markdown
stringlengths
0
1.02M
code
stringlengths
0
832k
output
stringlengths
0
1.02M
license
stringlengths
3
36
path
stringlengths
6
265
repo_name
stringlengths
6
127
As expected, this plot shows that all of the agreement metrics as well as PRMSE ranks the systems correctly since all of them are being computed against the same rater pair. Note that systems known to have better performance rank "higher". Step 3: Evaluate each systems against a different pair of ratersNow, we change...
# first let's get rater pairs within each category rater_pairs_per_category = df_rater_metadata.groupby('rater_category')['rater_id'].apply(lambda values: itertools.combinations(values, 2)) # next let's combine all possible rater pairs across the categories combined_rater_pairs = [f"{rater_id1}+{rater_id2}" for rater_...
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
Before we proceed, let's see how the different system categories are distributed across different rater pairs.
# create a dataframe from our rater pairs with h1 and h2 as the two columns df_rater_pairs_with_categories = pd.DataFrame(data=rater_pairs_for_systems, columns=['h1', 'h2']) # add in the system metadata df_rater_pairs_with_categories['system_id'] = df_system_metadata['system_id'] df_rater_pairs_with_categories['syste...
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
As the table shows, we see that systems in different categories were evaluated against rater pairs with different level of agreement. For example, 3 out of 5 systems in "low" category were evaluated against raters with "high" agreement. At the same time for systems in "medium" category, 3 out of 5 systems were evaluate...
# initialize an empty list to hold the metric values for each system ID metric_values_list = [] # iterate over each system ID for system_id, rater_id1, rater_id2 in zip(df_rater_pairs_with_categories['system_id'], df_rater_pairs_with_categories['h1'], ...
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
Now that we have computed the metrics, we can plot each simulated system's measured performance via each of the metrics against its known performance, as indicated by its system category.
# now create a longer version of this data frame that's more amenable to plotting df_metrics_different_rater_pairs_with_categories_long = df_metrics_different_rater_pairs_with_categories.melt(id_vars=['system_id', 'system_category'], ...
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
From this plot, we can see that only PRMSE values accurately separate the systems from each other whereas the other metrics are not able to do. Next, let's plot how the different systems are ranked by each of the metrics and also compare these ranks to the ranks from the same-rater scenario.
# get the ranks for the metrics df_ranks_different_rater_pairs = compute_ranks_from_metrics(df_metrics_different_rater_pairs_with_categories) # and now get a longer version of this data frame that's more amenable to plotting df_ranks_different_rater_pairs_long = df_ranks_different_rater_pairs.melt(id_vars=['system_cat...
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
As this plot shows, only the PRMSE metric is still able to rank the systems accurately whereas all the other metrics are not. We can also make another plot that shows a more direct comparison between $R^2$ and PRMSE.
# plot PRMSE and R2 ranks only df_r2_prmse_ranks_long = df_all_ranks_long[df_all_ranks_long['metric'].isin(['R2', 'PRMSE'])] ax = sns.boxplot(x='system_category', y='rank_diff', hue='metric', hue_order=['R2', 'PRMSE'], data=df_r2_prmse_ranks_long,) ax....
_____no_output_____
MIT
notebooks/ranking_multiple_systems.ipynb
EducationalTestingService/prmse-simulations
Hyperparameter tuning with Cloud ML Engine **Learning Objectives:** * Improve the accuracy of a model by hyperparameter tuning
import os PROJECT = 'cloud-training-demos' # REPLACE WITH YOUR PROJECT ID BUCKET = 'cloud-training-demos-ml' # REPLACE WITH YOUR BUCKET NAME REGION = 'us-central1' # REPLACE WITH YOUR BUCKET REGION e.g. us-central1 os.environ['TFVERSION'] = '1.8' # Tensorflow version # for bash os.environ['PROJECT'] = PROJECT os.envir...
_____no_output_____
Apache-2.0
courses/machine_learning/deepdive/05_artandscience/b_hyperparam.ipynb
cjqian/training-data-analyst
Create command-line programIn order to submit to Cloud ML Engine, we need to create a distributed training program. Let's convert our housing example to fit that paradigm, using the Estimators API.
%%bash rm -rf trainer mkdir trainer touch trainer/__init__.py %%writefile trainer/house.py import os import math import json import shutil import argparse import numpy as np import pandas as pd import tensorflow as tf def train(output_dir, batch_size, learning_rate): tf.logging.set_verbosity(tf.logging.INFO) # ...
_____no_output_____
Apache-2.0
courses/machine_learning/deepdive/05_artandscience/b_hyperparam.ipynb
cjqian/training-data-analyst
Create hyperparam.yaml
%%writefile hyperparam.yaml trainingInput: hyperparameters: goal: MINIMIZE maxTrials: 5 maxParallelTrials: 1 hyperparameterMetricTag: average_loss params: - parameterName: batch_size type: INTEGER minValue: 8 maxValue: 64 scaleType: UNIT_LINEAR_SCALE - parameterName...
_____no_output_____
Apache-2.0
courses/machine_learning/deepdive/05_artandscience/b_hyperparam.ipynb
cjqian/training-data-analyst
Big Brother - Healthcare edition Building a classifier using the [fastai](https://www.fast.ai/) library
from fastai.tabular import * #hide path = Path('./covid19_ml_education') df = pd.read_csv(path/'covid_ml.csv') df.head(3)
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Independent variableThis is the value we want to predict
y_col = 'urgency_of_admission'
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Dependent variableThe values on which we can make a prediciton
cat_names = ['sex', 'cough', 'fever', 'chills', 'sore_throat', 'headache', 'fatigue'] cat_names = ['sex', 'cough', 'fever', 'headache', 'fatigue'] cont_names = ['age'] #hide procs = [FillMissing, Categorify, Normalize] #hide test = TabularList.from_df(df.iloc[660:861].copy(), path = path, cat_names= cat_names, cont_nam...
