bookkeeping) % smooth out the photocell idxPhoto = photo > 0.05; photo(idxPhoto) = 0.3; photo(~idxPhoto) = 0; dphoto = diff(photo); dphoto = [0; dphoto]; % to take care of the indexing issue idxOn = find(dphoto == 0.3); % photocell on idxOff = find(dphoto == -0.3); % note, the first one will be due to the PTB syncing routine idxOn(1) = []; idxOff(1) = []; % - also importantly, note that the times that the photo is on/off % photoTS(idxOn(i)) or photoTS(idxOff(i)) % smooth out triggers idxTrig = trig > 0.1; trig(idxTrig) = 0.5; trig(~idxTrig) = 0; dTrig = diff(trig); dTrig = [0; dTrig]; idxTstart = find(dTrig == 0.5); idxTstop = find(dTrig == -0.5); %% just find when the flashes happened in the "storeSuccess" lists timeFlashesStarts = []; timeFlashesEnds = []; nonzeroidx = find(storeSuccess); for ti = 1:length(find(storeSuccess)) trl = storeSuccess(nonzeroidx(ti)); thisIndFlashes = find(idxOn > idxTstart(trl) & idxOn < idxTstop(trl)); thisNumFlashes = length(thisIndFlashes); if thisNumFlashes == 5 tflashes = nan(10,1); tflashes([1,3,5,7,9]) = photoTS(idxOn(thisIndFlashes)); tflashes([2,4,6,8,10]) = photoTS(idxOff(thisIndFlashes)); timeFlashesStarts(end+1:end+10,1) = tflashes; tflashes = nan(10,1); tflashes([1,3,5,7,9]) = photoTS(idxOff(thisIndFlashes)); tflashes([2,4,6,8]) = photoTS(idxOn(thisIndFlashes(2:end))); tflashes(10) = tflashes(9)+nanmean(diff(tflashes)); timeFlashesEnds(end+1:end+10,1) = tflashes; else fprintf('Possible Error with Trial: %i. Num flashes = %i. \n', trl, thisNumFlashes) storeSuccess(ti) = 0; end end if ~hasPrintedOnce, fprintf('Num Flashes Detected: %i. Num storeXlocs: %i. \n', length(timeFlashesStarts), length(storeXlocs)); hasPrintedOnce = true; end %% find every spike and determine what the frame was some time before it % screen resolution 1024x768 ycent = 384; % the yaxis is flipped storeYlocs = -(storeYlocs - ycent); numavgs = 1; for spi = 1:length(spktimes) thisspiketime = spktimes(spi) - window; idxFS = find(timeFlashesStarts < thisspiketime, 1, 'last'); if ~isempty(idxFS) && timeFlashesStarts(idxFS) < thisspiketime && timeFlashesEnds(idxFS) > thisspiketime frmat = frmat.*numavgs; for nf = 1:size(storeXlocs,2) % find out which row/col it goes into and add the appropriate value row = find(hbins > storeXlocs(idxFS,nf), 1, 'first'); col = find(vbins > storeYlocs(idxFS,nf), 1, 'first'); % add 1 to the location of where the stimulus was frmat(row,col) = frmat(row,col) + 1; end% nf % recompute average numavgs = numavgs + 1; frmat = frmat./numavgs; end end end clf, hold on figure(1), heatmap(frmat'); % the data is averaged on the go axis([0.5 xdiv 0.5 ydiv]) title(['Time Back: ' num2str(allwindow(wi))]) colorbar ax = axis; line(ax(1:2),[mean(ax(3:4)) mean(ax(3:4))], 'LineStyle', '--','LineWidth', 2, 'Color', 'k') line([mean(ax(1:2)) mean(ax(1:2))], ax(3:4), 'LineStyle', '--','LineWidth', 2, 'Color', 'k') drawnow end
% Extract pixel pairs from both HSI and RGB images manually and let the % algorithm to pick up some random pixels for manifold alignment. if ~exist('hsiToRgbManifoldDatasetGen', 'var') hsiToRgbManifoldDatasetGen = false; end if hsiToRgbManifoldDatasetGen fb_HSI_Image = RotateHsiImage(hsiCubeData.DataCube, -90); rgbFromHsi = ConstructRgbImage(fb_HSI_Image, 2, 3, 4); else fb_HSI_Image = RotateHsiImage(reflectanceCube.DataCube, -90); rgbFromHsi = GetTriBandRgbImage(fb_HSI_Image); end classesInHsi = 7; noOfSample = 10; % Positions in [height, width] format. extractedHsiPixels = zeros(noOfSample * classesInHsi, 2); extractedHsiVector = zeros(noOfSample, 2); extractedRgbPixels = zeros(noOfSample * classesInHsi, 2); extractedRgbVector = zeros(noOfSample, 2); for nId = 1: classesInHsi [topLeft, btmRight] = PickUpPixelArea(rgbFromHsi); H = (btmRight(1) - topLeft(1)); W = (btmRight(2) - topLeft(2)); totalSamples = H * W; coordList = zeros(totalSamples, 2); for nX = 1:H for nY = 1:W coordList(((nX - 1) * W) + nY, :) = [(topLeft(1) - 1) + nX, (topLeft(2) - 1) + nY]; end end % Randomize the data and pick n number of samples. randIds = randperm(totalSamples); for px = 1: noOfSample extractedHsiVector(px, :) = coordList(randIds(px), :); end extractedHsiPixels((((nId - 1) * noOfSample) + 1): (((nId - 1) * noOfSample) + noOfSample), :) = extractedHsiVector(:,:); % end %% RGB % Positions in [height, width] format. % for nId = 1: classesInHsi [topLeft, btmRight] = PickUpPixelArea(higResRgb); H = (btmRight(1) - topLeft(1)); W = (btmRight(2) - topLeft(2)); totalSamples = H * W; coordList = zeros(totalSamples, 2); for nX = 1:H for nY = 1:W coordList(((nX - 1) * W) + nY, :) = [(topLeft(1) - 1) + nX, (topLeft(2) - 1) + nY]; end end % Randomize the data and pick n number of samples. randIds = randperm(totalSamples); for px = 1: noOfSample extractedRgbVector(px, :) = coordList(randIds(px), :); end extractedRgbPixels((((nId - 1) * noOfSample) + 1): (((nId - 1) * noOfSample) + noOfSample), :) = extractedRgbVector(:,:); close all; end %% Clear variables vars = {'noOfSample', 'nId', 'btmRight', 'classesInHsi', 'coordList',... 'randIds', 'px', 'nX', 'nY', 'H', 'W', 'topLeft', 'totalSamples'}; clear(vars{:});
function [ out ] = nonan_Image( img ) out = zeros(size(img)); for k=1:size(img,3) for i=1:size(img,1) for j=1:size(img,2) if ( isnan(img(i,j,k))) out(i,j,k)=0; else out(i,j,k) = img(i,j,k); end end end end end
function RunBatNew( filename ) %RUNBAT runs a batch file via Windows start menu using Java Robot class %Input: % <filename> batch file (.bat) %CL % Get Windows version winvers = system_dependent('getos'); % Create java robot to generate native system input events import java.awt.Robot; import java.awt.event.*; robot = Robot(); % Open the windows start menu % scrnsize = get(0,'screensize'); % robot.mouseMove(0,scrnsize(4)); % robot.mousePress(InputEvent.BUTTON1_MASK); % robot.mouseRelease(InputEvent.BUTTON1_MASK); % Open Win+R (Run) window instead robot.keyPress(KeyEvent.VK_WINDOWS) robot.keyPress(KeyEvent.VK_R) robot.keyRelease(KeyEvent.VK_WINDOWS) % Execute the batch file by typing the filename if ~isempty(strfind(winvers,'10')) % For Windows 10, batch execution via 'run' pause(1) lang = get(0,'Language'); if strcmp('de',lang(1:2)) typestr('Ausführen') else typestr('run') end robot.keyPress(KeyEvent.VK_ENTER) pause(1) % Type filename typestr(filename) robot.keyPress(KeyEvent.VK_ENTER) else % For Windows 7 and 8, batch execution via 'search' typestr(filename) robot.keyPress(KeyEvent.VK_ENTER) end end function typestr( str ) %TYPESTRING types a given string on the keyboard using the java robot % Create java robot to generate native system input events import java.awt.Robot; import java.awt.event.*; robot = Robot(); pause(0.1) for i = 1:length(str) c = str(i); if isstrprop(c,'punct') % Punctuation character switch c case '-' robot.keyPress(KeyEvent.VK_MINUS) case '.' robot.keyPress(KeyEvent.VK_PERIOD) case ':' robot.keyPress(KeyEvent.VK_SHIFT) robot.keyPress(KeyEvent.VK_PERIOD) robot.keyRelease(KeyEvent.VK_SHIFT) case '\' % ASCII value robot.keyPress(KeyEvent.VK_ALT); robot.keyPress(KeyEvent.VK_NUMPAD9); robot.keyPress(KeyEvent.VK_NUMPAD2); robot.keyRelease(KeyEvent.VK_ALT); end elseif isstrprop(c,'wspace') % Space robot.keyPress(KeyEvent.VK_SPACE); elseif isstrprop(c,'digit') % Number eval(strcat('robot.keyPress(KeyEvent.VK_NUMPAD',c,')')); elseif strcmp(c,'ü') robot.keyPress(KeyEvent.VK_ALT); robot.keyPress(KeyEvent.VK_NUMPAD1); robot.keyPress(KeyEvent.VK_NUMPAD5); robot.keyPress(KeyEvent.VK_NUMPAD4); robot.keyRelease(KeyEvent.VK_ALT); elseif strcmp(c,'ö') robot.keyPress(KeyEvent.VK_ALT); robot.keyPress(KeyEvent.VK_NUMPAD1); robot.keyPress(KeyEvent.VK_NUMPAD5); robot.keyPress(KeyEvent.VK_NUMPAD3); robot.keyRelease(KeyEvent.VK_ALT); elseif strcmp(c,'ä') robot.keyPress(KeyEvent.VK_ALT); robot.keyPress(KeyEvent.VK_NUMPAD1); robot.keyPress(KeyEvent.VK_NUMPAD4); robot.keyPress(KeyEvent.VK_NUMPAD2); robot.keyRelease(KeyEvent.VK_ALT); else c = upper(c); % Upper case character eval(strcat('robot.keyPress(KeyEvent.VK_',c,')')); end pause(0.01) end end
clear ; %clears all the variables close all; %closes extra windows clc %clears the screen % =============================================================== Part 1: K MEANS ====================================================== % ----------------------------------------------------------(a)---------------------------------------------------------------- X = load('attr.txt'); y = load('label.txt'); [m n] = size(X); K = 6; %no. of clusters tic idx = randperm(m, K); %generates unique K random integers between 1 to m inclusive. final_cost = k_means(X, K, idx) toc % ----------------------------------------------------------(b)---------------------------------------------------------------- tic J = zeros(10,1); min_cost_idx = zeros(1,K); min_cost = inf; for i = 1 : 10 i idx = randperm(m, K); J(i) = k_means(X, K, idx) if(J(i) < min_cost) min_cost = J(i); min_cost_idx = idx; end end J min_cost toc mu = X(min_cost_idx,:); %taking K random examples as the initial mu J = zeros(60,1); tic for i = 1 : 60 c = best_cluster(X, K, mu); %Expectation Step mu = update_means(X, K, c); %Minimization Step J(i) = computeErrorMetric(X, c, mu); %finding the cost end toc x = [1:60]; figure; plot(x, J, '-', 'LineWidth', 0.7) ylabel('Error'); % Set the y axis label xlabel('No. of iterations'); % Set the x axis label % ----------------------------------------------------------(c)---------------------------------------------------------------- mu = X(min_cost_idx,:); accuracy_vector = zeros(60,1); c_tmp = zeros(m, 1); for iter = 1 : 60 c = best_cluster(X, K, mu); %Expectation Step mu = update_means(X, K, c); %Minimization Step for i = 1 : K idx = find(c == i); c_tmp(idx) = mode(y(idx)); end accuracy_vector(iter) = length(find(c_tmp == y)) * 100 / m ; end figure; plot(x, accuracy_vector, '-', 'LineWidth', 0.7) ylabel('Accuracy'); % Set the y axis label xlabel('No. of iterations'); % Set the x axis label
function [ X, interactionX ] = createAgeGenderSizeRegressionMatrix(groupVector,polynomialDims,varargin) numCovars=length(varargin); % Create a balanced groupVector positiveGroup=find(groupVector > 0); numPositive=length(positiveGroup); negativeGroup=find(groupVector < 0); numNegative=length(negativeGroup); groupVector(positiveGroup)=sum(groupVector(positiveGroup))/numPositive*(numPositive/numNegative); groupVector(negativeGroup)=sum(groupVector(negativeGroup))/numNegative*(numNegative/numPositive); % Create polynomial expansions of the base covariates for c=1:numCovars baseCovar=cell2mat(varargin(c)); subX(1,:)=baseCovar-mean(baseCovar); if polynomialDims(c)>1 for p=2:polynomialDims(c) polyX=subX(1,:).^p; polyX=polyX-mean(polyX); polyX=polyX/max(polyX); [~,~,stats]=glmfit(subX',polyX'); subX(p,:)=stats.resid; end end if c==1 X=subX; else X=[X;subX]; end clear subX end % Create a matrix to test for interactions interactionX=X*0; % Multiply main effect covariates by the group vector to % create the interaction terms for i=1:size(X,1) interactionX(i,:)=X(i,:).*groupVector'; end % Render the interaction terms orthogonal to the main effect for i=1:size(X,1) interaction=interactionX(i,:); mainEffect=X(i,:); [~,~,stats]=glmfit(mainEffect,interaction); interactionX(i,:)=stats.resid; end % detect any interaction terms that are all zeros and remove them goodColumns=ones(size(X,1),1); for i=1:size(X,1) if (max(interactionX(i,:))-min(interactionX(i,:)))<0.01 goodColumns(i)=0; end end interactionX=interactionX(find(goodColumns==1),:); gribble=1; end
function interactive_graph_gui % data showLabels = true; % flag to determine whether to show node labels prevIdx = []; % keeps track of 1st node clicked in creating edges selectIdx = []; % used to highlight node selected in listbox pts = zeros(0,2); % x/y coordinates of vertices adj = sparse([]); % sparse adjacency matrix (undirected) edd=sparse([]); MinXLim=0; MaxXLim=100; MinYLim=0; MaxYLim=100; % create GUI h = initGUI(); function h = initGUI() h.fig = figure('Name','Interactive Graph', 'Resize','off'); h.ax = axes('Parent',h.fig, 'ButtonDownFcn',@onMouseDown, ... 'XLim',[MinXLim MaxXLim], 'YLim',[MinYLim MaxYLim], 'XTick',[], 'YTick',[], 'Box','on', ... 'Units','pixels', 'Position',[160 20 380 380]); h.list = uicontrol('Style','listbox', 'Parent',h.fig, 'String',{}, ... 'Min',1, 'Max',1, 'Value',1, ... 'Position',[20 80 130 320], 'Callback',@onSelect); uicontrol('Style','pushbutton', 'Parent',h.fig, 'String','Clear', ... 'Position',[20 20 60 20], 'Callback',@onClear); uicontrol('Style','pushbutton', 'Parent',h.fig, 'String','Export', ... 'Position',[90 20 60 20], 'Callback',@onExport); uicontrol('Style','pushbutton', 'Parent',h.fig, 'String','Delete', ... 'Position',[50 50 60 20], 'Callback',@onDelete); h.cmenu = uicontextmenu('Parent',h.fig); h.menu = uimenu(h.cmenu, 'Label','Show labels', 'Checked','off', ... 'Callback',@onCMenu); set(h.list, 'UIContextMenu',h.cmenu) h.pts = line(NaN, NaN, 'Parent',h.ax, 'HitTest','off', ... 'Marker','o', 'MarkerSize',10, 'MarkerFaceColor','b', ... 'LineStyle','none'); h.selected = line(NaN, NaN, 'Parent',h.ax, 'HitTest','off', ... 'Marker','o', 'MarkerSize',10, 'MarkerFaceColor','y', ... 'LineStyle','none'); h.prev = line(NaN, NaN, 'Parent',h.ax, 'HitTest','off', ... 'Marker','o', 'MarkerSize',20, 'Color','r', ... 'LineStyle','none', 'LineWidth',2); h.edges = line(NaN, NaN, 'Parent',h.ax, 'HitTest','off', ... 'LineWidth',2, 'Color','g'); h.txt1 = []; h.txt2 = []; end function onMouseDown(~,~) % get location of mouse click (in data coordinates) p = get(h.ax, 'CurrentPoint'); % determine whether normal left click was used or otherwise if strcmpi(get(h.fig,'SelectionType'), 'Normal') % add a new node pts(end+1,:) = p(1,1:2); adj(end+1,end+1) = 0; else % add a new edge (requires at least 2 nodes) if size(pts,1) < 2, return; end % hit test (find node closest to click location: euclidean distnce) [dst,idx] = min(sum(bsxfun(@minus, pts, p(1,1:2)).^2,2)); % if sqrt(dst) > 0.025, return; end if isempty(prevIdx) % starting node (requires a second click to finish) prevIdx = idx; else % add the new edge adj(prevIdx,idx) =sqrt(((pts(prevIdx,1)-pts(idx,1))^2)+((pts(prevIdx,2)-pts(idx,2))^2)); prevIdx = []; end end % update GUI selectIdx = []; redraw() end function onDelete(~,~) % check that list of nodes is not empty if isempty(pts), return; end % delete selected node idx = get(h.list, 'Value'); pts(idx,:) = []; adj(:,idx) = []; adj(idx,:) = []; % clear previous selections if prevIdx == idx prevIdx = []; end selectIdx = []; % update GUI set(h.list, 'Value',max(min(idx,size(pts,1)),1)) redraw() end function onClear(~,~) % reset everything prevIdx = []; selectIdx = []; pts = zeros(0,2); adj = sparse([]); % update GUI set(h.list, 'Value',1) redraw() end function onExport(~,~) % export nodes and adjacency matrix to base workspace assignin('base', 'adj',adj+adj') % make it symmetric assignin('base', 'xy',pts) end function onSelect(~,~) % update index of currently selected node selectIdx = get(h.list, 'Value'); redraw() end function onCMenu(~,~) % flip state showLabels = ~showLabels; redraw() end function redraw() % edges p = nan(3*nnz(adj),2); [i,j] = find(adj); p(1:3:end,:) = pts(i,:); p(2:3:end,:) = pts(j,:); set(h.edges, 'XData',p(:,1), 'YData',p(:,2)) % nodes set(h.pts, 'XData',pts(:,1), 'YData',pts(:,2)) set(h.prev, 'XData',pts(prevIdx,1), 'YData',pts(prevIdx,2)) set(h.selected, 'XData',pts(selectIdx,1), 'YData',pts(selectIdx,2)) % list of nodes set(h.list, 'String',num2str(pts,'(%.3f,%.3f)')) % node labels if ishghandle(h.txt1) delete(h.txt1); end if ishghandle(h.txt2) delete(h.txt2); end if showLabels set(h.menu, 'Checked','on') h.txt1 = text(pts(:,1)+0.01, pts(:,2)+0.01, ... num2str((1:size(pts,1))'), ... 'HitTest','off', 'FontSize',8, ... 'VerticalAlign','bottom', 'HorizontalAlign','left'); else set(h.menu, 'Checked','off') end if ~isempty(find(adj, 1)) [xi,yi,val]=find(adj); h.txt2 = text((pts(xi,1)+pts(yi,1))/2, (pts(xi,2)+pts(yi,2))/2, ... num2str( val), ... 'HitTest','off', 'FontSize',8, ... 'VerticalAlign','bottom', 'HorizontalAlign','left'); end % force refresh drawnow end end