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
ModelHere we build our machine learning model that will learn from the dataset to classify between patients Using Focal Loss
import torch import torch.nn as nn import torch.nn.functional as F from torch.autograd import Variable class FocalLoss(nn.Module): def __init__(self, gamma=0, alpha=None, size_average=True): super(FocalLoss, self).__init__() self.gamma = gamma self.alpha = alpha if isinstance(alpha,...
/usr/local/lib/python3.7/site-packages/pandas/core/indexing.py:205: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy self._setitem_with_i...
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Making predictionsWe've taken out a test set to see how well our model works, by making predictions on them.Interestingly, all those predicted with 'High' urgency have a common trait of absence of **chills** and **sore throat**
testdf.urgency.value_counts() testdf.predictions.value_counts() from sklearn.metrics import classification_report print(classification_report(testdf.predictions, testdf.urgency, labels = ["High", "Low"])) print(classification_report(testdf.predictions, testdf.urgency, labels = ["High", "Low"])) testdf = pd.read_csv('pr...
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Profile after focal loss
import seaborn as sns import pandas as pd fig, ax = plt.subplots() fig.set_size_inches(7,5) df_cm = pd.DataFrame(cm_test, index = ['Actual Low','Actual High'], columns = ['Predicted Low','Predicted High']) sns.set(font_scale=1.2) sns.heatmap(df_cm, annot=True, ax = ax) ax.set_ylim([0,2]); ax.set_titl...
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Experimental SectionTrying to figure out top
for i in range(len(testdf)): row = testdf.iloc[i][1:] testdf.probability.iloc[i] = round(float(learn.predict(row[1:-1])[2][0]),5) testdf.head() testdf.sort_values(by=['probability'],ascending = False, inplace = True) # cumulative lift gain baseline model - test 20% Cost based affection Give kits only top 20%...
_____no_output_____
Unlicense
UnivAiBlog/.ipynb_checkpoints/CoVID-49-checkpoint.ipynb
hargun3045/covid19-app
Edgar Holdings The examples in this notebook demonstrate using the GremlinPython library to connect to and work with a Neptune instance. Using a Jupyter notebook in this way provides a nice way to interact with your Neptune graph database in a familiar and instantly productive environment. Connect to the Neptune Data...
!pip install --upgrade pip !pip install futures !pip install gremlinpython !pip install SPARQLWrapper !pip install tornado !pip install tornado-httpclient-session !pip install tornado-utils !pip install matplotlib !pip install numpy !pip install pandas !pip install networkx
Requirement already up-to-date: pip in /home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages (19.0.3) Requirement already satisfied: futures in /home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages (3.1.1) Requirement already satisfied: gremlinpython in /home/ec2-user/anaconda3/envs/python3/lib/...
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Establish access to our Neptune instanceBefore we can work with our graph we need to establish a connection to it. This is done using the `DriverRemoteConnection` capability as defined by Apache TinkerPop and supported by GremlinPython. The `neptune.py` helper module facilitates creating this connection.Once this cell...
from gremlin_python import statics from gremlin_python.structure.graph import Graph from gremlin_python.process.graph_traversal import __ from gremlin_python.process.strategies import * from gremlin_python.driver.driver_remote_connection import DriverRemoteConnection endpoint="wss://dbfindata.carpeooi4ov5.us-east-1.ne...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Let's find out a bit about the graphLet's start off with a simple query just to make sure our connection to Neptune is working. The queries below look at all of the vertices and edges in the graph and create two maps that show the demographic of the graph. As we are using the air routes data set, not surprisingly, the...
vertices = g.V().groupCount().by(T.label).toList() edges = g.E().groupCount().by(T.label).toList() print(vertices) print(edges)
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Find routes longer than 8,400 milesThe query below finds routes in the graph that are longer than 8,400 miles. This is done by examining the `dist` property of the `routes` edges in the graph. Having found some edges that meet our criteria we sort them in descending order by distance. The `where` step filters out the ...
paths = g.V().hasLabel('airport').as_('a') \ .outE('route').has('dist',gt(8400)) \ .order().by('dist',Order.decr) \ .inV() \ .where(P.lt('a')).by('code') \ .path().by('code').by('dist').by('code').toList() for p in paths: print(p)
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Draw a Bar Chart that represents the routes we just found.One of the nice things about using Python to work with our graph is that we can take advantage of the larger Python ecosystem of libraries such as `matplotlib`, `numpy` and `pandas` to further analyze our data and represent it pictorially. So, now that we have ...
import matplotlib.pyplot as plt; plt.rcdefaults() import numpy as np import matplotlib.pyplot as plt import pandas as pd routes = list() dist = list() # Construct the x-axis labels by combining the airport pairs we found # into strings with with a "-" between them. We also build a list containing # the distance valu...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Explore the distribution of airports by continentThe next example queries the graph to find out how many airports are in each continent. The query starts by finding all vertices that are continents. Next, those vertices are grouped, which creates a map (or dict) whose keys are the continent descriptions and whose valu...
# Return a map where the keys are the continent names and the values are the # number of airports in that continent. m = g.V().hasLabel('continent') \ .group().by('desc').by(__.out('contains').count()) \ .order(Scope.local).by(Column.keys) \ .next() for c,n in m.items(): print('%4d %s' %...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Draw a pie chart representing the distribution by continentRather than return the results as text like we did above, it might be nicer to display them as percentages on a pie chart. That is what the code in the next cell does. Rather than return the descriptions of the continents (their names) this time our Gremlin qu...
import matplotlib.pyplot as plt; plt.rcdefaults() import numpy as np # Return a map where the keys are the continent codes and the values are the # number of airports in that continent. m = g.V().hasLabel('continent').group().by('code').by(__.out().count()).next() fig,pie1 = plt.subplots() pie1.pie(m.values() \ ...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Find some routes from London to San Jose and draw themOne of the nice things about connected graph data is that it lends itself nicely to visualization that people can get value from looking at. The Python `networkx` library makes it fairly easy to draw a graph. The next example takes advantage of this capability to d...
import matplotlib.pyplot as plt; plt.rcdefaults() import numpy as np import matplotlib.pyplot as plt import pandas as pd import networkx as nx # Find up to 15 routes from LHR to SJC that make one stop. paths = g.V().has('airport','code','LHR') \ .out().out().has('code','SJC').limit(15) \ .pat...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
PART 2 - Examples that use iPython GremlinThis part of the notebook contains examples that use the iPython Gremlin Jupyter extension to work with a Neptune instance using Gremlin. Configuring iPython Gremlin to work with NeptuneBefore we can start to use iPython Gremlin we need to load the Jupyter Kernel extension an...
# Create a string containing the full Web Socket path to the endpoint # Replace <neptune-instance-name> with the name of your Neptune instance. # which will be of the form myinstance.us-east-1.neptune.amazonaws.com #neptune_endpoint = '<neptune-instance-name>' neptune_endpoint = os.environ['NEPTUNE_CLUSTER_ENDPOINT'] ...
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Run this cell if you need to reload the Gremlin extension.Occaisionally it becomes necessary to reload the iPython Gremlin extension to make things work. Running this cell will do that for you.
# Re-load the iPython Gremlin Jupyter Kernel extension. %reload_ext gremlin
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
A simple query to make sure we can connect to the graph. Find all the airports in England that are in London. Notice that when using iPython Gremlin you do not need to use a terminal step such as `next` or `toList` at the end of the query in order to get it to return results. As mentioned earlier in this post, the `%r...
%reset -f %gremlin g.V().has('airport','region','GB-ENG') \ .has('city','London').values('desc')
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
You can store the results of a query in a variable just as when using Gremlin Python.The query below is the same as the previous one except that the results of running the query are stored in the variable 'places'. We can then work with that variable in our code.
%reset -f places = %gremlin g.V().has('airport','region','GB-ENG') \ .has('city','London').values('desc') for p in places: print(p)
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Treating entire cells as GremlinAny cell that begins with `%%gremlin` tells iPython Gremlin to treat the entire cell as Gremlin. You cannot mix Python code into these cells.
%%gremlin g.V().has('city','London').has('region','GB-ENG').count()
_____no_output_____
MIT-0
neptune-sagemaker/notebooks/edgar/.ipynb_checkpoints/01-edar-checkpoint.ipynb
JanuaryThomas/amazon-neptune-samples
Creating a pandas data frame for experimental PF valuesWan et al. (JCTC 2020) trained a forward model on experimentally measured HDX protection factors:* 72 PF values for backbone amides of ubiquitin taken from Craig et al.* 30 (of 53) for amibackbone amides of BPTI taken from Persson et al.In this notebook we have co...
import os, sys import numpy as np import pandas as pd ### Ubiquitin ubiquitin_text ="""#residue\tresnum\tln PF \\ GLN & 2 & 6.210072 \\ ILE & 3 & 13.7372227 \\ PHE & 4 & 13.4839383 \\ VAL & 5 & 13.1523661 \\ LYS & 6 & 10.6909026 \\ THR & 7 & 10.4629467 \\ LEU & 8 & 0.67005226 \\ THR & 9 & 0 \\ GLY & 10 & 3.93511792 \...
all_lnPF_values.shape (102,) all_lnPF_values_HONGBIN.shape (102,) 6.210072 6.210071995804942 13.737222699999998 13.737222664802477 13.4839383 13.483938304573131 13.1523661 13.152366051181989 10.6909026 10.690902586771355 10.4629467 10.462946662564942 0.67005226 0.6700522620612673 0.0 0.0 3.93511792 3.9351179239268244 5...
MIT
experimental-data/create_pd.ipynb
vvoelz/HDX-forward-model
Implementation of a Devito skew self adjoint variable density visco- acoustic isotropic modeling operator -- Correctness Testing -- This operator is contributed by Chevron Energy Technology Company (2020)This operator is based on simplfications of the systems presented in:**Self-adjoint, energy-conserving second-order...
from scipy.special import hankel2 import numpy as np from examples.seismic import RickerSource, Receiver, TimeAxis, Model, AcquisitionGeometry from devito import (Grid, Function, TimeFunction, SpaceDimension, Constant, Eq, Operator, solve, configuration) from devito.finite_differences import Deriva...
_____no_output_____
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
1. Analytic response in the far fieldTest that data generated in a wholespace matches analogous analytic data away from the near field. We copy/modify the material shown in the [examples/seismic/acoustic/accuracy.ipynb](https://github.com/devitocodes/devito/blob/master/examples/seismic/acoustic/accuracy.ipynb) noteboo...
# Define the analytic response def analytic_response(fpeak, time_axis, src_coords, rec_coords, v): nt = time_axis.num dt = time_axis.step v0 = v.data[0,0] sx, sz = src_coords[0, :] rx, rz = rec_coords[0, :] ntpad = 20 * (nt - 1) + 1 tmaxpad = dt * (ntpad - 1) time_axis_pad = TimeAxis(sta...
_____no_output_____
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
Reset default shapes for subsequent tests
npad = 10 fpeak = 0.010 qmin = 0.1 qmax = 500.0 tmax = 1000.0 shape = (101, 81)
_____no_output_____
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
2. Modeling operator linearity test, with respect to sourceFor random vectors $s$ and $r$, prove:$$\begin{aligned}F[m]\ (\alpha\ s) &\approx \alpha\ F[m]\ s \\[5pt]F[m]^\top (\alpha\ r) &\approx \alpha\ F[m]^\top r \\[5pt]\end{aligned}$$We first test the forward operator, and in the cell below that the adjoint operato...
#NBVAL_INGNORE_OUTPUT solver = acoustic_ssa_setup(shape=shape, dtype=dtype, space_order=8, tn=tmax) src = solver.geometry.src a = -1 + 2 * np.random.rand() rec1, _, _ = solver.forward(src) src.data[:] *= a rec2, _, _ = solver.forward(src) rec1.data[:] *= a # Check receiver wavefeild linearity # Normalize by rms of re...