clear; close all; clc; %% Threshhold, Adaptive Equal % I = imread('fingerprint.png'); % I = double(I); % fingim = otsu(I); % figure('Name','Adaptive Equalization, Fingerprint'); % subplot(1,3,1);imagesc(I); title('Original Image'); % subplot(1,3,2);imagesc(fingim); colormap gray; title('Otsu Threshold'); % adt = AdaptiveEqualize(I); % otAdt = otsu(adt); % subplot(1,3,3); imagesc(otAdt);colormap gray; title('Otsu after Adaptive Equalizatoin'); % %% Gaussain and Bilateral % part a,b,c,d I = imread('lena.png'); I = double(I); subplot(2,3,1);imagesc(I); colormap gray; title('Original Im'); G = GaussianFilter(I,1); subplot(2,3,2);imagesc(G); colormap gray; title('GaussianFltr sigma = 3'); B = BilateralFilter(I,3,10); subplot(2,3,3);imagesc(B); colormap gray; title('BilateralFltr sd = 3 sr = 10'); dG = abs(I-G); subplot(2,3,5);imagesc(dG); colormap gray; title('Original-Gaussian'); dB = abs(I-B); sumdb = sum(sum(dB)); subplot(2,3,6); imagesc(dB); colormap gray;title('Original-Bilateral'); %% Part e I = imread('lena.png'); I = double(I); figure(); subplot(1,3,1);imagesc(I); colormap gray; title('Original Im'); B = BilateralFilter(I,3,100); subplot(1,3,2);imagesc(B); colormap gray; title('BilateralFltr sd = 3 sr = 100'); dB = abs(I-B); sumdb = sum(sum(dB)); subplot(1,3,3); imagesc(dB); colormap gray;title('Original-Bilateral'); %% part f and g % I = imread('lena.png'); % I = double(I); % figure(); % subplot(1,3,1);imagesc(I); colormap gray; title('Original Im'); % % B = BilateralFilter(I,1000,10); % subplot(1,3,2);imagesc(B); colormap gray; title('BilateralFltr sd = 1000 sr = 10'); % % B = BilateralFilter(I,3,1000); % subplot(1,3,3);imagesc(B); colormap gray; title('BilateralFltr sd = 3 sr = 1000'); %% House.png % f = imread('house.png'); % f = double(f); % g = f+20*randn(size(f)); % % G = GaussianFilter(f,1); % mn = size(f,1)*size(f,2); % MSEf = sum(sum(minus(f,G).^2))/mn; % MSEg = sum(sum(minus(f,g).^2))/mn; % mseNoise = sprintf('Noise; MSE = %f',MSEg); % mseFilter = sprintf('GaussianFlter sgm =3; MSE = %f',MSEf); % % subplot(1,3,1);imagesc(f); colormap gray; title('Original'); % subplot(1,3,2);imagesc(g); colormap gray; title(mseNoise); % subplot(1,3,3);imagesc(G); colormap gray; title(mseFilter); % % figure(); % mse = [0,0,0,0,0,0]; % x = [5,10,20,40,60,100]; % B1 = BilateralFilter(g,3,5); % mse(1) = 1./mn.*sum(sum(minus(f,B1).^2)); % mseFilter1 = sprintf('sgm =5; MSE = %f',mse(1)); % subplot(3,2,1);imagesc(B1); colormap gray;title(mseFilter1); % % B2 = BilateralFilter(g,3,10); % mse(2) = 1./mn.*sum(sum(minus(f,B2).^2)); % mseFilter2 = sprintf('sgm =10; MSE = %f',mse(2)); % subplot(3,2,2);imagesc(B2); colormap gray;title(mseFilter2); % % B3 = BilateralFilter(g,3,20); % mse(3) = 1./mn.*sum(sum(minus(f,B3).^2)); % mseFilter3 = sprintf('sgm =20; MSE = %f',mse(3)); % subplot(3,2,3);imagesc(B3); colormap gray;title(mseFilter3); % % B4 = BilateralFilter(g,3,40); % mse(4) = 1./mn.*sum(sum(minus(f,B4).^2)); % mseFilter4 = sprintf('sgm =40; MSE = %f',mse(4)); % subplot(3,2,4);imagesc(B4); colormap gray;title(mseFilter4); % % B5 = BilateralFilter(g,3,60); % mse(5) = 1./mn.*sum(sum(minus(f,B5).^2)); % mseFilter5 = sprintf('sgm =60; MSE = %f',mse(5)); % subplot(3,2,5);imagesc(B5); colormap gray;title(mseFilter5); % % B6 = BilateralFilter(g,3,100); % mse(6) = 1./mn.*sum(sum(minus(f,B6).^2)); % mseFilter6 = sprintf('sgm =100; MSE = %f',mse(6)); % subplot(3,2,6);imagesc(B6); colormap gray;title(mseFilter6); % % figure();plot(x,mse);
%time_score driver % sets paths to kml files and a wrfout, then runs the time_score function kml_path = '/bigdisk/james.haley/wrfcycling/wrf-fire/wrfv2_fire/test/TIFs/doc.kml'; wrfout_path = '/bigdisk/james.haley/wrfcycling/wrf-fire/wrfv2_fire/test/cycling_best_ig_single/wrfout_d01_2013-08-13_00:00:00'; time_score(kml_path,wrfout_path)
function d = mismatch( h1, h2 ) % MISMATCH computes the mismatch distance between the two histograms d = sum ( h1 ~= h2 ); end
clear clc close all A = tf([1.3],[1,1.3]) G = tf([1],[.00303,0,0]) H = tf([1000],[1,1000]) GOL = A*G*H; bode(GOL) margin(GOL) figure() rlocus(GOL) C1 = tf([1,1],[1,650]) C2 = tf([1,5],[1,300]) C = C1*C2; CGOL = C*GOL bode(CGOL) margin(CGOL) figure() rlocus(CGOL) K = 5; GCL = K*CGOL/(1+K*CGOL); figure() step(GCL) figure() pzmap(GCL)
function mtlscale(infile, outfile, factor) %function mtlscale(infile, outfile, factor) [file_type,info_blocks,n_channels,n_lines, sampling_rate,... first_line,last_line,n_directions,comment1, comment2] = mtlrh(infile) if (file_type ~= 2) error('Wrong filetype'); return; end; if (nargin < 3) factor=1; end; index=0; %if direction_matrix if ((info_blocks-2) > 0) direction_matrix=mtlrdir(infile,n_directions); if (index ~= 0) direction_matrix(:) = direction_matrix(:,index); end; % if time_delay exists read time_delay if (( (info_blocks-2)*256 - n_directions*8) == (n_channels*4)) [delay_l,delay_r]=mtlrdel(infile,n_directions,binaural); if (index ~= 0) delay_l = delay_l(index); if (binaural == 1) delay_r = delay_r(index); end; end; end; end; % write output_header mtlwh(outfile,file_type,info_blocks,n_channels,n_lines, sampling_rate,... first_line,last_line,n_directions,... comment1, comment2); % write directions; % write delays; if ((n_channels/n_directions) == 2) binaural=1; end; n_dir=n_directions; if (n_dir > 0) n_byte=n_dir*2*4; sn_del=length(delay_l); if (sn_del > 0) n_byte=n_byte+n_dir*4*(binaural+1); end; n_blocks=ceil(n_byte/256); end; info_blocks=2+n_blocks; comment1 = [comment1 '; after scaling']; mtlwh(outfile,file_type,info_blocks,n_channels,n_lines, sampling_rate,... first_line,last_line,n_directions,... comment1, comment2); [fid,message] = fopen(outfile,'a','ieee-le'); if fid==-1 % file can't be opened disp(message); return; end; if ((n_blocks) > 0) fseek(fid,512,'bof'); fwrite(fid,zeros(n_blocks*256,1),'float32'); fclose(fid); end %if direction_matrix if ((n_blocks) > 0) status=mtlwdir(outfile,direction_matrix) if(length(delay_l) > 0) mtlwdel(outfile, delay_l,delay_r); end; end for n=1:n_channels clc disp(['MTLSCALE: channel ', num2str(n), ' of ', num2str(n_channels)] ); channel=mtlrch(infile,n); status = mtlwch(outfile,channel*factor,n) end; fclose(fid);
function result = xcat(dim,varargin) % Array concatenation with singleton expansion. % Author: Vladimir Golkov if ~isscalar(dim) && ~isempty(dim) % dim not provided, first argument is array varargin(2:end+1) = varargin; % TODO is this fast? varargin{1} = dim; end sizelist = []; for i=1:numel(varargin) siz = size(varargin{i}); sizelist(:,end+1 : numel(siz)) = 1; % in case ndims(varargin{i}) is larger than largest previous ndims sizelist(i,1:numel(siz)) = siz; sizelist(i,numel(siz)+1 : end) = 1; % in case ndims(varargin{i}) is smaller than largest previous ndims end if isempty(dim) dim = size(sizelist,2)+1; % determine dim automatically: choose next higher dimension end sizelist(:,end+1 : dim) = 1; % in case every ndims(varargin{i}) is smaller than concatenation dimension resultsize = max(sizelist,[],1); resultsize(dim) = sum(sizelist(:,dim)); e_dim = false(1,size(resultsize,2)); e_dim(dim) = true; % disregard concatenation dimension when checking dimension consistency if ~all(all(bsxfun(@eq,sizelist,resultsize) | sizelist==1) | e_dim) error('xcat:InsonsistentDimensions','Inconsistent dimensions. For each dimension, array sizes must either match or be equal to one.'); end for i=1:numel(varargin) newsize = resultsize./sizelist(i,:); newsize(dim) = 1; % don't repeat along concatenation dimension varargin{i} = repmat(varargin{i},newsize); % TODO alternative to save working memory: in each loop iteration, do repmat, cat, discard end % prevent "Error using cat: Dimension for sparse matrix concatenation must be <= 2." if dim>2 && any(cellfun(@issparse,varargin)) warning('Converting sparse to full...') varargin = full(varargin); % this uses @cell/full.m end result = cat(dim,varargin{:});
function [T] = makebins(Set,Y,bins) % MAKEBINS divides a given data set into cell bins % -Set is the Data set (usually consisting of alignment identity score % values and their respective sequences % -Y is the result of running MATLAB's discretize function on Set % -bins is the number of bins that the data will be divided into T=cell(bins,1); for i=1:size(Y,1) num=Y(i,1); elt=Set(i,:); T{num,1}=[T{num,1}; elt]; end end
% testing the stereo door corner camera addpath(genpath('../utils/pyr')); addpath(genpath('../corner_cam')); addpath(genpath('../stereo_cam')); close all; clear; npx = 80; width = npx/10; floornpx = 80; nsamples = 80; % door is centered at origin door_corner1 = [-1 0 0]; door_corner2 = [1 0 0]; door_width = abs(door_corner2(1) - door_corner1(1)); % we look at the shadow at the two corners of the door xlocs1 = linspace(door_corner1(1), door_corner1(1)-1, floornpx); xlocs2 = linspace(door_corner2(1), door_corner2(1)+1, floornpx); ylocs = linspace(0, 1, floornpx); % images behind the door imdepths = [-5, -7]; imstarts = [2*width, 7*width]; nims = length(imstarts); imgs = zeros([npx, npx, nims]); scenex = linspace(door_corner1(1), door_corner2(1), npx); sceneys = zeros([nims, npx]); scenez = linspace(0, door_corner2(1), npx); gt_depths = zeros(size(scenex)); obs1 = zeros([floornpx*floornpx, 1]); obs2 = zeros(size(obs1)); figure; for i = 1:nims start = imstarts(i); imgs(:,start:start+width,i) = 1; sceneys(i,:) = ones(size(scenex)) * imdepths(i); subplot(nims,1,i); imagesc(imgs(:,:,i)); axis off; title(sprintf('%d units depth', -imdepths(i))); gt_depths = gt_depths + abs(imgs(1,:,i) .* sceneys(i,:)); % get observations amat1 = cornerAmat(xlocs1, ylocs, scenex, sceneys(i,:)); obs1 = obs1 + amat1 * reshape(imgs(:,:,i), [npx*npx,1]); amat2 = cornerAmat(xlocs2, ylocs, scenex, sceneys(i,:)); obs2 = obs2 + amat2 * reshape(imgs(:,:,i), [npx*npx,1]); end sigma = 0.5; beta = 100; obs1_noisy = obs1 + sigma * randn(size(obs1)); obs2_noisy = obs2 + sigma * randn(size(obs1)); % plotting the shadow observations figure; subplot(211); imagesc(xlocs1, ylocs, reshape(obs1_noisy, [floornpx, floornpx])); subplot(212); imagesc(xlocs2, ylocs, reshape(obs2_noisy, [floornpx, floornpx])); % reconstructing a 1d picture from each door corner thetas1 = [pi, pi/2]; tdir1 = sign(thetas1(2) - thetas1(1)); [amat1_1d, ~, ~] = allPixelAmat(door_corner1, floornpx, nsamples, thetas1); angles1 = linspace(thetas1(1), thetas1(2), nsamples) + tdir1*pi; thetas2 = [0, pi/2]; tdir2 = sign(thetas2(2) - thetas2(1)); [amat2_1d, ~, ~] = allPixelAmat(door_corner2, floornpx, nsamples, thetas2); angles2 = linspace(thetas2(1), thetas2(2), nsamples) + tdir2*pi; % spatial prior bmat = eye(nsamples) - diag(ones([nsamples-1,1]), 1); x1_1d = (amat1_1d'*amat1_1d/sigma^2 + bmat'*bmat*beta)\(amat1_1d'*obs1_noisy/sigma^2); x2_1d = (amat2_1d'*amat2_1d/sigma^2 + bmat'*bmat*beta)\(amat2_1d'*obs2_noisy/sigma^2); % x1_1d(x1_1d<0) = 0; % x1_1d(x1_1d>1) = 1; % x2_1d(x2_1d<0) = 0; % x2_1d(x2_1d>1) = 1; x1_1d = (x1_1d - min(x1_1d))/(max(x1_1d) - min(x1_1d)); x2_1d = (x2_1d - min(x2_1d))/(max(x2_1d) - min(x2_1d)); figure; subplot(211); imagesc(angles1, 1:npx/2, repmat(x1_1d', [npx/2, 1])); subplot(212); imagesc(angles2, 1:npx/2, repmat(x2_1d', [npx/2, 1])); % estimate depth % need to flip one of them; reverse orientations x1_1d = x1_1d(end:-1:1); angles1 = angles1(end:-1:1); % for every window in x1_1d, we find the best matching window in x2_1d winsize = 2; energy = zeros(nsamples-winsize); for i = 1:nsamples-winsize win1 = x1_1d(i:i+winsize); for j = 1:nsamples-winsize win2 = x2_1d(j:j+winsize); % energy(i,j) = -sum((win2-mean(win2)).*(win1-mean(win1)))/(std(win1)*std(win2)); energy(i,j) = sum((win2-win1).^2); end end figure; imagesc(energy); title('energy between two reconstructions'); % the corresponding angles from the door corners, for each point % only positive angles for this calculation door_angle1 = abs(angles1(1:nsamples-winsize)); depths = zeros(size(door_angle1)); for i = 1:length(depths) row = energy(i,:); [val, idx] = min(row); secondbest = row(row > val); val2 = min(secondbest); if val/val2 > 0.6 % too close together, assume we're not resolving to an actual object depths(i) = 1e-2; else % take the angle of corner 2 of the best match a2 = angles2(idx) - tdir2*pi; a1 = door_angle1(i); depths(i) = cot(a1) + cot(a2); end end depths = abs(door_corner1(1) - door_corner2(1)) ./ depths; locs1 = cot(door_angle1) .* depths - 1; angle_ends = abs(angles1(1+winsize:nsamples)); locs2 = cot(angle_ends) .* depths - 1; % [minval, minidx] = min(energy, [], 2); % door_angle2 = angles2(minidx) - tdir2*pi; % for each point, compute depth % depths = abs(door_corner1(1) - door_corner2(1))./(cot(door_angle1) + cot(door_angle2)); figure; subplot(211); plot(locs1(depths < 200), depths(depths < 200), 'ro'); hold on; plot(locs2(depths < 200), depths(depths < 200), 'bo'); xlim([-1, 1]); title('preliminary depth estimation of each point'); subplot(212); plot(scenex(gt_depths > 0), gt_depths(gt_depths > 0), 'bo'); xlim([-1, 1]); title('ground truth depths');
function LoadGamma(CLUT) physicalDisplay = 1; Display.ScreenID = 0; [OriginalGammaTable, dacbits, reallutsize] = Screen('ReadNormalizedGammaTable', Display.ScreenID);%, physicalDisplay); save('OriginalSetup3CLUT.mat', 'OriginalGammaTable'); CLUT = 'NIH_Setup3_ASUS_V27.mat'; load(CLUT); AllScreens = Screen('Screens'); Stereomode = 4; Text.Size = 60; Eye = {'Left buffer (0)','Right buffer (1)'}; %================== IDENTIFY LEFT/RIGHT SCREENS AND BUFFERS =============== [win Rect] = Screen('OpenWindow',0, 128, [],[],[],Stereomode); for s = 1:2 Screen('SelectStereoDrawBuffer', win, s-1); % 0 = left eye buffer; 1 = right eye buffer Screen('TextSize', win, Text.Size); DrawFormattedText(win, Eye{s}, 'center', 'center', [0 0 0], []); Screen('LoadNormalizedGammaTable', AllScreens, inverseCLUT{s}); end Screen('Flip',win); KbWait; sca; %================================== LOAD CLUTS ============================ % for i=1:numel(Lum) % fprintf('CLUT %d for %s\n', i, Lum(i).DisplayName); % end % for Display = 1:2 % Screen('SelectStereoDrawBuffer', AllScreens, Display-1); % Screen('LoadNormalizedGammaTable', AllScreens, inverseCLUT{Display}); % Load the original gamma table % end
% min_angle = minAngle(angles_array, indices_of_min_energy) % This function extracts the angles corresponding to minima in energy; function min_angle = minAngle(energies, indices) min_angle = angles_array(indices_of_min_energy); end
close all clear all clc %% code_folder = '/Users/akira/Documents/GitHub/MN-Model/Test Size'; data_folder = '/Users/akira/Documents/GitHub/MN-Model/MN Parameter'; %% type = 'S'; %% model_folder = '/Users/akira/Documents/GitHub/Twitch-Based-Muscle-Model/Model Parameters/Model_8'; cd(model_folder) load('modelParameter') cd(code_folder) [val,loc] = min(modelParameter.U_th); testMN = loc; MDR_d = modelParameter.MDR(testMN); PDR_d = modelParameter.PDR(testMN); I_max = 100; I_th = 100*val; slope_target = (PDR_d-MDR_d)/(I_max-I_th); %% % find a template MN whose minimal discharge rate is closest to the target % minimal discharge rate cd(data_folder) load([type '_MDR']) cd(code_folder) [~,loc_min] = min(abs(MDR_d-MDR)); nMN = num2str(loc_min); cd(data_folder) load([type '_' nMN]) cd(code_folder) %% count = 1; perturbation_amp = 0.3; %% % Fit the sloe of the f-I relationship by adjusting parameter g_Ks for j = 1:5 j amp_vec = 0:1:100; mean_FR = zeros(1,length(amp_vec)); CoV_FR = zeros(1,length(amp_vec)); for i = 1:length(amp_vec) Fs = 10000; time = 0:1/Fs:5; noise_amp = 0; amp = amp_vec(i); input = zeros(1,length(time)); input(1*Fs+1:end) = amp; inputOpt = 1; pltOpt = 0; %% [binary,V_s,V_d] = Cisi2008_function(param,time,input,Fs,noise_amp,inputOpt,pltOpt); spike_time = find(binary(end-2*Fs+1:end)); ISI = diff(spike_time)/(Fs/1000); mean_FR(i) = mean(1./ISI*1000); CoV_FR(i) = std(1./ISI*1000)/mean_FR(i)*100; end %% index_t = find(~isnan(mean_FR)); amp_th = amp_vec(index_t(1)); min_DR = mean_FR(index_t(1)); p_fit = polyfit(amp_vec(index_t),mean_FR(index_t),1); y1 = polyval(p_fit,amp_vec(index_t)); error = slope_target - p_fit(1); if error > 0 param.g_Ks = param.g_Ks - (param.g_Ks*perturbation_amp)./2.^(count-2); else param.g_Ks = param.g_Ks + (param.g_Ks*perturbation_amp)./2.^(count-2); end count = count + 1; end %% % Fit the recruitment threshold by adjusting parameter I_r mean_FR = zeros(1,length(amp_vec)); CoV_FR = zeros(1,length(amp_vec)); perturbation_amp = 0.3; count = 1; for j = 1:5 j amp_vec = 0:0.1:50; mean_FR = zeros(1,length(amp_vec)); CoV_FR = zeros(1,length(amp_vec)); error = I_th - amp_th; if error > 0 param.I_r = param.I_r + (param.I_r*perturbation_amp)./2.^(count-2); else param.I_r = param.I_r - (param.I_r*perturbation_amp)./2.^(count-2); end for i = 1:length(amp_vec) Fs = 10000; time = 0:1/Fs:5; noise_amp = 0; amp = amp_vec(i); input = zeros(1,length(time)); input(1*Fs+1:end) = amp; inputOpt = 1; pltOpt = 0; [binary,V_s,V_d] = Cisi2008_function(param,time,input,Fs,noise_amp,inputOpt,pltOpt); spike_time = find(binary(end-2*Fs+1:end)); ISI = diff(spike_time)/(Fs/1000); mean_FR(i) = mean(1./ISI*1000); CoV_FR(i) = std(1./ISI*1000)/mean_FR(i)*100; end index_t = find(~isnan(mean_FR)); amp_th = amp_vec(index_t(1)); min_DR = mean_FR(index_t(1)); p_fit = polyfit(amp_vec(index_t),mean_FR(index_t),1); y1 = polyval(p_fit,amp_vec(index_t)); count = count + 1; end %% Plot the result amp_vec = 0:0.1:100; mean_FR = zeros(1,length(amp_vec)); CoV_FR = zeros(1,length(amp_vec)); for i = 1:length(amp_vec) Fs = 10000; time = 0:1/Fs:5; noise_amp = 0; amp = amp_vec(i); input = zeros(1,length(time)); input(1*Fs+1:end) = amp; inputOpt = 1; pltOpt = 0; [binary,V_s,V_d] = Cisi2008_function(param,time,input,Fs,noise_amp,inputOpt,pltOpt); spike_time = find(binary(end-2*Fs+1:end)); ISI = diff(spike_time)/(Fs/1000); mean_FR(i) = mean(1./ISI*1000); CoV_FR(i) = std(1./ISI*1000)/mean_FR(i)*100; end index_t = find(~isnan(mean_FR)); amp_th = amp_vec(index_t(1)); min_DR = mean_FR(index_t(1)); p_fit = polyfit(amp_vec(index_t),mean_FR(index_t),1); y1 = polyval(p_fit,amp_vec(index_t)); figure(1) plot(amp_vec,mean_FR,'LineWidth',2) hold on plot(amp_vec(index_t),y1,'--k','LineWidth',1) plot([min(amp_vec) max(amp_vec)],[MDR_d MDR_d],'--k','LineWidth',1) plot([min(amp_vec) max(amp_vec)],[PDR_d PDR_d],'--k','LineWidth',1) xlabel('Injected Current (nA)','FontSize',14) ylabel('Dischage Rate (Hz)','FontSize',14) set(gca,'TickDir','out'); set(gca,'box','off')
function z = polynomial_probs(data,n,resolution,lim) C = mean(data,1); R = max( vecnorm(data-C,2,2) ); P = @(L) max( 1 - (norm(L-C)/(3*R))^n, 0 ); X = linspace(-lim,lim,resolution); Y = X; [Xg,Yg] = meshgrid(X,Y); Z = zeros(size(Xg)); for i=1:resolution^2 xx = Xg(i); yy = Yg(i); Z(i) = P([xx yy]); end z = Z/sum(sum(Z)); end
% function m = Gaussian(n) clear; clc; close all; n = 10; m = zeros(ceil(6*n),ceil(6*n)); sigma = 3*n; % for i = 1:n % for j = 1:n % top = -(i^2+j^2)/(2*sigma^2) % m(i,j) = 1/(2*pi*sigma^2) * exp(top); % end % end io = floor(size(m,1)/2); jo = floor(size(m,2)/2); for i = 1:size(m,1) for j = 1:size(m,2) top = -((i-io)^2/(2*sigma^2)+(j-jo)^2/(2*sigma^2)); m(i,j) = 1/(2*pi*sigma^2) * exp(top); end end surf(m);
function [ output_args ] = StressPlot( Maps ) %STRESSPLOT Summary of this function goes here % Detailed explanation goes here figure % S11 = Maps.S12_F; % S12 = Maps.S12_F; % S13 = Maps.S13_F; % S22 = Maps.S22_F; % S23 = Maps.S23_F; % S33 = Maps.S33_F; S11 = Maps.crop.S{1,1}; S12 = Maps.crop.S{1,2}; S13 = Maps.crop.S{1,3}; S22 = Maps.crop.S{2,2}; S23 = Maps.crop.S{2,3}; S33 = Maps.crop.S{3,3}; W = Maps.crop.W; GNDs = Maps.crop.GNDs; % Remove extreme values from plot factor = 10; S11(abs(S11)>factor*nanmean(abs(S11(:)))) = NaN; S12(abs(S12)>factor*nanmean(abs(S12(:)))) = NaN; S13(abs(S13)>factor*nanmean(abs(S13(:)))) = NaN; S22(abs(S22)>factor*nanmean(abs(S22(:)))) = NaN; S23(abs(S23)>factor*nanmean(abs(S23(:)))) = NaN; S33(abs(S33)>factor*nanmean(abs(S33(:)))) = NaN; minA = min([min(min(S11)) min(min(S12)) min(min(S13)) min(min(S22)) min(min(S23)) min(min(S33))]); maxA = max([max(max(S11)) max(max(S12)) max(max(S13)) max(max(S22)) max(max(S23)) max(max(S33))]); clim = [minA maxA]; Xvec = Maps.crop.X; Yvec = Maps.crop.Y; colormap jet h1 = subplot(3,3,1); pcolor(Xvec,Yvec,S11);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}11') h2 = subplot(3,3,2); pcolor(Xvec,Yvec,S12);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}12') h3 = subplot(3,3,3); pcolor(Xvec,Yvec,S13);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}13') h4 = subplot(3,3,5); pcolor(Xvec,Yvec,S22);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}22') h5 = subplot(3,3,6); pcolor(Xvec,Yvec,S23);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}23') h6 = subplot(3,3,9); pcolor(Xvec,Yvec,S33);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('\fontsize{16}\sigma\fontsize{10}33') %should be close to zero h7 = subplot(3,3,4); pcolor(Xvec,Yvec,W);shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colorbar title('W') h8 = subplot(3,3,7); pcolor(Xvec,Yvec,GNDs); ;shading interp; set(gca,'Ydir','normal'); axis equal xlim([min(Xvec) max(Xvec)]) ylim([min(Yvec) max(Yvec)]) colormap(jet(256)); set(gcf,'position',[500,100,950,700]); set(gca,'ColorScale','log'); set(gca,'CLim',[10^13 10^15.5]); c = colorbar; c.Label.String = 'log';%labelling title('GNDs') set(gcf,'position',[800,80,1000,900]) % L = max([abs(minA) abs(maxA)]); % set([h1 h2 h3 h4 h5 h6],'clim',[-L L]); end
function T_new = transform_TFT(T_old,M1,M2,M3,inverse) % Tranformed TFT % % short function to transform the TFT when an algebraic transformation % has been aplied to the image points. % % if inverse==0 : % from a TFT T_old assossiated to P1_old, P2_old, P3_old, find the new TFT % T_new associated to P1_new=M1*P1_old, P2_new=M2*P2_old, P3_new=M3*P3_old. % if inverse==1 : % from a TFT T_old assossiated to P1_old, P2_old, P3 _old, find the new TFT % T_new associated to M1*P1_new=P1_old, M2*P2_new=P2_old, M3*P3_new=P3_old. % % Copyright (c) 2017 Laura F. Julia <laura.fernandez-julia@enpc.fr> % All rights reserved. % % This program is free software: you can redistribute it and/or modify % it under the terms of the GNU General Public License as published by % the Free Software Foundation, either version 3 of the License, or % (at your option) any later version. % % This program is distributed in the hope that it will be useful, % but WITHOUT ANY WARRANTY; without even the implied warranty of % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the % GNU General Public License for more details. % % You should have received a copy of the GNU General Public License % along with this program. If not, see <http://www.gnu.org/licenses/>. if nargin<5 inverse=0; end if inverse==0 M1i=inv(M1); T_new=zeros(3,3,3); T_new(:,:,1)=M2*(M1i(1,1)*T_old(:,:,1) + M1i(2,1)*T_old(:,:,2) + M1i(3,1)*T_old(:,:,3) )*M3.'; T_new(:,:,2)=M2*(M1i(1,2)*T_old(:,:,1) + M1i(2,2)*T_old(:,:,2) + M1i(3,2)*T_old(:,:,3) )*M3.'; T_new(:,:,3)=M2*(M1i(1,3)*T_old(:,:,1) + M1i(2,3)*T_old(:,:,2) + M1i(3,3)*T_old(:,:,3) )*M3.'; elseif inverse==1 M2i=inv(M2); M3i=inv(M3); T_new=zeros(3,3,3); T_new(:,:,1)=M2i*(M1(1,1)*T_old(:,:,1) + M1(2,1)*T_old(:,:,2) + M1(3,1)*T_old(:,:,3) )*M3i.'; T_new(:,:,2)=M2i*(M1(1,2)*T_old(:,:,1) + M1(2,2)*T_old(:,:,2) + M1(3,2)*T_old(:,:,3) )*M3i.'; T_new(:,:,3)=M2i*(M1(1,3)*T_old(:,:,1) + M1(2,3)*T_old(:,:,2) + M1(3,3)*T_old(:,:,3) )*M3i.'; end T_new=T_new/norm(T_new(:)); end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % 本函数用于下载指定证券代码(多个请用‘,’隔开)相应指定起始和截止日期之间的日 % 行情数据,保存到backtest Cache中,文件名为证券代码。 % 这里的日行情数据格式如下: % 732686 4.460868254 4.48874868 4.419047614 4.453898147 11022101 6.39 215.9386515 % 732687 4.426017721 4.432987827 4.286615588 4.377226974 21426743 6.28 212.2213977 % 732688 4.377226974 4.426017721 4.321466121 4.342376441 9749282 6.23 210.5317369 % 732689 4.328436228 4.377226974 4.314496014 4.377226974 6720529 6.28 212.2213977 % 732690 4.356316654 4.426017721 4.307525908 4.328436228 7996673 6.21 209.8558726 % 日期 前复权开盘 前复权最高 前复权最低 前复权收盘 成交量 不复权收盘 后复权收盘 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function tool_download_callback(Path_backtest) % 获取证券代码 prompt = {'证券代码(多个请用‘,’隔开):','StartDate','EndDate'}; dlg_title = '输入下载证券代码'; def ={'000300.SHI,000001.SHI','2012-01-01','2012-06-01'}; num_lines = 1; options.Resize = 'on'; input_arg = inputdlg(prompt,dlg_title,num_lines,def,options); if ~isempty(input_arg) StartDate = input_arg{2}; EndDate = input_arg{3}; s = input_arg{1}; token = cell(1,1); index = 1; [token{index}, remain] = strtok(s,','); % 下载指定证券代码日行情数据并保存到Cache中 writeintocache_backtest(token{index},StartDate,EndDate,Path_backtest); while ~isempty(remain) s = remain; index = index + 1; [token{index}, remain] = strtok(s,','); writeintocache_backtest(token{index},StartDate,EndDate,Path_backtest); end end end
%% Utility functions classdef Util < handle %% methods (Static) %% % Scale and randomize data (matrix or filename). Use for % convenience instead of individual functions. function [data outfile] = scaleAndRandomize(data, numOutputs, inputMinValue, inputMaxValue, outputMinValue, outputMaxValue) if nargin < 2 error('scaleAndRandomize(): Not enough input arguments.'); end outfile = ''; if ischar(data) outfile = Util.appendFilename(data, '_scaled_randomized'); data = xlsread(data); end switch nargin case 2 data = Util.scale(data, numOutputs); case 4 data = Util.scale(data, numOutputs, inputMinValue, inputMaxValue); case 6 data = Util.scale(data, numOutputs, inputMinValue, inputMaxValue, outputMinValue, outputMaxValue); otherwise error('scaleAndRandomize(): Invalid number of arguments.'); end data = Util.randomize(data); if ~isempty(outfile) xlswrite(outfile, data); end end %% % Scale `data` (a matrix or an xls filename) to the given min/max values. % default input limits are [-1, 1] default output limits are [0, 1]. % If data is a filename, filename_scaled.xls will also be written. function [scaled outfile] = scale(data, numOutputs, inputMinValue, inputMaxValue, outputMinValue, outputMaxValue) if nargin < 2 error('scale(): Not enough input arguments.'); end outfile = ''; if ischar(data) outfile = Util.appendFilename(data, '_scaled'); data = xlsread(data); end if isempty(data) error('scale(): Data is empty.'); end if numOutputs < 1 error('scale(): There must be at least one output.'); end numInputs = size(data, 2) - numOutputs; if numInputs < 1 error('scale(): There must be at least one input.'); end numSamples = size(data, 1); if nargin == 3 || nargin == 5 error('scale(): maxValue must be defined if minValue is defined.'); end if nargin == 4 || nargin == 6 if inputMinValue >= inputMaxValue || outputMinValue >= outputMaxValue error('scale(): minValue must be less than maxValue.'); end else inputMinValue = -1; inputMaxValue = 1; outputMinValue = 0; outputMaxValue = 1; end scaled = zeros(size(data)); scaleDiff = inputMaxValue - inputMinValue; for i = 1 : numInputs minval = min(data(:, i)); maxval = max(data(:, i)); dataDiff = maxval - minval; for j = 1 : numSamples scaled(j, i) = inputMinValue + ((data(j, i) - minval) * scaleDiff) / dataDiff; end end scaleDiff = outputMaxValue - outputMinValue; for i = (numInputs + 1) : (numInputs + numOutputs) minval = min(data(:, i)); maxval = max(data(:, i)); dataDiff = maxval - minval; for j = 1 : numSamples scaled(j, i) = outputMinValue + ((data(j, i) - minval) * scaleDiff) / dataDiff; end end if ~isempty(outfile) xlswrite(outfile, scaled); end end %% % Randomize the rows in `data` (a matrix or an xls filename). % If data is a filename, filename_randomized.xls will also be written. function [randomized outfile] = randomize(data) if nargin < 1 error('randomize(): Not enough input arguments.'); end outfile = ''; if ischar(data) outfile = Util.appendFilename(data, '_randomized'); data = xlsread(data); end if isempty(data) error('randomize(): Data is empty.'); end randomized = data(randperm(size(data,1)),:); if ~isempty(outfile) xlswrite(outfile, randomized); end end %% % Insert text between the end of a file's name and the extension. function newname = appendFilename(filename, text) if nargin < 2 || ~ischar(filename) || ~ischar(text) error('appendFilename(): Two string arguments are required.'); end strcat(filename, ''); [~, ~, ext] = fileparts(filename); newname = strcat(filename(1:(length(filename) - length(ext))), text, ext); end end methods (Access = private) end end