Operator `IsoFwdOperator` run in 0.03 s Operator `IsoAdjOperator` run in 0.02 s Operator `IsoAdjOperator` run in 0.03 s
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
3. Modeling operator adjoint test, with respect to sourceFor random vectors $s$ and $r$, prove:$$r \cdot F[m]\ s \approx s \cdot F[m]^\top r$$
#NBVAL_INGNORE_OUTPUT src1 = solver.geometry.src rec1 = solver.geometry.rec rec2, _, _ = solver.forward(src1) # flip sign of receiver data for adjoint to make it interesting rec1.data[:] = rec2.data[:] src2, _, _ = solver.adjoint(rec1) sum_s = np.dot(src1.data.reshape(-1), src2.data.reshape(-1)) sum_r = np.dot(rec1.d...
Operator `IsoFwdOperator` run in 0.56 s Operator `IsoAdjOperator` run in 1.42 s
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
4. Nonlinear operator linearization test, with respect to modelFor initial velocity model $m$ and random perturbation $\delta m$ prove that the $L_2$ norm error in the linearization $E(h)$ is second order (decreases quadratically) with the magnitude of the perturbation.$$E(h) = \biggl\|\ f(m+h\ \delta m) - f(m) - h\ \...
#NBVAL_INGNORE_OUTPUT src = solver.geometry.src # Create Functions for models and perturbation m0 = Function(name='m0', grid=solver.model.grid, space_order=8) mm = Function(name='mm', grid=solver.model.grid, space_order=8) dm = Function(name='dm', grid=solver.model.grid, space_order=8) # Background model m0.data[:] ...
_____no_output_____
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
5. Jacobian operator linearity test, with respect to modelFor initial velocity model $m$ and random vectors $\delta m$ and $\delta r$, prove:$$\begin{aligned}\nabla F[m; q]\ (\alpha\ \delta m) &\approx \alpha\ \nabla F[m; q]\ \delta m \\[5pt](\nabla F[m; q])^\top (\alpha\ \delta r) &\approx \alpha\ (\nabla F[m; q])^\t...
#NBVAL_INGNORE_OUTPUT src0 = solver.geometry.src m0 = Function(name='m0', grid=solver.model.grid, space_order=8) m1 = Function(name='m1', grid=solver.model.grid, space_order=8) m0.data[:] = 1.5 # Model perturbation, box of random values centered on middle of model m1.data[:] = 0 size = 5 ns = 2 * size + 1 nx2, nz2 =...
Operator `IsoFwdOperator` run in 0.55 s Operator `IsoJacobianAdjOperator` run in 0.13 s Operator `IsoJacobianAdjOperator` run in 2.34 s
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
6. Jacobian operator adjoint test, with respect to model perturbation and receiver wavefield perturbation For initial velocity model $m$ and random vectors $\delta m$ and $\delta r$, prove:$$\delta r \cdot \nabla F[m; q]\ \delta m \approx \delta m \cdot (\nabla F[m; q])^\top \delta r$$
#NBVAL_INGNORE_OUTPUT src0 = solver.geometry.src m0 = Function(name='m0', grid=solver.model.grid, space_order=8) dm1 = Function(name='dm1', grid=solver.model.grid, space_order=8) m0.data[:] = 1.5 # Model perturbation, box of random values centered on middle of model dm1.data[:] = 0 size = 5 ns = 2 * size + 1 nx2, nz...
Operator `IsoFwdOperator` run in 0.09 s Operator `IsoJacobianFwdOperator` run in 4.05 s
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
7. Skew symmetry for shifted derivativesEnsure for random $x_1, x_2$ that Devito shifted derivative operators $\overrightarrow{\partial_x}$ and $\overrightarrow{\partial_x}$ are skew symmetric by verifying the following dot product test.$$x_2 \cdot \left( \overrightarrow{\partial_x}\ x_1 \right) \approx -\ x_1 \cdot \...
#NBVAL_INGNORE_OUTPUT # Make 1D grid to test derivatives n = 101 d = 1.0 shape = (n, ) spacing = (1 / (n-1), ) origin = (0., ) extent = (d * (n-1), ) dtype = np.float64 # Initialize Devito grid and Functions for input(f1,g1) and output(f2,g2) # Note that space_order=8 allows us to use an 8th order finite differen...
Operator `Kernel` run in 0.01 s
MIT
examples/seismic/skew_self_adjoint/ssa_03_iso_correctness.ipynb
dabiged/devito
Support Vector Regression (SVR) Importing the libraries
import numpy as np import matplotlib.pyplot as plt import pandas as pd
_____no_output_____
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Importing the dataset
dataset = pd.read_csv('Position_Salaries.csv') x = dataset.iloc[:, 1:-1].values y = dataset.iloc[:, -1].values print(x) print(y) y = y.reshape(len(y), 1) print(y)
[[ 45000] [ 50000] [ 60000] [ 80000] [ 110000] [ 150000] [ 200000] [ 300000] [ 500000] [1000000]]
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Feature Scaling
from sklearn.preprocessing import StandardScaler sc_x = StandardScaler() sc_y = StandardScaler() x = sc_x.fit_transform(x) y = sc_y.fit_transform(y) print(x) print(y)
[[-0.72004253] [-0.70243757] [-0.66722767] [-0.59680786] [-0.49117815] [-0.35033854] [-0.17428902] [ 0.17781001] [ 0.88200808] [ 2.64250325]]
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Training the SVR model on the whole dataset
from sklearn.svm import SVR regressor = SVR(kernel = 'rbf') regressor.fit(x, y)
/usr/local/lib/python3.7/dist-packages/sklearn/utils/validation.py:993: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel(). y = column_or_1d(y, warn=True)
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Predicting a new result
sc_y.inverse_transform([regressor.predict(sc_x.transform([[6.5]]))]) #requires 2D array, therefore added square brackets before regressor.predict
_____no_output_____
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Visualising the SVR results
m=regressor.predict(x) plt.scatter(sc_x.inverse_transform(x), sc_y.inverse_transform(y), color = 'red') plt.plot(sc_x.inverse_transform(x), sc_y.inverse_transform(m.reshape(len(m),1)), color='blue') # requires 2d array therefore used reshape() plt.title('truth or bluff (Support Vector Regression)') plt.xlabel('Level') ...