function [itrs, w_s, AUCs, timing] = OPAUC(X,y,X_test,y_test,eta,lambda,n_delta,u) % Usage: OPAUC algorithm % nargin < 8: standard OPAUC % nargin == 8: OPAUCr (OPAUC for large scale data) % Author: xzhang % Date: 2017.11.16 % Version: 1.2 tic; 'new~' % Returns AUCs = []; itrs = []; timing = []; w_s = []; y = sparse(y); % parameters [d,n] = size(X); if nargin < 8 || ~u init = @zeros; u = d; std = 1; else disp('large scale and going with sparse setting ...') std = 0; init = @sparse; R_hat_p = init(1,u); R_hat_n = init(1,u); end T_p = 0; T_n = 0; c_p = init(d,1); c_n = init(d,1); w = init(d,1); Gamma_p = init(d,u); Gamma_n = init(d,u); delta = floor(n/n_delta); sample = randsample(n, n, 'false'); for i = 1:n xi = X(:,sample(i)); yi = y(sample(i)); if yi == 1 T_p = T_p+1; c_p = c_p+(xi-c_p)/T_p; if std [Gamma_p,g] = std_update(xi,yi,w,lambda,Gamma_p,c_p,T_p,Gamma_n,c_n); else [Gamma_p,g,R_hat_p] = large_update(xi,yi,w,lambda,u,Gamma_p,R_hat_p,Gamma_n,c_n,R_hat_n,T_n); end else T_n = T_n+1; c_n = c_n+(xi-c_n)/T_n; if std [Gamma_n,g] = std_update(xi,yi,w,lambda,Gamma_n,c_n,T_n,Gamma_p,c_p); else [Gamma_n,g,R_hat_n] = large_update(xi,yi,w,lambda,u,Gamma_n,R_hat_n,Gamma_p,c_p,R_hat_p,T_p); end end w = w-eta*g; % update model % collect intermediate results if mod(i,delta) == 0 [~,~,~,AUC] = perfcurve(y_test,w'*X_test,1); disp(['***OPAUC: itr=', num2str(i), ', AUC=', num2str(AUC)]); AUCs = cat(1, AUCs, AUC); itrs = cat(1, itrs, i); timing = cat(1, timing, toc); w_s = cat(2, w_s, w); end end end function [Gamma,g] = std_update(xi,yi,w,lambda,Gamma,c,T,Gamma_op,c_op) tmp = c; Gamma = Gamma+(xi*xi'-Gamma)/T+tmp*tmp'-c*c'; g = lambda*w-yi*xi+yi*c_op+(xi-c_op)*(xi-c_op)'*w+Gamma_op*w; end function [Gamma,g,R_hat] = large_update(xi,yi,w,lambda,u,Gamma,R_hat,Gamma_op,c_op,R_hat_op,T_op) % ri = randn(u,1); ri = sparse(randn(u,1)); R_hat = R_hat + ri'/u; % update R_hat Gamma = Gamma + xi*ri'/u; % update Z tmp = xi-c_op; g = lambda*w-yi*(tmp)+(tmp)*((tmp)'*w); % if T_op > 0 % c_hat_op = (c_op*R_hat_op)/T_op; g = g + Gamma_op*(Gamma_op'*w)/T_op-(c_op*(R_hat_op*R_hat_op')*(c_op'*w))/(T_op^2); end end
%Name: % getu % %Purpose: % This method will be used to apply the Dirichlet boundary condition. % Where Vbound indicates we need the displacement to be zero. % %Parameters: % A - ((2x#vertices) x (2x#vertices)) stiffness matrix % F - ((2x#vertices) x 1) vector which represent the force value of each % vertex % Vbound (#vertices x 3) - matrix which shows which vertices were chosen % to be Dirichlet boundary points % %Return Values: % u - ((2x#vertices) x 1) vector which represents the displacment each % vertex will recieve (our solution) % %Author: % Shea Yonker % %Date: % 09/18/2017 function [u] = getu(A,F,Vbound) s = size(A,1); u=zeros(s,1); m=1; B=zeros(s); C=zeros(s,1); for j=1:(s/2) if (Vbound(j,3) == 1) else n=1; for i=1:s if (i<=(s/2)) if (Vbound(i,3) == 1) else B(m,n)=A(j,i); n=n+1; end else if (Vbound(i-(s/2),3) == 1) else B(m,n)=A(j,i); n=n+1; end end end C(m)=F(j); m=m+1; end end for j=((s/2)+1):s if (Vbound(j-(s/2),3) == 1) else n=1; for i=1:s if (i<=(s/2)) if (Vbound(i,3) == 1) else B(m,n)=A(j,i); n=n+1; end else if (Vbound(i-(s/2),3) == 1) else B(m,n)=A(j,i); n=n+1; end end end C(m)=F(j); m=m+1; end end Areduced=zeros(m-1); Freduced=zeros(m-1,1); for j=1:m-1 for i=1:m-1 Areduced(j,i)=B(j,i); end Freduced(j)=C(j); end ureduced=Areduced\Freduced; m=1; for j=1:(s/2) if (Vbound(j,3) == 1) u(j)=0; else u(j)=ureduced(m); m=m+1; end end for j=((s/2)+1):s if (Vbound(j-(s/2),3) == 1) u(j)=0; else u(j)=ureduced(m); m=m+1; end end end
clc clear % x = [1,2,3;4,5,6;]; x = double(rgb2gray(imread('/Users/INNOCENTBOY/Documents/MATLAB/pic/256/4.1.01.tiff'))); h = ones(5,5); % h = h/25; M= size(x,1); N= size(x,2); m= size(h,1); n= size(h,2); h2 = fft2(h,M+m-1,N+n-1); x2 = fft2(x,M+m-1,N+n-1); h1 = fft2(h,M,N); x1 = fft2(x,M,N); out = x1.*h1; out = ifft2(out); out1 = x2.*h2; out1 = ifft2(out1); figure subplot(2,2,1),imshow(mat2gray(x)); subplot(2,2,2),imshow(mat2gray(out)); subplot(2,2,3),imshow(mat2gray(out1));
function pixelgrid( h ) % Generates horizontal and vertical pixel grid lines for every pixel in an % open image specified by the handle h. Operates on an already displayed % figure and image. % % Required Input % ============== % h Figure handle correspoding to the open image to operate on % % Reference % ========= % http://blogs.mathworks.com/steve/2011/02/17/pixel-grid/ % h = findobj(h, 'type', 'image'); xdata = get(h, 'XData'); ydata = get(h, 'YData'); M = size(get(h, 'CData'), 1); N = size(get(h, 'CData'), 2); if M > 1 pixel_height = diff(ydata) / (M-1); else pixel_height = 1; end if N > 1 pixel_width = diff(xdata) / (M-1); else pixel_width = 1; end y_top = ydata(1) - (pixel_height / 2); y_bottom = ydata(2) + (pixel_height / 2); y = linspace(y_top, y_bottom, M+1); x_left = xdata(1) - (pixel_width / 2); x_right = xdata(2) + (pixel_width / 2); x = linspace(x_left, x_right, N+1); xv = zeros(1, 2*numel(x)); xv(1:2:end) = x; xv(2:2:end) = x; yv = repmat([y(1) ; y(end)], 1, numel(x)); yv(:,2:2:end) = flipud(yv(:,2:2:end)); xv = xv(:); yv = yv(:); yh = zeros(1, 2*numel(y)); yh(1:2:end) = y; yh(2:2:end) = y; xh = repmat([x(1) ; x(end)], 1, numel(y)); xh(:,2:2:end) = flipud(xh(:,2:2:end)); xh = xh(:); yh = yh(:); dark = [0.3 0.3 0.3]; light = [0.8 0.8 0.8]; hold on; ax = ancestor(h, 'axes'); line('Parent', ax, 'XData', xh, 'YData', yh, 'Color', dark, ... 'LineStyle', '-', 'Clipping', 'off'); line('Parent', ax, 'XData', xh, 'YData', yh, 'Color', light, ... 'LineStyle', '--', 'Clipping', 'off'); line('Parent', ax, 'XData', xv, 'YData', yv, 'Color', dark, ... 'LineStyle', '-', 'Clipping', 'off'); line('Parent', ax, 'XData', xv, 'YData', yv, 'Color', light, ... 'LineStyle', '--', 'Clipping', 'off'); hold off; end
function[nll] = negPopLogLikelihood(motion,f,r,params) % function[nll] = negPopLogLikelihood(motion,f,r,params) % % motion: (e.g., specifies [direction, speed] for xz motion) % f: tuning curve function handle % r: "neural responses" % params: parameters for tuning curves % % nll: negative loglikelihood % % see Graf, Kohn, Jazayeri, & Movshon (2011) % motion(2) = exp(motion(2)); ftemp = f(motion,params); % calculate tuning curve responses to particular motion weights = log(ftemp); ll = bsxfun(@times,weights,r) - ftemp - repmat(logfactorial(r),1,size(motion,1)); nll = -sum(ll);
function [mapout,xgrid] = sz1SM(data,paramstruct) % SZ1SM, Significant derivative Zero crossings % Steve Marron's matlab function % Creates color map (function of location and bandwidth), % showing statistical signicance of slope of smooth % Colored (or gray level) as: % blue (dark) at points where deriv sig > 0 % red (light) at points where deriv sig < 0 % purple (gray) at points where deriv roughly 0 % light gray where "effective sample size" < 5 % % Inputs: % % data - Case 1: density estimation: % either n x 1 column vector of 1-d data % or vector of bincts, when imptyp = -1 % Case 2: regression: % either n x 2 matrix of Xs (1st col) and Ys (2nd col) % or g x 3 matrix of bincts, when imptyp = -1 % (3rd column is wt'd bin avg's of squares) % % paramstruct - a Matlab structure of input parameters % Use: "help struct" and "help datatypes" to % learn about these. % Create one, using commands of the form: % % paramstruct = struct('field1',values1,... % 'field2',values2,... % 'field3',values3) ; % % where any of the following can be used, % these are optional, misspecified values % revert to defaults % % fields values % % vxgp vector of x grid parameters: % 0 (or not specified) - use endpts of data and 401 bins % [le; lr; nb] - le left, re right, and nb bins % % vhgp vector of h (bandwidth) grid parameters: % 0 (or not specified) - use (2*binwidth) to (range), % and 21 h's % [hmin; hmax; nh] - use hmin to hmax and nh h's. % % imptyp flag indicating implementation type: % -1 - binned version, and "data" is assumed to be % bincounts of prebinned data % CAUTION: for regression, also need % to include bincounts of squares % as a third column % 0 (or not specified) - linear binned version % and bin data here % % eptflag endpoint truncation flag (only has effect when imptyp = 0): % 0 (or not specified) - move data outside range to % nearest endpoint % 1 - truncate data outside range % % ibdryadj index of boundary adjustment % 0 (or not specified) - no adjustment % 1 - mirror image adjustment % 2 - circular design % (only has effect for density estimation) % % iregtdist index for using Student's t distribution in regression % 0 (or not specified) - use only Gaussian approximation % 1 - use Student's t distribution quantiles in testing % (only has effect for regression) % % alpha Usual "level of significance". I.e. C.I.s have coverage % probability 1 - alpha. (0.05 when not specified) % % simflag Confidence Interval type (simultaneous vs. ptwise) % 0 - Use Pointwise C.I.'s % 1 (or not specified) - Use Row-wise Simultaneous C.I.'s % 2 - Use Global Simultaneous C.I.'s % 3 - Use Mixture 1 (Bonferroni across rows) Simultaneous C.I.'s % 4 - Use Mixture 2 (Bonferroni across rows) Simultaneous C.I.'s % 5 - 3 times Eff. SS global % 6 - 3 times Eff. SS Row-wise % % icolor 1 (default) full color version of SiZer % 0 fully black and white version % % titlestr string with title (only has effect when plot is made here) % default is empty string, '', for no title % % titlefontsize font size for title % (only has effect when plot is made here, % and when the titlestr is nonempty) % default is empty [], for Matlab default % % xlabelstr string with x axis label % (only has effect when plot is made here) % default is empty string, '', for no xlabel % % ylabelstr string with y axis label % (only has effect when plot is made here) % default is empty string, '', for no ylabel % % labelfontsize font size for axis labels % (only has effect when plot is made here, % and when a label str is nonempty) % default is empty [], for Matlab default % % iplot 1 - plot even when there is numerical output % 0 - (default) only plot when no numerical output % % % Outputs: % (none) - Draws a gray level map (in the current axes) % mapout - output of indices into color map % xgrid - col vector grid of points at which estimate(s) are % evaluted (useful for plotting), unless grid is input, % can also get this from linspace(le,re,nb)' % % Assumes path can find personal functions: % vec2matSM.m % lbinrSM.m % % Copyright (c) J. S. Marron 1997-2001 % First set all parameters to defaults vxgp = 0 ; vhgp = 0 ; imptyp = 0 ; eptflag = 0 ; ibdryadj = 0 ; iregtdist = 0 ; alpha = 0.05 ; simflag = 1 ; icolor = 1 ; titlestr = '' ; titlefontsize = [] ; xlabelstr = '' ; ylabelstr = '' ; labelfontsize = [] ; iplot = 0 ; % Now update parameters as specified, % by parameter structure (if it is used) % if nargin > 1 ; % then paramstruct has been added if isfield(paramstruct,'vxgp') ; % then change to input value vxgp = getfield(paramstruct,'vxgp') ; end ; if isfield(paramstruct,'vhgp') ; % then change to input value vhgp = getfield(paramstruct,'vhgp') ; end ; if isfield(paramstruct,'eptflag') ; % then change to input value eptflag = getfield(paramstruct,'eptflag') ; end ; if isfield(paramstruct,'ibdryadj') ; % then change to input value ibdryadj = getfield(paramstruct,'ibdryadj') ; end ; if isfield(paramstruct,'iregtdist') ; % then change to input value iregtdist = getfield(paramstruct,'iregtdist') ; end ; if isfield(paramstruct,'alpha') ; % then change to input value alpha = getfield(paramstruct,'alpha') ; end ; if isfield(paramstruct,'simflag') ; % then change to input value simflag = getfield(paramstruct,'simflag') ; end ; if isfield(paramstruct,'icolor') ; % then change to input value icolor = getfield(paramstruct,'icolor') ; end ; if isfield(paramstruct,'titlestr') ; % then change to input value titlestr = getfield(paramstruct,'titlestr') ; end ; if isfield(paramstruct,'titlefontsize') ; % then change to input value titlefontsize = getfield(paramstruct,'titlefontsize') ; end ; if isfield(paramstruct,'xlabelstr') ; % then change to input value xlabelstr = getfield(paramstruct,'xlabelstr') ; end ; if isfield(paramstruct,'ylabelstr') ; % then change to input value ylabelstr = getfield(paramstruct,'ylabelstr') ; end ; if isfield(paramstruct,'labelfontsize') ; % then change to input value labelfontsize = getfield(paramstruct,'labelfontsize') ; end ; if isfield(paramstruct,'iplot') ; % then change to input value iplot = getfield(paramstruct,'iplot') ; end ; if isfield(paramstruct,'imptyp') ; % then change to input value imptyp = getfield(paramstruct,'imptyp') ; end ; end ; % of resetting of input parameters % detect whether density or regression data % if size(data,2) == 1 ; % Then is density estimation xdat = data(:,1) ; idatyp = 1 ; itdist = 0 ; % use Gaussian distribution, not t else ; % Then assume regression ; xdat = data(:,1) ; ydat = data(:,2) ; idatyp = 2 ; if iregtdist == 1 ; itdist = 1 ; % use t distribution else ; itdist = 0 ; % use Gaussian distribution, not t end ; end ; % Set x grid stuff % n = length(xdat) ; if vxgp == 0 ; % then use standard default x grid vxgp = [min(xdat),max(xdat),401] ; end ; left = vxgp(1) ; right = vxgp(2) ; ngrid = vxgp(3) ; xgrid = linspace(left,right,ngrid)' ; % col vector to evaluate smooths at cxgrid = xgrid - mean(xgrid) ; % centered version, gives numerical stability % Set h grid stuff % if vhgp == 0 ; % then use standard default h grid range = right - left ; binw = range / (ngrid - 1) ; vhgp = [2 * binw,range,21] ; end ; hmin = vhgp(1) ; hmax = vhgp(2) ; nh = vhgp(3) ; if nh == 1 ; vh = hmax ; else ; if hmin < hmax ; % go ahead with vh construction vh = logspace(log10(hmin),log10(hmax),nh) ; else ; % bad inputs, warn and quit disp('!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!') ; disp('!!! Error from sz1SM.m: !!!') ; disp('!!! Bad inputs in vhgp, !!!') ; disp('!!! Reguires vhgp(1) < vhgp(2) !!!') ; disp('!!! Terminating execution !!!') ; disp('!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!') ; return ; end ; end ; % Bin the data (if needed) % if idatyp == 1 ; % Treating as density estimation if imptyp == -1 ; % Then data have already been binned bincts = data ; else ; % Then need to bin data bincts = lbinrSM(xdat,vxgp,eptflag) ; end ; elseif idatyp == 2 ; % Treating as regression if imptyp == -1 ; % Then data have already been binned bincts = data(:, [1 2]) ; bincts2 = data(:, [1 3]) ; else ; % Then need to bin data bincts = lbinrSM([xdat ydat],vxgp,eptflag) ; bincts2 = lbinrSM([xdat, ydat.^2],vxgp,eptflag) ; end ; end ; n = sum(bincts(:,1)) ; % put this here in case of truncations during binning % Construct Surfaces % mdsurf = [] ; % Derivative surface mesurf = [] ; % Effective sample size surface mvsurf = [] ; % Estimated Variance of Derivative if itdist == 0 ; vgq = [] ; % Vector of Gaussian Quantiles (for simultaneous CI's) elseif itdist == 1 ; mtq = [] ; % matrix of t Quantiles (for simultaneous CI's) end ; % Create grid for kernel values delta = (right - left) / (ngrid - 1) ; % binwidth k = ngrid - 1 ; % index of last nonzero entry of kernel vector % Loop through bandwidths ihprevcalc = 0 ; % indicator of whether have done the full calculation % (not done when all locations have ESS < 5) for ih = 1:nh ; h = vh(ih) ; % Set common values arg = linspace(0,k * delta / h,k + 1)' ; kvec = exp(-(arg.^2) / 2) ; kvec = [flipud(kvec(2:k+1)); kvec] ; % construct symmetric kernel if idatyp == 1 ; % Doing density estimation % main lines from gpkde.m, via nmsur5.m % do boundary adjustment if needed % if ibdryadj == 1 ; % then do mirror image adjustment babincts = [flipud(bincts); bincts; flipud(bincts)] ; elseif ibdryadj == 2 ; % then do circular design adjustment babincts = [bincts; bincts; bincts] ; else ; babincts = bincts ; end ; % Vector of Effective sample sizes % (notation "s0" is consistent with below) ve = conv(babincts,kvec) ; if ibdryadj == 1 | ibdryadj == 2 ; % then did boundary adjustment ve = ve(ngrid+k+1:k+2*ngrid) ; else ; ve = ve(k+1:k+ngrid) ; end ; % denominator of NW est. % (same as sum for kde) kvecd = -arg .* exp(-(arg.^2) / 2) / sqrt(2 * pi) ; kvecd = [-flipud(kvecd(2:k+1)); kvecd] ; % construct symmetric kernel vd = conv(babincts,kvecd) ; vv = conv(babincts,kvecd.^2) ; if ibdryadj == 1 | ibdryadj == 2 ; % then did boundary adjustment vd = vd(ngrid+k+1:k+2*ngrid) / (n * h^2) ; vv = vv(ngrid+k+1:k+2*ngrid) / (n * h^4) ; else ; vd = vd(k+1:k+ngrid) / (n * h^2) ; vv = vv(k+1:k+ngrid) / (n * h^4) ; end ; vv = vv - vd.^2 ; vv = vv / n ; % did this outsidea loop in nmsur5.m else ; % Doing regression % using modification of lines from gpnpr.m % main lines from gpnpr.m, via szeg4.m % Vector of Effective sample sizes % (notation "s0" is consistent with below) ve = conv(bincts(:,1),kvec) ; ve = ve(k+1:k+ngrid) ; % denominator of NW est. % (same as sum for kde) flag = (ve < 1) ; % locations with effective sample size < 1 ve(flag) = ones(sum(flag),1) ; % replace small sample sizes by 1 to avoid 0 divide % no problem below, since gets grayed out s1 = conv(bincts(:,1) .* cxgrid , kvec) ; s1 = s1(k+1:k+ngrid) ; s2 = conv(bincts(:,1) .* cxgrid.^2 , kvec) ; s2 = s2(k+1:k+ngrid) ; t0 = conv(bincts(:,2),kvec) ; t0 = t0(k+1:k+ngrid) ; % numerator of NW est. xbar = conv(bincts(:,1) .* cxgrid , kvec) ; xbar = xbar(k+1:k+ngrid) ; % Weighted sum of X_i xbar = xbar ./ ve ; % Weighted avg of X_i t1 = conv(bincts(:,2) .* cxgrid , kvec) ; t1 = t1(k+1:k+ngrid) ; numerd = t1 - t0 .* xbar ; % sum(Y_i * (X_i - xbar)) * W (weighted cov(Y,X)) denomd = s2 - 2 * s1 .* xbar + ve .* xbar.^2 ; % sum((X_i - xbar)^2) * W (weighted var(X)) flag2 = denomd < (10^(-10) * mean(denomd)) ; % for local linear, need to also flag locations where this % is small, because could have many observaitons piled up % at one location ve(flag2) = ones(sum(flag2),1) ; % also reset these, because could have more than 5 piled up % at a single point, but nothing else in window flag = flag | flag2 ; % logical "or", which flags both types of locations % to avoid denominator problems denomd(flag) = ones(sum(flag),1) ; % this avoids zero divide problems, OK, since grayed out later mhat = t0 ./ ve ; vd = numerd ./ denomd ; % linear term from local linear fit (which est's deriv). % (sometimes called betahat) sig2 = conv(bincts2(:,2),kvec) ; sig2 = sig2(k+1:k+ngrid) ; sig2 = sig2 ./ ve - mhat.^ 2 ; flag2 = sig2 < (10^(-10) * mean(sig2)) ; ve(flag2) = ones(sum(flag2),1) ; % also reset these flag = flag | flag2 ; % logical "or", which flags both types of locations % to avoid denominator problems sig2(flag) = ones(sum(flag),1) ; % this avoids zero divide problems, OK, since grayed out later rho2 = vd.^2 .* (denomd ./ (sig2 .* ve)) ; sig2res = (1 - rho2) .* sig2 ; % get the residual variance from local linear reg. u0 = conv(bincts(:,1) .* sig2res , kvec.^2) ; u0 = u0(k+1:k+ngrid) ; u1 = conv(bincts(:,1) .* sig2res .* cxgrid , kvec.^2) ; u1 = u1(k+1:k+ngrid) ; u2 = conv(bincts(:,1) .* sig2res .* cxgrid.^2 , kvec.^2) ; u2 = u2(k+1:k+ngrid) ; vv = u2 - 2 * u1 .* xbar + u0 .* xbar.^2 ; vv = vv ./ denomd.^2 ; % vector of variances of slope est. (from local linear) end ; % Get quantiles, for CI's flag = (ve >= 5) ; % locations where effective sample size >= 5 if sum(flag) > 0 ; if simflag == 0 ; % do pt'wise CI's if itdist == 0 ; gquant = norminv(1 - (alpha / 2)) ; elseif itdist == 1 ; vtquant = tinv(1 - (alpha / 2),ve-1) ; end ; elseif simflag == 1 ; % do row-wise simultaneous CI's nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups beta = (1 - alpha)^(1/numind) ; if itdist == 0 ; gquant = -norminv((1 - beta) / 2) ; elseif itdist == 1 ; vtquant = -tinv((1 - beta) / 2,ve-1) ; end ; elseif simflag == 2 ; % do Global Simultaneous C.I.'s if ihprevcalc == 0 ; % then are at finest scale, % where have not done a calculation yet ihprevcalc = 1 ; % turn off this indicator, so won't calculate this again nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups numind = 2 * numind - 1 ; % modify to make global beta = (1 - alpha)^(1/numind) ; end ; % at other scales, continue to use this value if itdist == 0 ; gquant = -norminv((1 - beta) / 2) ; elseif itdist == 1 ; vtquant = -tinv((1 - beta) / 2,ve-1) ; end ; elseif simflag == 3 ; % do Mixture 1 (Bonferroni across rows) Simul. C.I.'s nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups beta = (1 - (alpha / nh))^(1/numind) ; % only change from simflag = 1 above is: alpha ---> alpha / nh % (Bonferroni across rows, calculation) if itdist == 0 ; gquant = -norminv((1 - beta) / 2) ; elseif itdist == 1 ; vtquant = -tinv((1 - beta) / 2,ve-1) ; end ; elseif simflag == 4 ; % do Mixture 2 (Bonferroni across rows) Simul. C.I.'s nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups if itdist == 0 ; gquant = numind ; elseif itdist == 1 ; vtquant = numind ; end ; % store the Effective number of groups, for later processing % CAUTION: in this case, don't return quantiles, but just % numind, for later turning into quantiles elseif simflag == 5 ; % do 3 times Global Simultaneous C.I.'s if ihprevcalc == 0 ; % then are at finest scale, % where have not done a calculation yet ihprevcalc = 1 ; % turn off this indicator, so won't calculate this again nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups numind = 6 * numind - 5 ; % modify to make global beta = (1 - alpha)^(1/numind) ; end ; % at other scales, continue to use this value if itdist == 0 ; gquant = -norminv((1 - beta) / 2) ; elseif itdist == 1 ; vtquant = -tinv((1 - beta) / 2,ve-1) ; end ; elseif simflag == 6 ; % do 3 times row-wise simultaneous CI's nxbar = mean(ve(flag)) ; % Crude average effective sample size numind = n / nxbar ; % Effective number of independent groups numind = 3 * numind - 2 ; beta = (1 - alpha)^(1/numind) ; if itdist == 0 ; gquant = -norminv((1 - beta) / 2) ; elseif itdist == 1 ; vtquant = -tinv((1 - beta) / 2,ve-1) ; end ; end ; else ; if itdist == 0 ; gquant = inf ; elseif itdist == 1 ; vtquant = inf * ones(ngrid,1) ; end ; end ; mdsurf = [mdsurf vd] ; mesurf = [mesurf ve] ; mvsurf = [mvsurf vv] ; if itdist == 0 ; vgq = [vgq gquant] ; elseif itdist == 1 ; mtq = [mtq vtquant] ; end ; end ; % finish vgq calculation (if needed) % (i.e. if only stored numind, instead of quantiles above) % if simflag == 4 ; % do Mixture 2 (Bonferroni across rows) Simul. C.I.'s if itdist == 0 ; vnumind = vgq ; % unpack column vector of number of indep blocks, stored above elseif itdist == 1 ; vnumind = mtq ; % unpack column vector of number of indep blocks, stored above end ; vrowfactor = vnumind .^ (0.5) ; vrowfactor = vrowfactor / sum(vrowfactor) ; vbeta = (1 - (alpha .* vrowfactor)).^ (1 ./ vnumind) ; if itdist == 0 ; vgq = -norminv((1 - vbeta) / 2) ; elseif itdist == 1 ; mtq = -tinv(vec2matSM((1 - vbeta) / 2,ngrid), vec2matSM(ve-1,nh)) ; % note first is an nh x 1 column vector % and second is a 1 x ngrid row vector end ; end ; %vrowfactor %1/nh %vnumind %vgq %pauseSM % Construct scale space CI surfaces % if itdist == 0 ; if nh > 1 ; % then have full matrices mloci = mdsurf - vec2matSM(vgq,ngrid) .* sqrt(mvsurf) ; % Lower confidence (simul.) surface for derivative mhici = mdsurf + vec2matSM(vgq,ngrid) .* sqrt(mvsurf) ; % Upper confidence (simul.) surface for derivative else ; % have only vectors (since only one h) mloci = mdsurf - vgq * sqrt(mvsurf) ; % Lower confidence (simul.) surface for derivative mhici = mdsurf + vgq * sqrt(mvsurf) ; % Upper confidence (simul.) surface for derivative end ; elseif itdist == 1 ; mloci = mdsurf - mtq .* sqrt(mvsurf) ; % Lower confidence (simul.) surface for derivative mhici = mdsurf + mtq .* sqrt(mvsurf) ; % Upper confidence (simul.) surface for derivative end ; % Construct "color map", really assignment % of pixels to integers, with idea: % 1 (very dark) - Deriv. Sig. > 0 % 2 (darker gray) - Eff. SS < 5 % 3 (lighter gray) - Eff. SS >= 5, but CI contains 0 % 4 (very light) - Deriv. Sig. < 0 mapout = 3 * ones(size(mloci)) ; % default is purple (middle lighter gray) flag = (mloci > 0) ; % matrix of ones where lo ci above 0 ssflag = sum(sum(flag)) ; if ssflag > 0 ; mapout(flag) = ones(ssflag,1) ; % put blue (dark grey) where significantly positive end ; flag = (mhici < 0) ; % matrix of ones where hi ci below 0 ssflag = sum(sum(flag)) ; if ssflag > 0 ; mapout(flag) = 4 * ones(ssflag,1) ; % put red (light gray) where significantly negative end ; flag = (mesurf <= 5) ; % matrix of ones where effective np <= 5 ; ssflag = sum(sum(flag)) ; if ssflag > 0 ; mapout(flag) = 2 * ones(ssflag,1) ; % put middle darker gray where effective sample size < 5 end ; % Transpose for graphics purposes mapout = mapout' ; % Make plots if no numerical output requested % if nargout == 0 | ... iplot == 1 ; % Then make a plot if icolor ~= 0 ; % Then go for nice colors in sizer and sicon % Set up colorful color map cocomap = [0, 0, 1; ... .35, .35, .35; ... .5, 0, .5; ... 1, 0, 0; ... 1, .5, 0; ... .35, .35, .35; ... 0, 1, 0; ... 0, 1, 1] ; colormap(cocomap) ; else ; % Then use gray scale maps everywhere % Set up gray level color map comap = [.2, .2, .2; ... .35, .35, .35; ... .5, .5, .5; ... .8, .8, .8] ; colormap(comap) ; end ; image([left,right],[log10(hmin),log10(hmax)],mapout) ; set(gca,'YDir','normal') ; if ~isempty(titlestr) ; if isempty(titlefontsize) ; title(titlestr) ; else ; title(titlestr,'FontSize',titlefontsize) ; end ; end ; if ~isempty(xlabelstr) ; if isempty(labelfontsize) ; xlabel(xlabelstr) ; else ; xlabel(xlabelstr,'FontSize',labelfontsize) ; end ; end ; if ~isempty(ylabelstr) ; if isempty(labelfontsize) ; ylabel(ylabelstr) ; else ; ylabel(ylabelstr,'FontSize',labelfontsize) ; end ; end ; end ;
classdef AbstractAnimator < handle % An abstract animator class % % @author omar @date 2017-06-01 % % Copyright (c) 2017, UMICH Biped Lab % All right reserved. % % Redistribution and use in source and binary forms, with or without % modification, are permitted only in compliance with the BSD 3-Clause % license, see % http://www.opensource.org/licenses/BSD-3-Clause properties(Access = private) tmr timer end properties (GetAccess = protected, SetAccess = immutable) fig axs end properties (Access = public) currentTime double speed double updateWorldPosition logical end properties (Access = public) isLooping logical pov frost.Animator.AnimatorPointOfView startTime double endTime double end properties (GetAccess = public, SetAccess = immutable) TimerDelta double end properties (Dependent, Access = public) isPlaying logical end properties (Access = private) x_all; t_all; text; ground; end properties (Access = public) display end methods function obj = AbstractAnimator(display, t, x, varargin) % if exist('f', 'var') % obj.fig = f; % obj.axs = axes(obj.fig); % else % obj.fig = figure(); % obj.axs = axes(obj.fig); % end obj.display = display; obj.fig = display.fig; obj.axs = display.axs; obj.t_all = t; obj.x_all = x; obj.startTime = t(1); obj.currentTime = obj.startTime; obj.endTime = t(end); % setup timer obj.speed = 1; obj.TimerDelta = round(1/30,3); obj.pov = frost.Animator.AnimatorPointOfView.Free; obj.tmr = timer; obj.tmr.Period = obj.TimerDelta; obj.tmr.ExecutionMode = 'fixedRate'; obj.tmr.TimerFcn = @(~, ~) obj.Animate(); % Define Terrain if isempty(varargin) [terrain.Tx, terrain.Ty] = meshgrid(-10:1:10, -10:1:10); terrain.Tz = 0.*terrain.Tx; else terrain = varargin{1}; end obj.ground = surf(terrain.Tx,terrain.Ty,terrain.Tz,'FaceColor',[0.5 0.8 0.5]); hold on; end function playing = get.isPlaying(obj) playing = strcmp(obj.tmr.Running, 'on'); end function set.isPlaying(obj, play) if ~obj.isPlaying && play start(obj.tmr); notify(obj, 'playStateChanged'); elseif obj.isPlaying && ~play stop(obj.tmr); notify(obj, 'playStateChanged'); end end function set.currentTime(obj, time) obj.currentTime = time; if obj.currentTime > obj.endTime %#ok<*MCSUP> obj.currentTime = obj.endTime; elseif obj.currentTime < obj.startTime obj.currentTime = obj.startTime; end end end methods (Sealed) function Animate(obj, Freeze) if ~exist('Freeze', 'var') Freeze = false; end if obj.currentTime >= obj.endTime obj.currentTime = obj.endTime; x = GetData(obj, obj.currentTime); notify(obj, 'newTimeStep', frost.Animator.TimeStepData(obj.currentTime, x)); obj.Draw(obj.currentTime, x); obj.HandleAxis(obj.currentTime, x); notify(obj, 'reachedEnd', frost.Animator.TimeStepData(obj.currentTime, x)); if obj.isLooping if ~Freeze obj.currentTime = obj.startTime; end else obj.isPlaying = false; end else x = GetData(obj, obj.currentTime); notify(obj, 'newTimeStep', frost.Animator.TimeStepData(obj.currentTime, x)); obj.Draw(obj.currentTime, x); obj.HandleAxis(obj.currentTime, x); if ~Freeze obj.currentTime = obj.currentTime + obj.TimerDelta*obj.speed; end end end end methods function HandleAxis(obj, t, x) [center, radius, yaw] = GetCenter(obj, t, x); if length(radius) == 1 axis(obj.axs, [center(1)-radius, center(1)+radius, center(2)-radius, center(2)+radius,center(3)-radius/3, center(3)+radius]); else axis(obj.axs, radius); end hAngle = 0; vAngle = 0; switch(obj.pov) case frost.Animator.AnimatorPointOfView.North hAngle = hAngle + 0; case frost.Animator.AnimatorPointOfView.South hAngle = hAngle + 180; case frost.Animator.AnimatorPointOfView.East hAngle = hAngle - 90; case frost.Animator.AnimatorPointOfView.West hAngle = hAngle + 90; case frost.Animator.AnimatorPointOfView.Front hAngle = hAngle + yaw; case frost.Animator.AnimatorPointOfView.Back hAngle = hAngle + 180 + yaw; case frost.Animator.AnimatorPointOfView.Left hAngle = hAngle - 90 + yaw; case frost.Animator.AnimatorPointOfView.Right hAngle = hAngle + 90 + yaw; case frost.Animator.AnimatorPointOfView.TopSouthEast hAngle = hAngle + 225; vAngle = vAngle + 45; case frost.Animator.AnimatorPointOfView.TopFrontLeft hAngle = hAngle + 225 + yaw; vAngle = vAngle + 45; end if obj.pov ~= frost.Animator.AnimatorPointOfView.Free view(obj.axs, hAngle, vAngle); end end function Draw(obj, t, x) obj.display.update(x); [center, radius, ~] = GetCenter(obj, t, x); delete(obj.text); obj.text = text(center(1)-radius/2,center(2),center(3)+radius-0.2,['t = ',sprintf('%.2f',t)]); %#ok<CPROPLC> obj.text.FontSize = 14; obj.text.FontWeight = 'Bold'; obj.text.Color = [0,0,0]; % set(obj.text); drawnow; end function [center, radius, yaw] = GetCenter(obj, t, x) center = [0,0,0]; radius = 3; yaw = 0; end function x = GetData(obj, t) t_start = obj.t_all(1); t_end = obj.t_all(end); delta_t = t_end - t_start; if t < t_start || t > t_end val = floor((t - t_start) / delta_t); t = t - val * delta_t; end if t < t_start t = t_start; elseif t > t_end t = t_end; end n = length(obj.t_all); x = obj.x_all(:, 1); % Default a = 1; b = n; while (a <= b) c = floor((a + b) / 2); if t < obj.t_all(c) x = obj.x_all(:, c); b = c - 1; elseif t > obj.t_all(c) a = c + 1; else x = obj.x_all(:, c); break; end end end end events newTimeStep playStateChanged reachedEnd end end