_____no_output_____
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
Visualising the SVR results (for higher resolution and smoother curve)
X_unscaled = sc_x.inverse_transform(x) y_unscaled = sc_y.inverse_transform(y) X_grid = np.arange(min(x), max(x), 0.1) X_grid = X_grid.reshape((len(X_grid), 1)) X_grid_unscaled = sc_x.inverse_transform(X_grid) y_pred_grid = regressor.predict(X_grid) y_pred_grid = sc_y.inverse_transform([y_pred_grid]) y_pred_grid = y_p...
_____no_output_____
MIT
Regression/Support_vector_regression.ipynb
AstitvaSharma/ML_Algorithms
The fundamental data structures are Series and DataFrame
s = pd.Series(np.random.randn(4), index =['a','f','t','y']) s s['a']
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
The data argument can be dict, ndarray, or even a scalar etc
data = {'a': 3, 'b':4, 'c': 5, 'f':'something_else'} data s_dict = pd.Series(data, index=data.keys()) s_dict
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
When the dict doesn't have a matching key, it's not added. And when the index key is missing a value attached, NaN is given
s_dict_with_nan = pd.Series(data, index = ['a','b', 'j']) s_dict_with_nan
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Works differently with scalars
s_scalar = pd.Series(3, index=range(5))
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Works the same way even if you pass a list with one element, like [3] but fails if you pass [3,4] because expects 5 and not 2
s_scalar
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Series works just like an ndarray, if you have worked with numpy before. I do not have a lot of experience with numpy so can't comment on the full capabilites but according to what I know, you can apply vectorized operations to get a better code performance, slice in the same way we do with numpy ndarrays.
s
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
MENTIONING PYTHON DATA TYPE IN VARIABLE NAME IS NOT GOOD PRACTICE, SO TRY NOT TO
s[s>s.median()] s[[3, 0, 1]] np.exp(s) s.values s.keys s.keys() s.index try : some_random_var = s['g'] # Raises key error except KeyError : print('Caught key Error') s.f # Can also access elements this way s.a s
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Vectorized operations - Start of the end of matlab RIP
s+s s*2 s*3 s/2
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Moving to Pandas DataFrame According to definition from https://pandas.pydata.org/pandas-docs/stable/dsintro.htmldsintro __DataFrame is a 2-dimensional labeled data structure with columns of potentially different types. You can think of it like a spreadsheet or SQL table, or a dict of Series objects. It is generally ...
data = {'one': pd.Series([1,2,3,4], index=['a','b','c','d']), 'two': pd.Series([3,4,5,56,6], index=['a','b','f','e','y'])} df = pd.DataFrame(data) df
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
As you can see that it merged the two index together and filled the rest of the values with NaN
df_1 = pd.DataFrame(data, index=['a','b','c','f']) df_1 df_2 = pd.DataFrame(data, columns=['two']) df_2 df df.index df.columns data_for_dict = { 'one': [1,2,4,5], 'two': [2.,5.4,4.,5]} df_from_dict = pd.DataFrame(data_for_dict) df_from_dict['one'] data = np.ones((2,8)) data data.keys() s = pd.Series([...
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
__A difference to note here is that Series does have a values and keys attribute whereas an ndarray doesn't. Good to know__
rows = [[1,2,3,43],[3,5,6,6],[66,6,6]] df_rows = pd.DataFrame(rows) df_rows df_rows_with_index = pd.DataFrame(rows, index=['first', 'second','third']) df_rows_with_index # Number of indices should always match the rows count else will raise a shape error try: df_test = pd.DataFrame({'one':[2,23,4,5,56], ...
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
As we can see from the above couple of test df, when we pass a dictionary as data to DataFrame, the arrays needs to be of the same length. But if we pass a row of rows, they are adjusted accordinglyStrange, but need to know the reasoning.
df_test_4
_____no_output_____
MIT
Untitled.ipynb
arora-anmol/pandas-practice
Working with POSTGIS In this chapter we will cover the following topics: Executing a PostGIS ST_Buffer analysis query and exporting it to GeoJSON Finding out whether a point is inside a polygon Splitting LineStrings at intersections using ST_Node Checking the validity of LineStrings Executing a spatial ...
#!/usr/bin/env python import psycopg2 import json from geojson import loads, Feature, FeatureCollection # Database connection information db_host = "localhost" db_user = "calvin" db_passwd = "planets" db_database = "py_test" db_port = "5432" # connect to database conn = psycopg2.connect(host=db_host, user=db_user, ...
_____no_output_____
MIT
spatial_GIS_RS/python_notebooks/4. Working_with_Postgis.ipynb
OpitiCalvin/Scripts_n_Tutorials
How it worksThe database connection is using the pyscopg2 module, so we import the libraries at the start alongside geojson and the standard json modules to handle our GeoJSON export.Our connection is created and then followed immediately with our SQL Buffer query string. The query uses three PostGIS functions. Workin...
#!/usr/bin/env python # -*- coding: utf-8 -*- import json import psycopg2 from geojson import loads, Feature, FeatureCollection # Database Connection Info db_host = "localhost" db_user = "pluto" db_passwd = "stars" db_database = "py_geoan_cb" db_port = "5432" # connect to DB conn = psycopg2.connect(host=db_host, us...
_____no_output_____
MIT
spatial_GIS_RS/python_notebooks/4. Working_with_Postgis.ipynb
OpitiCalvin/Scripts_n_Tutorials
You can now view your newly created GeoJSON file on a great little site created by Mapbox at http://www.geojson.io. Simply drag and drop your GeoJSON file from Windows Explorer in Windows or Nautilus in Ubuntu onto the http://www.geojson.io web page and, Bob's your uncle, you should see 50 or so schools that are locate...
#!/usr/bin/env python import psycopg2 import json from geojson import loads, Feature, FeatureCollection # Database Connection Info db_host = "localhost" db_user = "pluto" db_passwd = "stars" db_database = "py_geoan_cb" db_port = "5432" # connect to DB conn = psycopg2.connect(host=db_host, user=db_user, port=db_...
_____no_output_____
MIT
spatial_GIS_RS/python_notebooks/4. Working_with_Postgis.ipynb
OpitiCalvin/Scripts_n_Tutorials
Well, this was quite simple and we can now see that McNicoll Avenue is split at the intersection with Cypress Street. HOw it worksLooking at the code, we can see that the database connection remains the same and the only new thing is the query itself that creates the intersection. Here three separate PostGIS functions...
#!/usr/bin/env python # -*- coding: utf-8 -*- import psycopg2 # Database Connection Info db_host = "localhost" db_user = "pluto" db_passwd = "stars" db_database = "py_geoan_cb" db_port = "5432" # connect to DB conn = psycopg2.connect(host=db_host, user=db_user, port=db_port, password=db_passwd, database=db_data...
_____no_output_____
MIT
spatial_GIS_RS/python_notebooks/4. Working_with_Postgis.ipynb
OpitiCalvin/Scripts_n_Tutorials
This query should return an empty Python list, which means that we have no invalid geometries. If there are objects in your list, then you'll know that you have some manual work to do to correct those geometries. Your best bet is to fire up QGIS and get started with digitizing tools to clean things up. Executing a spa...
#!/usr/bin/env python # -*- coding: utf-8 -*- import psycopg2 # Database Connection Info db_host = "localhost" db_user = "pluto" db_passwd = "stars" db_database = "py_geoan_cb" db_port = "5432" # connect to DB conn = psycopg2.connect(host=db_host, user=db_user, port=db_port, password=db_passwd, database=db_database)...
_____no_output_____
MIT
spatial_GIS_RS/python_notebooks/4. Working_with_Postgis.ipynb
OpitiCalvin/Scripts_n_Tutorials
**Getting Started With Spark using Python** Estimated time needed: **15** minutes ![](http://spark.apache.org/images/spark-logo.png) The Python API Spark is written in Scala, which compiles to Java bytecode, but you can write python code to communicate to the java virtual machine through a library called py4j. P...
# Installing required packages !pip install pyspark !pip install findspark import findspark findspark.init() # PySpark is the Spark API for Python. In this lab, we use PySpark to initialize the spark context. from pyspark import SparkContext, SparkConf from pyspark.sql import SparkSession
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Exercise 1 - Spark Context and Spark Session In this exercise, you will create the Spark Context and initialize the Spark session needed for SparkSQL and DataFrames.SparkContext is the entry point for Spark applications and contains functions to create RDDs such as `parallelize()`. SparkSession is needed for SparkSQL...
# Creating a spark context class sc = SparkContext() # Creating a spark session spark = SparkSession \ .builder \ .appName("Python Spark DataFrames basic example") \ .config("spark.some.config.option", "some-value") \ .getOrCreate()
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 2: Initialize Spark sessionTo work with dataframes we just need to verify that the spark session instance has been created.
spark
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Exercise 2: RDDsIn this exercise we work with Resilient Distributed Datasets (RDDs). RDDs are Spark's primitive data abstraction and we use concepts from functional programming to create and manipulate RDDs. Task 1: Create an RDD.For demonstration purposes, we create an RDD here by calling `sc.parallelize()`\We creat...
data = range(1,30) # print first element of iterator print(data[0]) len(data) xrangeRDD = sc.parallelize(data, 4) # this will let us know that we created an RDD xrangeRDD
1
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 2: Transformations A transformation is an operation on an RDD that results in a new RDD. The transformed RDD is generated rapidly because the new RDD is lazily evaluated, which means that the calculation is not carried out when the new RDD is generated. The RDD will contain a series of transformations, or computa...
subRDD = xrangeRDD.map(lambda x: x-1) filteredRDD = subRDD.filter(lambda x : x<10)
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 3: Actions A transformation returns a result to the driver. We now apply the `collect()` action to get the output from the transformation.
print(filteredRDD.collect()) filteredRDD.count()
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 4: Caching Data This simple example shows how to create an RDD and cache it. Notice the **10x speed improvement**! If you wish to see the actual computation time, browse to the Spark UI...it's at host:4040. You'll see that the second calculation took much less time!
import time test = sc.parallelize(range(1,50000),4) test.cache() t1 = time.time() # first count will trigger evaluation of count *and* cache count1 = test.count() dt1 = time.time() - t1 print("dt1: ", dt1) t2 = time.time() # second count operates on cached data only count2 = test.count() dt2 = time.time() - t2 pri...
dt1: 1.997375726699829 dt2: 0.41718220710754395
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Exercise 3: DataFrames and SparkSQL In order to work with the extremely powerful SQL engine in Apache Spark, you will need a Spark Session. We have created that in the first Exercise, let us verify that spark session is still active.
spark
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 1: Create Your First DataFrame! You can create a structured data set (much like a database table) in Spark. Once you have done that, you can then use powerful SQL tools to query and join your dataframes.
# Download the data first into a local `people.json` file !curl https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBM-BD0225EN-SkillsNetwork/labs/data/people.json >> people.json # Read the dataset into a spark dataframe using the `read.json()` function df = spark.read.json("people.json").cache() # Prin...
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Task 2: Explore the data using DataFrame functions and SparkSQLIn this section, we explore the datasets using functions both from dataframes as well as corresponding SQL queries using sparksql. Note the different ways to achieve the same task!
# Select and show basic data columns df.select("name").show() df.select(df["name"]).show() spark.sql("SELECT name FROM people").show() # Perform basic filtering df.filter(df["age"] > 21).show() spark.sql("SELECT age, name FROM people WHERE age > 21").show() # Perfom basic aggregation of data df.groupBy("age").count(...