function pdMat = randPD(n) [u,~,~] = svd(rand(n,n)); pdMat = u * diag(rand(n,1)) * u';
% history points for smoothWavyWall % %-----------------------------------------------------% % points nx=10; % nx=number of x-points ny=100; % ny=number of y points per x-location h=0.50; % height to go up to in Y casename='smoothWavyWall'; %-----------------------------------------------------% % get x locations from maass % fil = 'maass/dat'; xM = zeros(1,nx); for i=1:nx nameM = [fil,num2str(i,'%0.2d')]; xM(i) = dlmread(nameM,'',[0 3 0 3]); end xM(end) = xM(end)-1; x = xM + 2; %-----------------------------------------------------% % geometry [x,y]=meshgrid(x,linspace(0,h,ny)); [x,y,xw,yw] = wavyWall(x,y,casename); x = reshape(x,[nx*ny,1]); y = reshape(y,[nx*ny,1]); z = 0*x; %-----------------------------------------------------% % create file casename.his format long A = [x,y,z]; casename=[casename,'.his']; fID = fopen(casename,'w'); fprintf(fID, [num2str(nx*ny) ' !=number of monitoring points\n']); dlmwrite(casename,A,'delimiter',' ','-append'); type(casename)
% Reads multiple folders recorded on same/different days and produces % matrix with all sessions and folders recorded % % Revision history: % % v1.0 September 1, 2016. Basic script prepared. % Data file is format: % [subjectName_Date_chChannelNumber_uUnitNumber_unitType.mat] % (for example: aq_20161231_ch1_u1_m.mat); function [y] = get_path_spikes_v10 (path1, subject_name) % Determine how many files with the subject name are there index_cur_dir = dir(path1); % Initialize output variables temp_ind_subj = cell(length(index_cur_dir), 1); temp_ind_dates = NaN(length(index_cur_dir), 1); temp_ind_channels = NaN (length(index_cur_dir), 1); temp_ind_units = NaN (length(index_cur_dir), 1); temp_ind_unit_class = cell (length(index_cur_dir), 1); %% Parse the file name % Temporary variables initiated fnames = cell (length(index_cur_dir), 1); % Temporary file name used fdates = cell (length(index_cur_dir), 1); % Dates in string format will be stored here % Extract file names & their sizes for i=1:length(index_cur_dir) fnames{i} = index_cur_dir(i).name; end %============ % Subject names %============ % Extract subject names for i=1:length(fnames) if length(fnames{i}) > length(subject_name) if strcmp(fnames{i}(1:length(subject_name)), subject_name) temp_ind_subj{i} = subject_name; % Save subject name fnames{i}(1:length(subject_name)) = []; else temp_ind_subj{i} = NaN; % Save subject name fnames{i} = NaN; end else temp_ind_subj{i} = NaN; fnames{i}=[]; % Discard if folder doesnt start with subject name end end % If there is a dash, extract it for i=1:length(fnames) if length(fnames{i}) > 1 if strcmp(fnames{i}(1), '_') fnames{i}(1) = []; end end end %=============== % Date detection %=============== % Detect chanel number for i=1:length(fnames) if length(fnames{i}) >= 2 [a,b] = strsplit(fnames{i}, '_'); % Find delimiter if ~isempty(b) % If delimiter exists if str2num(a{1}) temp_ind_dates(i) = str2num(a{1}); fnames{i}(1:length(a{1})) = []; else temp_ind_dates(i)=NaN; fnames{i}=NaN; end else temp_ind_dates(i)=NaN; fnames{i} = NaN; end end end % If there is a dash, extract it for i=1:length(fnames) if length(fnames{i}) > 1 if strcmp(fnames{i}(1), '_') fnames{i}(1) = []; end end end %================= % Channel detection %================= % Detect if next is channel notation for i=1:length(fnames) if length(fnames{i}) >= 2 if strcmp(fnames{i}(1:2), 'ch') fnames{i}(1:2) = []; else fnames{i}=NaN; end else fnames{i} = NaN; end end % Detect chanel number for i=1:length(fnames) if length(fnames{i}) >= 2 [a,b] = strsplit(fnames{i}, '_'); % Find delimiter if ~isempty(b) % If delimiter exists if str2num(a{1}) temp_ind_channels(i) = str2num(a{1}); fnames{i}(1:length(a{1})) = []; else temp_ind_channels(i) = NaN; fnames{i}=NaN; end else temp_ind_channels(i) = NaN; fnames{i} = NaN; end end end % If there is a dash, extract it for i=1:length(fnames) if length(fnames{i}) > 1 if strcmp(fnames{i}(1), '_') fnames{i}(1) = []; end end end %================= % Unit detection %================= % Detect if next is unit number notation for i=1:length(fnames) if length(fnames{i}) >= 1 if strcmp(fnames{i}(1), 'u') fnames{i}(1) = []; else fnames{i}=NaN; end else fnames{i} = NaN; end end % Detect unit number for i=1:length(fnames) if length(fnames{i}) >= 2 [a,b] = strsplit(fnames{i}, '_'); % Find delimiter if ~isempty(b) % If delimiter exists if str2num(a{1}) temp_ind_units(i) = str2num(a{1}); fnames{i}(1:length(a{1})) = []; else temp_ind_units(i) = NaN; fnames{i}=NaN; end else temp_ind_units(i) = NaN; fnames{i} = NaN; end end end % If there is a dash, extract it for i=1:length(fnames) if length(fnames{i}) > 1 if strcmp(fnames{i}(1), '_') fnames{i}(1) = []; end end end %================= % Unit classification detection %================= for i=1:length(fnames) if length(fnames{i}) >= 1 if strcmp(fnames{i}(1), 'u') || strcmp(fnames{i}(1), 's') || strcmp(fnames{i}(1), 'm') temp_ind_unit_class{i}(1) = fnames{i}(1); fnames{i}(1) = []; else temp_ind_unit_class{i}(1) = NaN; fnames{i}=NaN; end else temp_ind_unit_class{i}(1) = NaN; fnames{i} = NaN; end end %% For each date and session prepare a list of paths % Find unique days used unique_dates = unique(temp_ind_dates); ind=isnan(unique_dates); unique_dates(ind)=[]; index_cur_dir; % This is necessary input. Contains all file names and file ordering % Create output matrix file_name1 = cell(1); dates1 = []; subject1 = cell(1); unit1 = []; channel1 = []; unit_class1 = cell(1); path_temp1 = cell(1); for i=1:length(unique_dates) index1 = find(temp_ind_dates == unique_dates(i)); for j=1:length(index1); if length(file_name1)==1 && isempty(file_name1{1}) ind = 1; else ind = ind+1; end file_name1{ind,1} = index_cur_dir(index1(j)).name; dates1(ind,1) = temp_ind_dates(index1(j)); subject1{ind,1} = temp_ind_subj{index1(j)}; unit1(ind,1) = temp_ind_units(index1(j)); channels1(ind,1) = temp_ind_channels(index1(j)); unit_class1{ind,1} = temp_ind_unit_class{index1(j)}; path_temp1{ind,1} = [path1, index_cur_dir(index1(j)).name]; end end %% Output y.index_dates=dates1; y.index_unique_dates = unique_dates; y.index_subjects = subject1; y.index_file_name = file_name1; y.index_unit = unit1; y.index_channel = channels1; y.index_unit_type = unit_class1; y.index_path = path_temp1;
function summary=SG_analyse_bootstrap(N,restricted) if nargin==0 N = 1000 ; end %nDS = numel(dir(sprintf('../results/bootstrap/data*_r%d.mat',restricted)))/45 ; nDS = 100;% %effect_types = {'dynamics','choiceUpdate','successUpdate','forceUpdate'}; % for each random group parfor iB=1:N rng('shuffle'); % draw 18 random subjects dummy_group = randperm(nDS,18); %load data [summary_group]=SG_analyse('bootstrap',restricted,dummy_group); summary_sub(iB) = summary_group ; end summary.all = summary_sub; % for iT=1:numel(effect_types) % % MC = [summary_sub.MC]; % % Ef = struct.extract([MC.(effect_types{iT})],'Ef'); % xp = struct.extract([MC.(effect_types{iT})],'xp'); % % summary.(effect_types{iT}).Ef_mean = mean(Ef); % summary.(effect_types{iT}).Ef_std = std(Ef); % % summary.(effect_types{iT}).xp_mean = mean(xp); % summary.(effect_types{iT}).xp_std = std(xp); % % end % % effect_param = fields(summary_sub(1).BMA.significance); % for iT=1:numel(effect_param) % BMA = [summary_sub.BMA]; % % pp = struct.extract([BMA.significance],effect_param{iT}) ; % summary.(effect_param{iT}).p_mean = mean(pp); % summary.(effect_param{iT}).p_std = std(pp); % % val = struct.extract([BMA.effects],effect_param{iT}) ; % summary.(effect_param{iT}).val_mean = mean(val); % summary.(effect_param{iT}).val_std = std(val); % end
%% *_Branin Function_* % % Syntax: % Y = UQ_BRANIN(X) % Y = UQ_BRANIN(X,P) % % The model contains M=2 (independent) random variables (X=[X_1,X_2]) % and 6 scalar parameters of type double (P=[a,b,c,r,s,t]). % % Input: % X N x M matrix including N samples of M stochastic parameters % P vector including parameters % by default: a = 1; b = 5.1/(2*pi)^2; c = 5/pi; r = 6; s = 10; % t = 1/(8*pi); % % Output/Return: % Y column vector of length N including evaluations using branin % function % %%% function Y = uq_branin(X,P) %% Check % narginchk(1,2) assert(size(X,2)==2,'only 2 input variables allowed') %% Evaluation % % $$f(\mathbf{x}) = a(x_2 - bx_1^2 + cx_1 - r)^2 +s(1-t)\cos(x_1) +s$$ % if nargin==1 % Constants a = 1; b = 5.1/(2*pi)^2; c = 5/pi; r = 6; s = 10; t = 1/(8*pi); Y = a*(X(:,2) - b*X(:,1).^2 + c*X(:,1) - r).^2 + s*(1-t)*cos(X(:,1)) + s; end if nargin==2 Y = P(1)*(X(:,2) - P(2)*X(:,1).^2 + P(3)*X(:,1) - P(4)).^2 + P(5)*(1-P(6))*cos(X(:,1)) + P(5); end end
function varargout = colorSelector(varargin) % COLORSELECTOR MATLAB code for colorSelector.fig % COLORSELECTOR, by itself, creates a new COLORSELECTOR or raises the existing % singleton*. % % H = COLORSELECTOR returns the handle to a new COLORSELECTOR or the handle to % the existing singleton*. % % COLORSELECTOR('CALLBACK',hObject,eventData,handles,...) calls the local % function named CALLBACK in COLORSELECTOR.M with the given input arguments. % % COLORSELECTOR('Property','Value',...) creates a new COLORSELECTOR or raises the % existing singleton*. Starting from the left, property value pairs are % applied to the GUI before colorSelector_OpeningFcn gets called. An % unrecognized property name or invalid value makes property application % stop. All inputs are passed to colorSelector_OpeningFcn via varargin. % % *See GUI Options on GUIDE's Tools menu. Choose "GUI allows only one % instance to run (singleton)". % % See also: GUIDE, GUIDATA, GUIHANDLES % Edit the above text to modify the response to help colorSelector % Last Modified by GUIDE v2.5 25-Aug-2019 19:52:32 % Begin initialization code - DO NOT EDIT gui_Singleton = 1; gui_State = struct('gui_Name', mfilename, ... 'gui_Singleton', gui_Singleton, ... 'gui_OpeningFcn', @colorSelector_OpeningFcn, ... 'gui_OutputFcn', @colorSelector_OutputFcn, ... 'gui_LayoutFcn', [] , ... 'gui_Callback', []); if nargin && ischar(varargin{1}) gui_State.gui_Callback = str2func(varargin{1}); end if nargout [varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:}); else gui_mainfcn(gui_State, varargin{:}); end % End initialization code - DO NOT EDIT % --- Executes just before colorSelector is made visible. function colorSelector_OpeningFcn(hObject, eventdata, handles, varargin) % This function has no output args, see OutputFcn. % hObject handle to figure % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % varargin command line arguments to colorSelector (see VARARGIN) % Choose default command line output for colorSelector handles.output = hObject; handles.choose = [0, 0, 0]; imshow(ones(500,500,3)) % Update handles structure guidata(hObject, handles); % UIWAIT makes colorSelector wait for user response (see UIRESUME) % uiwait(handles.figure1); % --- Outputs from this function are returned to the command line. function varargout = colorSelector_OutputFcn(hObject, eventdata, handles) % varargout cell array for returning output args (see VARARGOUT); % hObject handle to figure % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) % Get default command line output from handles structure varargout{1} = handles.output; % --- Executes on button press in pushbutton1. function pushbutton1_Callback(hObject, eventdata, handles) % hObject handle to pushbutton1 (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) colorImage =ColorImage(10,10,50,50,[1,1, 0]); colorImage.setChoose(handles.choose); %gca(handles.axes1); colorImage.plot(true); handles.colorImage = colorImage; guidata(hObject, handles); % --- Executes on button press in checkboxRed. function checkboxRed_Callback(hObject, eventdata, handles) % hObject handle to checkboxRed (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) choose = handles.choose; choose(1) = get(hObject, 'Value'); handles.colorImage.setChoose(choose); if(any(choose)) handles.colorImage.plot(false); else handles.colorImage.plot(true); end handles.choose = choose; guidata(hObject, handles); % Hint: get(hObject,'Value') returns toggle state of checkboxRed % --- Executes on button press in checkboxGreen. function checkboxGreen_Callback(hObject, eventdata, handles) % hObject handle to checkboxGreen (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) choose = handles.choose; choose(2) = get(hObject, 'Value'); handles.colorImage.setChoose(choose); if(any(choose)) handles.colorImage.plot(false); else handles.colorImage.plot(true); end handles.choose = choose; guidata(hObject, handles); % Hint: get(hObject,'Value') returns toggle state of checkboxGreen % --- Executes on button press in checkboxBlue. function checkboxBlue_Callback(hObject, eventdata, handles) % hObject handle to checkboxBlue (see GCBO) % eventdata reserved - to be defined in a future version of MATLAB % handles structure with handles and user data (see GUIDATA) choose = handles.choose; choose(3) = get(hObject, 'Value'); handles.colorImage.setChoose(choose); if(any(choose)) handles.colorImage.plot(false); else handles.colorImage.plot(true); end handles.choose = choose; guidata(hObject, handles); % Hint: get(hObject,'Value') returns toggle state of checkboxBlue
ps = ProcessDays(psparse,'IntraGroup','Cells',ndresp.SessionDirs,'AnalysisLevel','AllIntraGroup'); InspectGUI(ps) figure; plot(ps) % compare mean lifetime sparseness and median population sparseness load sparsityobjs % get indices grouped according to site gind = groupDirs(sparseframe); % find sites with more than 1 cell since we need more than 1 cell to % compute population sparseness si = find(sum(~isnan(gind))>1) gind2 = gind(:,si); % get the sparseness data grouped according to site gdata = nanindex(sparseframe.data.sparsity,gind2) % find mean sparseness for each site gm = nanmean(gdata) pm = prctile(ps.data.psparse,50); % compute spree plot % first compute variance for each cell sfv = sparseframe.data.Values; sv = nanstd(sfv).^2; % now get variance for each site vdata = nanindex(sv,gind2); % find maximum variance mv = nanmax(vdata); % normalize by maximum variance vdata2 = flipud(sort(vdata ./ repmat(mv,size(vdata,1),1))); % data for 50 responsive cells plot([nan nan 0 1],vdata2(:,[1 3 4 10 13]),'go-','MarkerSize',16) hold on plot([nan 0 0.5 1],vdata2(:,[2 5 6 7 9 11 12]),'bo-','MarkerSize',16) plot([0 0.33 0.66 1],vdata2(:,8),'ro-','MarkerSize',16) hold off % data for 88 single units subplot(1,6,1) plot([nan nan nan nan nan 0 1],vdata2(:,[2 4 5 11 15 18 20]),'kx-','MarkerSize',16) set(gca,'TickDir','out','Box','off') % hold on subplot(1,6,2) plot([nan nan nan nan 0 0.5 1],vdata2(:,[3 10 12 14 16 21]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,3) plot([nan nan nan 0 0.33 0.66 1],vdata2(:,[1 17 19]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,4) plot([nan nan 0 0.25 0.50 0.75 1],vdata2(:,[6 13]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,5) plot([nan 0 0.2 0.4 0.6 0.8 1],vdata2(:,9),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,6) plot([0 0.16 0.33 0.5 0.66 0.83 1],vdata2(:,[7 8]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') % compute the area for 50 responsive cells vd2 = vdata2(:,[1 3 4 10 13]); vd3 = vdata2(:,[2 5 6 7 9 11 12]); vd4 = vdata2(:,8); a2 = 0.5 * (vd2(4,:) + 1); a3 = 0.5 * repmat(0.5,1,2) * (vd3(2:3,:) + vd3(3:4,:)); a4 = 0.5 * repmat(1/3,1,3) * (vd4(1:3,:) + vd4(2:4,:)); screarea = [a2 a3 a4]; mean(screarea) std(screarea) % compute the area for 88 single units vd2 = vdata2(:,[2 4 5 11 15 18 20]); vd3 = vdata2(:,[3 10 12 14 16 21]); vd4 = vdata2(:,[1 17 19]); vd5 = vdata2(:,[6 13]); vd6 = vdata2(:,9); vd7 = vdata2(:,[7 8]); a2 = 0.5 * (vd2(7,:) + 1); a3 = 0.5 * repmat(0.5,1,2) * (vd3(5:6,:) + vd3(6:7,:)); a4 = 0.5 * repmat(1/3,1,3) * (vd4(4:6,:) + vd4(5:7,:)); a5 = 0.5 * repmat(1/4,1,4) * (vd5(3:6,:) + vd5(4:7,:)); a6 = 0.5 * repmat(1/5,1,5) * (vd6(2:6,:) + vd6(3:7,:)); a7 = 0.5 * repmat(1/6,1,6) * (vd7(1:6,:) + vd7(2:7,:)); screarea = [a2 a3 a4 a5 a6 a7]; mean(screarea) std(screarea) % compute scree plots by first normalizing each cell to its mean so that we % can compare the variances across sites % create normalized scree plot sm = nanmean(sfv); % normalize the spike counts by the mean so the variances % between sites are equivalent sfv2 = sfv ./ repmat(sm,size(sfv,1),1); sv2 = nanstd(sfv2).^2; % now get variance for each site vdata3 = nanindex(sv2,gind2); % find maximum variance mv2 = nanmax(vdata3); % normalize by maximum variance vdata4 = flipud(sort(vdata3 ./ repmat(mv2,size(vdata3,1),1))); % plot for 50 responsive cells plot([nan nan 0 1],vdata4(:,[1 3 4 10 13]),'k:') hold on plot([nan 0 0.5 1],vdata4(:,[2 5 6 7 9 11 12]),'k--') plot([0 0.33 0.66 1],vdata4(:,8),'k-') hold off % plot for 88 single units subplot(1,6,1) plot([nan nan nan nan nan 0 1],vdata4(:,[2 4 5 11 15 18 20]),'kx-','MarkerSize',16) set(gca,'TickDir','out','Box','off') % hold on subplot(1,6,2) plot([nan nan nan nan 0 0.5 1],vdata4(:,[3 10 12 14 16 21]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,3) plot([nan nan nan 0 0.33 0.66 1],vdata4(:,[1 17 19]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,4) plot([nan nan 0 0.25 0.50 0.75 1],vdata4(:,[6 13]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,5) plot([nan 0 0.2 0.4 0.6 0.8 1],vdata4(:,9),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') subplot(1,6,6) plot([0 0.16 0.33 0.5 0.66 0.83 1],vdata4(:,[7 8]),'kx-','MarkerSize',16) set(gca,'YTickLabel','','TickDir','out','Box','off') % compute the area for 50 responsive cells vd2a = vdata4(:,[1 3 4 10 13]); vd3a = vdata4(:,[2 5 6 7 9 11 12]); vd4a = vdata4(:,8); a2a = 0.5 * (vd2a(4,:) + 1); a3a = 0.5 * repmat(0.5,1,2) * (vd3a(2:3,:) + vd3a(3:4,:)) a4a = 0.5 * repmat(1/3,1,3) * (vd4a(1:3,:) + vd4a(2:4,:)) screarea2 = [a2a a3a a4a]; mean(screarea2) std(screarea2) % compute the area for 88 single units vd2a = vdata4(:,[2 4 5 11 15 18 20]); vd3a = vdata4(:,[3 10 12 14 16 21]); vd4a = vdata4(:,[1 17 19]); vd5a = vdata4(:,[6 13]); vd6a = vdata4(:,9); vd7a = vdata4(:,[7 8]); a2a = 0.5 * (vd2a(7,:) + 1); a3a = 0.5 * repmat(0.5,1,2) * (vd3a(5:6,:) + vd3a(6:7,:)); a4a = 0.5 * repmat(1/3,1,3) * (vd4a(4:6,:) + vd4a(5:7,:)); a5a = 0.5 * repmat(1/4,1,4) * (vd5a(3:6,:) + vd5a(4:7,:)); a6a = 0.5 * repmat(1/5,1,5) * (vd6a(2:6,:) + vd6a(3:7,:)); a7a = 0.5 * repmat(1/6,1,6) * (vd7a(1:6,:) + vd7a(2:7,:)); screarea2 = [a2a a3a a4a a5a a6a a7a]; mean(screarea2) std(screarea2)
function [dat] = mainFrequencyAnalysis(dat); %H1 Line -- loop to calculate on frequency analysis on a structure array %Help Text -- %input requirements: dat is a structure array (output of impHYDAT) % %output details: dat, is the same arry entered in output with a filed that % contain the frequency analysis. The analysis is run for % each flow station. %Author: Laurence Chaput-Desrochers %date:august 19th 2013 %************************************************************************** nbFiles = size(dat,1); %compute log-PearsonIII frequency analysis %************************************************************************** for n = 1:nbFiles; [out,tableN] = logPearsonIIIgeV2(dat(n,1).discharges,dat(n,1).year); dat(n,1).freqDataPearson = tableN; dat(n,1).logPearsonIII = out; clear out tableN end%end loop n clear n %************************************************************************** end%end of function mainFrequancyAnalysis
function [ret] = showPossible(i_x, i_y) global board; thisPiece = board(i_x,i_y).possible; thisPiece(isnan(thisPiece)) = 0; ret = thisPiece; end
%Author : Vignesh Waren Sunder %Signal Digitisation and Reconstruction %Coded using MATLAB R2019a- academic use %GitHub: vicky-ML close all clc %Generating Signal t = 0:0.001:2; % set the time domain range (1 to 2 with 1000 steps) f1 = 6; f2 = 9; x = sin(2*pi*f1*t)+sin(2*pi*f2*t); %Sanalog figure subplot(211) plot(t,x); title("Sanalog") xlabel("Time(s)"); Fs = 100; %Our Sampling Frequency Ts = 1/Fs; n = 0:Ts:2; %set the time domain range x_sampled= sin(2*pi*f1*n)+sin(2*pi*f2*n); %Ssam subplot(212) stem(n,x_sampled,'fill'); title("Ssam"); xlabel("Time(s) Fs=100Hz"); %Quantisation maxx= max(x_sampled); %max value of Ssam minn = min(x_sampled); % min value of Ssam range = (maxx-minn)/16; % getting range with appropriate quatisation level %we perform "for" loop to get the y_axis value for each step in x_axis and %then round it to the nearest integer for n1 = 1:length(x_sampled) x_quant=round((x_sampled-minn)/range)*range+minn; end %Signal to quantisation noise ratio powerinsig = x_sampled.^2; %Power in Signal powerinnoise = (x_sampled-x_quant).^2; %Power in Noise sqnr = 10*log10 (powerinsig/powerinnoise); %SQNR figure plot(t,x); hold stairs(n,x_quant) title('Quantised') xlabel("Time(s)"); %fft %The value of Fs changes. Fs=100 for FFT of Ssamp | Fs=1000 for FFT of %Sanalog. Kindly change the value when running the code. xft = fft(x_sampled(1:length(x_sampled)-1)); %perform fft with a point delay(-1) arr=0:Fs/(2*Fs):Fs-Fs/(2*Fs); %To get the range of the frequency spectrum %(arr is 0 to 200 with 2000 steps) gives for one cycle %We generated and plotted Frequency spectrum only for One Cycle so that we %can easily perform FFTShift and extract the required frequency component xabs=abs(xft); %gives the magnitude figure subplot(211) plot(arr,xabs/Fs) title('FFT of Sanalog') xlabel("Frequency(Hz)"); ylabel("Magnitude"); xxftshift = fftshift(xft); %perform FFTShift subplot(212) plot(arr-0.5*Fs, abs(xxftshift)/Fs)%plot fftshift with desired range title("FFTShift of Sanalog"); xlabel("Frequency(Hz)"); ylabel("Magnitude"); %Low Pass Filter lpfilter = zeros (1,length(xxftshift)); %zero magnitude for this range for number1 = 1:length(xxftshift) %magnitude is 1 for these range of points if number1< 0.5*length(xxftshift) lpfilter(number1)=1; end end figure plot(arr,xabs/Fs.*lpfilter) %perform cyclic convolution by using point wise %multiplicaiton in Frequency domain. The required frequency signal is %extracted title("filtered spectrum Sanalog"); xlabel("Frequency(Hz)"); ylabel("Magnitude"); figure iff = ifft(xft/Fs.*lpfilter*2); %perform ifft with extracted signal from %the lpfilter %iff = ifft(xft/Fs); plot(n(1:length(n)-1),iff*Fs, 'r') %Plot the reconstructed Signal hold %stem(n,x_sampled,'fill'); %to get a stemed reconstructed signal output %hold plot(t,x, 'k') title("Reconstructed Signal and Sanalog") xlabel("Time (s)"); ylabel("Amplitude");
clc clear all; close all; global epsilon; global p; %ep = 0.001; eps = linspace(3,3,1); iter = linspace(1,1,1); epsilon = 0.01; op = odeset('reltol',1e-9,'abstol',1e-11); for i=iter figure(i) hold on; p=eps(i); epsilon = 0.1/(p^3) val = epsilon*(p^3) [t,y] = ode45('mms2d',[0,1000],0.01,op); plot(t,y,'DisplayName','p ='+string(p)); mean(y) %[t,y] = ode45('mms2d_avg',[0,400],0.01,op); %plot(t,y,'DisplayName','p_avg ='+string(p),'linewidth',2); %hold off; title('$\epsilon$ = '+string(epsilon),'interpreter','latex') xlabel('t') ylabel('x') legend end
function h = mrpln02b_PlotPairwiseFactorsInFG(fg, values, plot_flags) factor_color = plot_flags.factor_color; factor_line_width = plot_flags.factor_line_width; factor_linestyle = plot_flags.factor_linestyle; h = []; odometryPathXCoords = []; odometryPathYCoords = []; for i=0:double(fg.size)-1 if ~fg.exists(i), continue; end f = fg.at(i); keys = f.keys(); delta = abs(double(gtsam.mrsymbolIndex(keys.front()))-double(gtsam.mrsymbolIndex(keys.back()))); if delta == 1 % continue; % skip odometry factors for better visibility factor_linestyle = '-'; else continue % skip LC factors factor_linestyle = '--'; end edge = mrpln02c_PlotPairwiseFactor(f, values, factor_color, factor_line_width, factor_linestyle); h = [h edge]; if delta == 1 odometryPathXCoords = [odometryPathXCoords edge.XData]; odometryPathYCoords = [odometryPathYCoords edge.YData]; hold on; plot(edge.XData, edge.YData, 'Color', plot_flags.factor_color, 'LineWidth', plot_flags.factor_line_width+2) end end hold on; plot(odometryPathXCoords, odometryPathYCoords, '.', 'Color', plot_flags.factor_color, 'MarkerSize', 30) % mark observation poses
function second = Q_ra(x_1,x_1_r,x_2,x_2_r,P_a,P_b) % Eq. 4 in Ref.[2] second=P_b*(x_2-x_2_r)/((x_1-x_1_r+P_a*(x_2-x_2_r))^2+(P_b*(x_2-x_2_r))^2); end
%NAME: Jadeja Jaydevsinh G. %ROLL NO: 12MEC08 %BATCH: ME-2nd sem,2013 %DEPARTMENT: Electronics & Communication. %SUBJECT: ADC %1.Aim: To verify the nyquist theoram. close all; clear all; t=-10:0.01:10; T=8; fm=1/T; x=sin(2*pi*fm*t); subplot(311);plot(t,x); n=-10:1:10; j=ones(1,length(n)); subplot(312);stem(n,j);axis([-10 10 0 2]); xn=sin(2*pi*n*fm); subplot(313);stem(n,xn);hold on;plot(n,xn);
%% This code is the code of the part 3 %% Calcul of matrix Ks and Ms % function Part3(Samcef) close all run Dimensions run Geometry nElementT = 0; for i = 1 : 45 nElementT = nElementT + element(i,5); end nodeList = zeros(nElementT,3); basic_dofs = zeros(29,6); for i = 1:29 basic_dofs(i,:) = [(i-1)*6+1 (i-1)*6+2 (i-1)*6+3 (i-1)*6+4 (i-1)*6+5 (i-1)*6+6]; end nBeamG = 12; % number of big beams nBeamM = 29; % number of middel beams nBeamP = 4; % number of little beam for i = 1:45 dofList(((i-1)*element(i,5))+1,:) = basic_dofs(element(i,1),:); nodeList(((i-1)*element(i,5))+1,:) = Node(element(i,1),:); for j = 1:element(i,5)-2 dofList((element(i,5)*(i-1))+1+j,:) = (29*6+((i-1)*(element(i,5)-2)+j-1)*6)+(1:6); nodeList((element(i,5)*(i-1))+1+j,:) = Node(element(i,1),:) + ((Node(element(i,2),:)-Node(element(i,1),:))./(element(i,5)-1)*j); end dofList((i*element(i,5)),:) = basic_dofs(element(i,2),:); nodeList((i*element(i,5)),:) = Node(element(i,2),:); end for i = 1:45 for j = 1:element(i,5)-1 Kel(:,:,(i-1)*(element(i,5)-1)+j) = matriceRaideur(element(i,3), element(i,4), nodeList((i-1)*element(i,5)+j,:), nodeList((i-1)*element(i,5)+(j+1),:)); Mel(:,:,(i-1)*(element(i,5)-1)+j) = matriceMasse(element(i,3), element(i,4), nodeList((i-1)*element(i,5)+j,:), nodeList((i-1)*element(i,5)+(j+1),:)); end end % Assembly of the structural and mass matrices nNode = 0; for i = 1 : 45 nNode = nNode + element(i,5) - 2; end nNode = nNode + 29; Ks = zeros(nNode*6,nNode*6); Ms = zeros(nNode*6,nNode*6); for i = 1:45 for j = 1:element(i,5)-1 dofs1 = dofList((i-1)*element(i,5)+j,:); dofs2 = dofList((i-1)*element(i,5)+(j+1),:); locel = [dofs1 dofs2]; for k=1:12 for l=1:12 Ks(locel(k), locel(l)) = Ks(locel(k), locel(l)) + Kel(k,l,(i-1)*(element(i,5)-1)+j); Ms(locel(k), locel(l)) = Ms(locel(k), locel(l)) + Mel(k,l,(i-1)*(element(i,5)-1)+j); end end end end fixedDof = [10*6+1:10*6+6 11*6+1:11*6+6 12*6+1:12*6+6 13*6+1:13*6+6]; Ks(fixedDof, :) = []; Ks(:, fixedDof) = []; Ms(fixedDof, :) = []; Ms(:, fixedDof) = []; nb_vp = 10; % nombre de modes propres et valeurs propres à afficher clear Ms clear Ks Ms = [1 0; 0 1] Ks =[20000 -10000; -10000 20000] C = [3 -1; -1 3]; [V, D]=eigs(Ks, Ms, 2, 'sm'); w = sqrt(diag(D)); % [rad/s] f = w /2/pi; V2 = V; %% Calcul of parameters A = zeros(2, 2); epsR = zeros(2,1); epsR(:,1) = 0.005; result = zeros(2,1); gamma = zeros(2,1); mu = zeros(2,1); V2t = V2'; omega = 2*pi; result = A\epsR; %C = result(1)*Ks + result(2)*Ms; %% Codes run ComparisonStep run ComparisonSamcef run FRF
% Marco Bettiol - 586580 - BATCH SIZE ESTIMATE % % CBT Simple Test % % This script implements a simulation of the estimate obtained % using CBT in a batch of size n. % % Nodes are initially uniformily picked-up in the interval [0,1) clear all; close all; clc; n=16; % batch size disp(['Size :' int2str(n)]); nodes=rand(n,1); % virtual node with value 1 to get easier search algorithm % among the nodes % asc sorting nodes=[sort(nodes); 1]; if (n<2) error('BRA must start with a collision'); end % CBT Simulation % true if we got a success in the last transmission lastwassuccess=false; %false to end CBT waitforconsecutive=true; imax=length(nodes); %index of the first node in the batch imin=1; % index of the first node in the batch xmin=0; % starting interval [0,1/2) xlen=1/2; % we suppose a collision already occurred. while (waitforconsecutive) [e,imin,imax]=cbtsplit(nodes,imin,imax,xmin,xlen); % update next analyzed interval if(e==1) xmin=xmin+xlen; %xlen=xlen; elseif (e==0) xmin=xmin+xlen; xlen=xlen/2; else %xmin=xmin; xlen=xlen/2; end if (lastwassuccess==true && e==1) disp(' '); disp('CBT completed :'); disp(['Estimate :' num2str(1/xlen)]); disp(['Last node transmitting :' int2str(imin-1)]); waitforconsecutive=false; end if(e==1) lastwassuccess=true; else lastwassuccess=false; end end % DEBUG % estimate is given by the first serie of descending differences in the % nodes ID's dif=-1*ones(n-1,1); %negative init for i=1:n-1 dif(i)=nodes(i+1)-nodes(i); end nodes dif