+----+-----+ | age|count| +----+-----+ | 19| 1| |null| 1| | 30| 1| +----+-----+ +----+-----+ | age|count| +----+-----+ | 19| 1| |null| 0| | 30| 1| +----+-----+
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
*** Question 1 - RDDs Create an RDD with integers from 1-50. Apply a transformation to multiply every number by 2, resulting in an RDD that contains the first 50 even numbers.
# starter code # numbers = range(1, 50) # numbers_RDD = ... # even_numbers_RDD = numbers_RDD.map(lambda x: ..) # Code block for learners to answer numbers = range(1, 50) numbers_RDD = sc.parallelize(data, 4) even_numbers_RDD = numbers_RDD.map(lambda x: x * 2) even_numbers_RDD.collect()
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Question 2 - DataFrames and SparkSQL Similar to the `people.json` file, now read the `people2.json` file into the notebook, load it into a dataframe and apply SQL operations to determine the average age in our people2 file.
# starter code # !curl https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBM-BD0225EN-SkillsNetwork/labs/data/people2.json >> people2.json # df = spark.read... # df.createTempView.. # spark.sql("SELECT ...") # Code block for learners to answer !curl https://cf-courses-data.s3.us.cloud-object-storage.ap...
% Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 326 100 326 0 0 478 0 --:--:-- --:--:-- --:--:-- 478
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
Double-click **here** for a hint.<!-- The hint is below:1. The SQL query "Select AVG(column_name) from.." can be used to find the average value of a column. 2. Another possible way is to use the dataframe operations select() and mean()--> Double-click **here** for the solution.<!-- The answer is below:df = spark.read('...
# Code block for learners to answer spark.stop()
_____no_output_____
MIT
Coursera/Apache_Spark_Fundamentals.ipynb
SokolovVadim/Big-Data
There are two ways to load the pretrained model, the first way is to load from local model directory. A model directory should consist of two files: config.pkl that describes the configuration of the model and model.pt, which is the model weights. If you use model.save(MODEL_DIR) to save the model, then, it should be g...
path = utils.download_pretrained_model('Morgan_AAC_DAVIS') net = models.model_pretrained(path_dir = path) net.config
_____no_output_____
BSD-3-Clause
DEMO/load_pretraining_models_tutorial.ipynb
markcheung/DeepPurpose
For models that provided by us, you can directly use the pre-designated model names. The full list is in the Github README https://github.com/kexinhuang12345/DeepPurpose/blob/master/README.mdpretrained-models
net = models.model_pretrained(model = 'MPNN_CNN_DAVIS') net.config
Beginning Downloading MPNN_CNN_DAVIS Model... Downloading finished... Beginning to extract zip file... pretrained model Successfully Downloaded...
BSD-3-Clause
DEMO/load_pretraining_models_tutorial.ipynb
markcheung/DeepPurpose
SIR example Deterministic model
def SIR(t, y, b, d, beta, u, v): N = y[0]+y[1]+y[2] return [b*N - d*y[0] - beta*y[1]/N*y[0] - v*y[0], beta*y[1]/N*y[0] - u*y[1] - d*y[1], u*y[1] - d*y[2] + v*y[0]] # Time interval for the simulation t0 = 0 t1 = 120 t_span = (t0, t1) t_eval = np.linspace(t_span[0], t_span[1], 10000) # Init...
_____no_output_____
MIT
LectureNotes/MC/MC.ipynb
enigne/ScientificComputingBridging
Stochastic model
import numpy as np # Plotting modules import matplotlib.pyplot as plt def sample_discrete(probs): """ Randomly sample an index with probability given by probs. """ # Generate random number q = np.random.rand() # Find index i = 0 p_sum = 0.0 while p_sum < q: p_sum += pro...
_____no_output_____
MIT
LectureNotes/MC/MC.ipynb
enigne/ScientificComputingBridging
Solve SIR in stochastic method
# Column changes S, I, R simple_update = np.array([[-1, 1, 0], [0, -1, 1], [-1, 0, 1], [-1, 0, 0], [0, -1, 0], [0, 0, -1], [1, 0, 0]], dtype=np.int) # Specify para...
_____no_output_____
MIT
LectureNotes/MC/MC.ipynb
enigne/ScientificComputingBridging
Simulate interest rate path by the CIR model
import math import numpy as np import matplotlib.pyplot as plt def cir(r0, K, theta, sigma, T=1., N=10,seed=777): np.random.seed(seed) dt = T/float(N) rates = [r0] for i in range(N): dr = K*(theta-rates[-1])*dt + \ sigma*math.sqrt(abs(rates[-1]))*np.random.normal() rates...
_____no_output_____
MIT
LectureNotes/MC/MC.ipynb
enigne/ScientificComputingBridging
Import data and drop redundant data (rates)
# import data df = pd.read_csv('../../data/deepsolar_tract.csv', encoding = "utf-8")
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Clean Data
df = drop_redundant_columns(df) # Create our target column 'has_tiles', and drop additional redundant columns df = create_has_tiles_target_column(df) df.shape # # # Figure out which variables are highly correlated, remove the most correlated ones one by one # corr = pd.DataFrame((df.corr() > 0.8).sum()) # corr.sort_v...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Checking for missing values
nulls = pd.DataFrame(df.isna().sum()) nulls.columns = ["missing"] nulls[nulls['missing']>0].head() # drop all missing values df = df.dropna(axis = 0) # Check class imbalance df.has_tiles.value_counts() df.shape
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Train test split
X = df.drop('has_tiles', axis = 1) y = df['has_tiles'] df.shape X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, stratify = y)
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Sampling Techniques
# smote, undersampling, or oversampling X_train, y_train = pick_sampling_method(X_train, y_train, method = 'oversampling') y_train.value_counts()
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Scale Data
scaler = StandardScaler() X_train = scaler.fit_transform(X_train) X_test = scaler.transform(X_test) X_train.shape
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Modeling
from sklearn.metrics import classification_report
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Vanilla Decision Tree 0.74
## DUMMY dummy = DecisionTreeClassifier() dummy.fit(X_train, y_train) y_pred = dummy.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred))) print("Recall: {}".format(recall_score(y_test, y_pred))) print("Accuracy: {}".format(accuracy_score(y_test, y_pred))) print("F1 Score: {}".format(f1_score(y...
precision recall f1-score support 0 0.24 0.82 0.37 2500 1 0.79 0.21 0.33 8320 accuracy 0.35 10820 macro avg 0.51 0.51 0.35 10820 weighted avg 0.66 0.35 0.34 ...