function [A,X,res] = DN_NMF(Y,opts) %function [A,X,res] = nmf_qp_cg(Y,A,X,MaxIter,tol_c,lambda) % The Damped Newton (DN) algorithm developed by % Copyrirght R. Zdunek, A.-H. Phan, and A. Cichocki % % DN %% line 61 is changed defopts = struct('NumOfComp',[],'A0',[],'X0',[],'MaxIter',500,'Tol',1e-6,'lambda',1); % Theta maps checkStep. if ~exist('opts','var') opts = struct; end [r,A,X,MaxIter,tol_c,lambda] ... =scanparam(defopts,opts); if isempty(r) r=size(Y,2); end [I,T] = size(Y); epsil = eps; Y = max(eps,Y); n = 0; k = 0; if isempty(A) A=rand(I,r); end if isempty(X) X=rand(r,T); end % Number of components r = size(A,2); % Scaling da = sqrt(sum(A.^2,1))+eps; A = bsxfun(@rdivide,A,da); X = bsxfun(@times,da',X); k = 0; res = []; c = []; while k < MaxIter k = k + 1; % if ~mod(k,50) % disp(['Iteration: ',num2str(k)]); % end % UPDATES FOR A % ====================================================================== % Memory prelocation M = I*r; Q = spalloc(M,M,M*r); Q_tilde = spalloc(M,M,M*r); B = X*X'; C = X*Y'; Q = kron(speye(I),B); % Hessian of D(Y||AX) with respect to A Qd = zeros(M,1); At = A'; at = At(:); atp = at; GAt = B*A' - C; active = find((at <= epsil) & (GAt(:) > 0)); % active set degradated = find((at <= epsil) & (abs(GAt(:)) < epsil), 1); if ~isempty(degradated) disp(['Degenerate: ',num2str(k)]); end inactive = setdiff(1:length(at),active); % inactive set if ~isempty(active) a_tilde = at; a_tilde(active) = 0; c_tilde = C(:); c_tilde(active) = 0; Q_tilde = Q; Q_tilde(:,active) = 0; Q_tilde(active,:) = 0; Qd(active) = 1; Q_tilde = Q_tilde + spdiags(Qd,0,M,M) + lambda*speye(M); else a_tilde = at; Q_tilde = Q + lambda*speye(M); c_tilde = C(:); end % CGS tol = 1e-6 + exp(-k); at = zeros(M,1); Io = []; Ao = setdiff(1:M,Io); Qd = zeros(M,1); while ~isempty(Ao) [a_tilde,FLAG,RELRES,ITER] = pcg(Q_tilde,c_tilde,tol); Ao = find(a_tilde < 0); Q_tilde(Ao,:) = 0; Q_tilde(:,Ao) = 0; Q_tilde = Q_tilde + spdiags(Qd,0,M,M); c_tilde(Ao) = 0; end at = a_tilde; At = (max(epsil,reshape(at,r,I))); A = At'; da = eps+sqrt(sum(A.^2,1)); A = bsxfun(@rdivide,A,da); X = bsxfun(@times,da',X); % UPDATES FOR X % ====================================================================== HX = A'*A + 1e-12*eye(r); GX = HX*X - A'*Y; I_active = find((X < eps)&(GX > 0)); UX = GX; UX = HX\UX; UX(I_active) = 0; X = max(eps,X - UX); % Normalization % da = eps+sqrt(sum(A.^2,1)); % A = bsxfun(@rdivide,A,da); % X = bsxfun(@times,da',X); % dx = eps+sqrt(sum(X.^2,2)); X = bsxfun(@ldivide,dx,X); A = bsxfun(@times,dx',A); res(k) = norm(Y - A*X,'fro')/norm(Y,'fro'); % Stagnation breaking if k > 3 c(k) = res(k-1) - res(k); if (c(k)) < tol_c lambda = lambda/2; end if k > 30 & c(k) < 0 & c(k-1) < 0 & c(k-2) < 0 break end end end % while k
%clear % ---------------------------------------------------------------------- % % 実験設定。 % クエリ数を変更した際にはデータセットを新しく用意する必要があります。 % ---------------------------------------------------------------------- % %実験名 exName = 'ex1204stl'; %実験日 todaysDate = datetime('now','TimeZone','local','Format','yMMd'); % iExperiment: 何周目の実験か。パラメータ設定以外では使用しない for iExperiment = 1 % ---------------------------------------------------------------------- % %フォルダ(outputフォルダ & mファイル) [filepath, name, ext] = fileparts([mfilename('fullpath'),'.m']); Dir_kume = filepath; % モデルサイズの設定。有効な値は以下の通り。 % 16, 32, 40, 48, 56, 64, 80, 96 ModelSize = 40; % モデルサイズの設定。任意の値が有効。 imgSize = 64; % モデルタイプごとのクエリ数。基本5。 nQuery = 5; % データセットを新しく用意するかどうか。 % true - 用意する/false - 用意しない PrepareDataSet = false; % クエリ・データベースのGeodesic sphereのID。 DatabaseGeodesicID = 6; QueryGeodesicID = 6; % Geodesic: ID(1 〜 7)に対応する投影枚数。 % 例:Geodesic IDが1のとき,投影枚数は12枚。 nProjections = [12, 42, 92, 162, 252, 362, 492]; % ssht: ID(1001 〜 1016) generated by 'makePrjPointList_SSHT(L)' % mirror: ID(20XX) % nProjections = [1, 6, 15, 28, 45, 66, 91, 120, 153, 190, 231, 276, 325, 378, 435, 496] % % 1次元投影: ID に '-(マイナス)' を付ける generated by 'makePrjPointList_projection1d(gdid)' % 作成できた2次元投影を 'Projection1dFromProjection2d.m' に突っ込むと % マイナスを付ける前の ID の頂点に対応した1次元投影を求められる % % クエリ・データベースの部分空間の使用次元数。 rDatabase = 6; rQuery = 6; % 特徴の高周波成分削減回数。 nDivide = 0; % 結果を Excel に出力するためのパラメータ % 【対象のExcelファイルは開じる(開いてると書き込めない)】 % 【MacOS,Linux は非対応, 可能なら代わりにCSVファイルで出力される】 % Excel ファイル名(拡張子 '.xlsx' 忘れずに。上書き注意!) excelFileName = [exName, '_', char(todaysDate), '.xlsx']; % シート番号 sheet = iExperiment; % 書き込みを始めるセル % 例:[5,2] -> 'B5' から書き始める startCell = [1,1]; % 有効数字(Excelへの出力時、複素数のためにString型へ変換) precision = 6; % ---------------------------------------------------------------------- % current_path = pwd; cRetrieval = tic; % モデルタイプの名称とモデルタイプの数。 % 基本的にはClutch, Die, Gearの3種類。 % TypeName = {'Clutch'; 'Die'; 'Gear';}; % TypeName = {'clutch'; 'die'; 'gear';}; TypeName = {'clutch'; 'die'; 'gear'; 'mould'; 'hydraulic';}; % TypeName = {'clutch'; 'die'; 'gear'; 'mould'; 'hydraulic'; 'pvb'; 'steamtrap'; 'swashplate'; 'clutch_die'; 'die_gear'; 'gear_mould'; 'mould_clutch'; 'clutch_8'; 'keo'; 'morton';}; nModelType = length(TypeName); % 投影計算 & 処理時間計測 CulcProcessingTime_forSato(exName, DatabaseGeodesicID, QueryGeodesicID, imgSize, TypeName); % 投影の統合 [QueryProjection,DatabaseProjection] = InteglatePrj_forSato(exName, DatabaseGeodesicID, QueryGeodesicID, imgSize, TypeName); % %{ DatabaseName = fieldnames(DatabaseProjection); QueryName = fieldnames(QueryProjection); cDatabaseExtraction = tic; % データベース側の特徴抽出処理 disp('データベース側の特徴抽出処理を実行中'); for iDatabase = 1 : length(DatabaseName) tDatabaseProjection = DatabaseProjection.(DatabaseName{iDatabase}); % 投影画像の重み付け DatabaseWeightedProjection = ... WeightingProjection(tDatabaseProjection); % 特徴抽出処理 DatabaseFeature.(DatabaseName{iDatabase}) = ... DescriptorForMSM(DatabaseWeightedProjection, nDivide); end ExperimentTime.DatabaseExtraction = toc(cDatabaseExtraction); fprintf('処理時間:%f[s]\n', ExperimentTime.DatabaseExtraction); % データベース側の部分空間の生成 cDatabaseSubspace = tic; disp('データベース側の部分空間生成処理を実行中'); for iDatabase = 1 : length(DatabaseName) tDatabaseFeature = DatabaseFeature.(DatabaseName{iDatabase}); LabelList = fieldnames(tDatabaseFeature); for iLabel = 1 : length(LabelList) [tDatabaseSubspace] = ... EVD(tDatabaseFeature.(LabelList{iLabel}), rDatabase); DatabaseSubspace.(DatabaseName{iDatabase}).(LabelList{iLabel}) = ... tDatabaseSubspace; end end ExperimentTime.DatabaseSubspace = toc(cDatabaseSubspace); fprintf('処理時間:%f[s]\n', ExperimentTime.DatabaseSubspace); cQueryExtraction = tic; % クエリ側の特徴抽出処理 disp('クエリ側の特徴抽出処理を実行中'); for iQuery = 1 : length(QueryName) tQueryProjection = QueryProjection.(QueryName{iQuery}); % 投影画像の重み付け QueryWeightedProjection = ... WeightingProjection(tQueryProjection); % 特徴抽出処理 QueryFeature.(QueryName{iQuery}) = ... DescriptorForMSM(QueryWeightedProjection, nDivide); end ExperimentTime.QueryExtraction = toc(cQueryExtraction); fprintf('処理時間:%f[s]\n', ExperimentTime.QueryExtraction); % クエリ側の部分空間の生成 cQuerySubspace = tic; disp('クエリ側の部分空間生成処理を実行中'); for iQuery = 1 : length(QueryName) tQueryFeature = QueryFeature.(QueryName{iQuery}); LabelList = fieldnames(tQueryFeature); for iLabel = 1 : length(LabelList) [tQuerySubspace] = ... EVD(tQueryFeature.(LabelList{iLabel}), rQuery); QuerySubspace.(QueryName{iQuery}).(LabelList{iLabel}) = ... tQuerySubspace; end end ExperimentTime.QuerySubspace = toc(cQuerySubspace); fprintf('処理時間:%f[s]\n', ExperimentTime.QuerySubspace); Similarity = zeros(length(QueryName), length(DatabaseName)); % 各モデルとクエリとの類似度を格納する配列。 ExperimentTime.Matching = 0; disp('モデル間の距離を相互部分空間法によって計算中'); for iQuery = 1 : length(QueryName) tQuerySubspace = QuerySubspace.(QueryName{iQuery}); for iDatabase = 1 : length(DatabaseName) tDatabaseSubspace = DatabaseSubspace.(DatabaseName{iDatabase}); % データベースモデルとの距離を計算 cMatching = tic; LabelList = fieldnames(tQuerySubspace); % 部分空間による距離計算 tSimilarity = 0; for iLabel = 1 : length(LabelList) U = tQuerySubspace.(LabelList{iLabel}); V = tDatabaseSubspace.(LabelList{iLabel}); tSimilarity = tSimilarity + MSM(U, V); end tSimilarity = tSimilarity / length(LabelList); Similarity(iQuery, iDatabase) = tSimilarity; ExperimentTime.Matching = ExperimentTime.Matching + toc(cMatching); end end fprintf('処理時間:%f[s]\n', ExperimentTime.Matching); % 正解率を算出する。 for iModelType = 1 : nModelType TrueModelIndex = 5 * (iModelType - 1) + (1:5)'; TopIndex = nQuery * (iModelType - 1) + 1; BottomIndex = nQuery * iModelType; tSimilarity = Similarity(TopIndex : BottomIndex, :); [~, MinIndex] = min(tSimilarity, [], 2); Accuracy.(TypeName{iModelType}) = sum(MinIndex == TrueModelIndex) / nQuery; end ExperimentTime.Total = toc(cRetrieval); fprintf('プログラム全体の処理時間:%f[s]\n', ExperimentTime.Total); % 結果を構造体としてまとめる。 ExperimentResult.SimilarityMatrix = Similarity; ExperimentResult.Accuracy = Accuracy; ExperimentResult.ExperimentTime = ExperimentTime; % データセットを構造体としてまとめる。 %DataSet.ShiftVector = ShiftVector; %DataSet.RotateVector = RotateVector; % 結果を保存する。 FileName = sprintf('Result_GDID_D%dQ%d_Query%d_%s', ... DatabaseGeodesicID, QueryGeodesicID, nQuery, exName); % save(FileName, 'ExperimentResult', 'DataSet'); save([Dir_kume, filesep, FileName], 'ExperimentResult'); % 結果を Excel にまとめる bodyStruct = ExperimentResult; % output structure headerStruct = struct; headerStruct.exName = exName; headerStruct.DatabaseGeodesicID = DatabaseGeodesicID; headerStruct.QueryGeodesicID = QueryGeodesicID; headerStruct.nQuery = nQuery; headerStruct.matfile = [FileName,'.mat']; savePath = [Dir_kume, filesep, excelFileName]; saveStructure(savePath,bodyStruct,headerStruct,startCell,sheet,precision); % ---------------------------------------------------------------------- % %} end % for iExpetiment
function [ array_stddev, array_mean ] = my_stddev( one_dim_array ) % pusing bro, ada obat nyamuk? % dapatkan panjang array n_array = length(one_dim_array); % dapatkan nilai mean array_mean = my_mean(one_dim_array); % dapatkan stddev kuadrat stddev_sum = 0; for i = 1:n_array stddev_sum = stddev_sum+((one_dim_array(i)-array_mean)^2); end array_stddev = sqrt(stddev_sum/n_array); end
%% A. Benchmark ngrid=8361; grouprankmtgppv7alls120g8361mars=zeros(ngrid,12); for ns=1:ngrid [transdat,lambda] = boxcox((senscoreMARSgppv7alls120g8361(ns,:)+1)'); Y = pdist(transdat); Z = linkage(Y,'ward'); [~,T] = dendrogram(Z,4); orimax=zeros(4,1); for i=1:4 orimax(i,1)=max(transdat(find(T==i))'); end [stmax,index]=sort(orimax); tT=T; for i=1:4 tT(find(T==index(i)),1)=5-i; end grouprankmtgppv7alls120g8361mars(ns,:)=tT; end % Group rank statistics grouprankfreqgppv7alls120g8361=zeros(4,12); for i=1:4 for j=1:12 for k=1:ngrid if (grouprankmtgppv7alls120g8361mars(k,j)==i) grouprankfreqgppv7alls120g8361(i,j)=grouprankfreqgppv7alls120g8361(i,j)+1; end end end end finalgrouprankgppv7alls120g8361=zeros(1,12); for i=1:12 finalgrouprankgppv7alls120g8361(i)=find(grouprankfreqgppv7alls120g8361(:,i)==max(grouprankfreqgppv7alls120g8361(:,i))); end finalgridportiongppv7alls120g8361=zeros(1,12); for i=1:12 finalgridportiongppv7alls120g8361(i)=grouprankfreqgppv7alls120g8361(finalgrouprankgppv7alls120g8361(i),i)/ngrid; end
%Q6* function [image, correctLabel, predictedLabel] = PredictTest (n) allTrainImages = loadMNISTImages('./train-images.idx3-ubyte'); allTrainLabels = loadMNISTLabels('./train-labels.idx1-ubyte'); mdl = fitcknn(allTrainImages', allTrainLabels); allTestImages = loadMNISTImages('./t10k-images.idx3-ubyte'); allTestLabels = loadMNISTLabels('./t10k-labels.idx1-ubyte'); image = allTestImages(:, n); predictedLabel = predict(mdl, image'); correctLabel = allTestLabels(n); image = reshape(image, 28, 28); end
function s = accuracy(x,y,value) [ m, n ] = size(x); if ( n != 1) error('x is not a column vector'); endif [ m, n ] = size(y); if ( n != 1) error('y is not a column vector'); endif if (length(x) != length(y)) error('inputs column vectors x and y should have the same length'); endif R = x >= value; TP = sum(R.*y); TN = sum((1-R).*(1-y)); s = (TP+TN)/m; endfunction
function [ digit ] = FindDigit( I ) dir = 'letters/'; ext = '.bmp'; max_coeff = -2; for i = 1:10 im = imread([dir sprintf('%d', i-1) ext]); coeff = corr2(im, I); if coeff > max_coeff max_coeff = coeff; digit = i-1; end end end
function saveAsAvi_AI_ratio(matFilesPath, matFilesPrefix, aviFilesPath, aviFilesPrefix, ... radius, nSides) %% saving a video for changing activator/inhibitor levels % in the colony. files = dir([matFilesPath filesep matFilesPrefix '*.mat']); nTimePoints = numel(files); videoPath = [aviFilesPath filesep aviFilesPrefix '_' matFilesPrefix]; %% fig = figure; set(fig, 'Position', [68 800 500 400]); fullVideoPath = [videoPath '.avi']; v = VideoWriter(fullVideoPath); v.FrameRate = 5; open(v); counter = 1; for ii = 1:nTimePoints outputFile = [matFilesPath filesep matFilesPrefix '_t' int2str(ii) '.mat']; load(outputFile, 'storeStates'); if ii == 1 % get the colony boundary nSquares = size(storeStates,1); edgeLength = 1; lattice = false(nSquares); [~, ~, colonyState] = specifyColonyInsideLattice(lattice, radius, nSides); [colonyEdgeIdx] = specifyRegionWithinColony(colonyState, edgeLength); end timeSteps = size(storeStates,4); for kk = [1:timeSteps] % each file has multiple timesteps activator = storeStates(:,:,1,kk); inhibitor = storeStates(:,:,2,kk); values = inhibitor./activator; values(isnan(values)) = 0; values = values.*colonyState; imagesc(values); colorbar; caxis([0 5]); title(['activator/inhibitor' 't=' int2str(counter)]); counter = counter+1; % if component == 1 % caxis([0 1.1]); % elseif component < 3 % caxis([0 2.5]); % else % caxis([0 0.5]); % end hold on; plot(colonyEdgeIdx(:,2), colonyEdgeIdx(:,1), 'k.', 'MarkerSize', 8); axis off; drawnow; pause(0.01); M = getframe(fig); writeVideo(v,M); end end close(v); end
clear % For plotting h = figure; hold on; grid on; title('Earth temperature vs albedo'); xlabel('Temperature (K)'); ylabel('Albedo'); % Variables alpha = 0; % Albedo - describes how much sunlight Earth reflects. sigma = 5.67 * 1e-8; % Stefan-Boltzmann constant - total intensity. I = 344; % Incident - short-wave radiation from the sun. % Base function - balance between amount of emitted/absorbed energy: % (1 - alpha) * I = sigma * T^4 % Case 1: % When temperature on Earth is T1 = 235 K, % then alpha is 0.5 (high albedo - there is ice and clouds reflecting sun). alpha1 = 0.5; T1_mean = nthroot((1-alpha1)*I/sigma, 4); % 234 K line([200 250],[alpha1 alpha1],'LineWidth',1); % Case 2: % When temperature is between 250 and 270 K, then albedo has functional % dependency of (270 - T) / 40. After solving balance equation and summing % temperatures, then mean is 259 K. T2 = 250:1:270; alpha2 = (270 - T2) / 40; syms T; % Define symbol for solving equation eqnLeft = T; % Left side of equation eqnRight = nthroot((1-(270 - T) / 40)*I/sigma, 4); % Right side of equation initialGuess = 300; % Initial guess where solution converges to 260 K. T_mean = vpasolve(eqnLeft == eqnRight, T, initialGuess); % 260 K plot(T2, alpha2); % Case 3: % When temperature on Earth is T3 = 280 K, % then alpha is 0 (low albedo - no ice and clouds reflecting sun). alpha3 = 0; T3_mean = nthroot((1-alpha3)*I/sigma, 4); % 279 K line([270 300],[alpha3 alpha3],'LineWidth',1); % Save plot to file saveas(h, 'ex4','jpg'); % Summary: % In first case average temperature is 234 K, % in second case 259 K % and in third case 279 K.
End of preview. Expand in Data Studio
README.md exists but content is empty.
Downloads last month
4