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Decision Tree with Hyperparameter Tuning
dt = find_hyperparameters(pipe_dt, params_dt, X_train, y_train) dt.best_params_ best_dt = dt.best_estimator_ # Decision Tree: {'dt__max_depth': 2, 'dt__min_samples_leaf': 1, 'dt__min_samples_split': 2} best_dt.fit(X_train, y_train) best_dt.score(X_test, y_test) # Decision Tree: 0.755637707948244 y_pred_dt = best_dt.pre...
Precision: 0.9131785238869989 Recall: 0.7420673076923077 Accuracy: 0.7474121996303142 F1 Score: 0.8187785955838471
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Final Model
# with oversampling rf = RandomForestClassifier(max_features = 'sqrt', max_depth = 5, min_samples_leaf = 5, n_estimators = 30) rf.fit(X_train, y_train) y_pred = rf.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred))) print("Recall: {}".format(recall_score(y_test, y_pred))) print("Accuracy: {}"...
C:\Users\allis\Anaconda3\lib\site-packages\sklearn\base.py:197: FutureWarning: From version 0.24, get_params will raise an AttributeError if a parameter cannot be retrieved as an instance attribute. Previously it would return None. FutureWarning) C:\Users\allis\Anaconda3\lib\site-packages\yellowbrick\classifier\base....
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Vanilla Random Forests
rf = RandomForestClassifier() rf.fit(X_train, y_train) y_pred = rf.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred))) print("Recall: {}".format(recall_score(y_test, y_pred))) print("Accuracy: {}".format(accuracy_score(y_test, y_pred))) print("F1 Score: {}".format(f1_score(y_test, y_pred))) p...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Random Forests with Hyperparameter Tuning
### Random Forests rf = find_hyperparameters(pipe_rf, params_rf, X_train, y_train) print(rf.best_params_) best_rf = rf.best_estimator_ #first hyperparamter tuning: {'rf__max_features': 'sqrt', 'rf__min_samples_leaf': 20, 'rf__n_estimators': 30} # Second tuning: {'rf__min_samples_leaf': 5, 'rf__n_estimators': 50} best_...
Precision: 0.8924837003321442 Recall: 0.8719951923076923 Accuracy: 0.8207948243992607 F1 Score: 0.8821204936470303 Balanced Accuracy: 0.7611975961538462
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Vanilla SVC
svc = SVC() svc.fit(X_train, y_train) y_pred_svc = svc.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred_svc))) print("Recall: {}".format(recall_score(y_test, y_pred_svc))) print("Accuracy: {}".format(accuracy_score(y_test, y_pred_svc))) print("F1 Score: {}".format(f1_score(y_test, y_pred_svc)...
Precision: 0.9094472225976483 Recall: 0.8087740384615385 Accuracy: 0.7910351201478744 F1 Score: 0.8561613334181563
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
SVC with Hyperparameter Tuning
svc = find_hyperparameters(pipe_svc, params_svc, X_train, y_train) print(svc.best_params_) best_svc = svc.best_estimator_ best_svc.fit(X_train, y_train) best_svc.score(X_test, y_test) y_pred_svc = best_svc.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred_svc))) print("Recall: {}".format(recal...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Vanilla KNN
knn = KNeighborsClassifier() knn.fit(X_train, y_train) y_pred_knn = knn.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred_knn))) print("Recall: {}".format(recall_score(y_test, y_pred_knn))) print("Accuracy: {}".format(accuracy_score(y_test, y_pred_knn))) print("F1 Score: {}".format(f1_score(y_...
Precision: 0.9354388413889374 Recall: 0.6443509615384615 Accuracy: 0.6923290203327171 F1 Score: 0.7630773610419187
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
KNN with Hyperparameter Tuning
knn = find_hyperparameters(pipe_knn, params_knn, X_train, y_train) print(knn.best_params_) best_knn = knn.best_estimator_ best_knn.fit(X_train, y_train) best_knn.score(X_test, y_test) y_pred_knn = best_knn.predict(X_test) print("Precision: {}".format(precision_score(y_test, y_pred_knn))) print("Recall: {}".format(recal...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Preliminary Conclusions: Model Performance Comparisons Based on comparisons of both accuracy and balanced accuracy scores, our Random Forest Classifier model performed the best with oversampling methods and hyperparameter tuning.
falsepositives = isFalsePositive(df, X_test, y_test, rf) inversefalsepositives = scaler.inverse_transform(falsepositives) inversefalsepositives = pd.DataFrame(inversefalsepositives) inversefalsepositives = inversefalsepositives.set_axis(falsepositives.columns, axis=1, inplace=False) len(inversefalsepositives) ozdf = pd...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis
Running Model on Entire Dataset
ozdf = ozdf['Census_Tract_Number'] merged = df.merge(ozdf, how = 'left', left_on='fips',right_on='Census_Tract_Number') merged = merged.dropna() merged.drop('fips', axis = 1, inplace = True) merged['has_tiles'].value_counts() X_ozdf = merged.drop('has_tiles', axis = 1) y_ozdf = merged['has_tiles'] y_pred_ozdf = rf.pred...
_____no_output_____
MIT
EDA_Notebooks/EDA_Allison.ipynb
BudBernhard/Mod4Project-DeepSolarAnalysis