markdown stringlengths 0 37k | code stringlengths 1 33.3k | path stringlengths 8 215 | repo_name stringlengths 6 77 | license stringclasses 15
values |
|---|---|---|---|---|
Indeed, there are 9 cases of 300-year olds patient ICU admittances that look suspicious, but after that, the maximum age of the patients is 89. Actually, the 300 age is intentionally introduced in the MIMIC-III datasets for privacy protection of patients whose age is 90 or beyond: any patient in this age group got thei... | run_query('''
WITH co AS (
SELECT
icu.subject_id,
icu.hadm_id,
icu.icustay_id,
pat.dob,
TIMESTAMP_DIFF(icu.outtime, icu.intime, DAY) AS icu_length_of_stay,
DATE_DIFF(DATE(icu.intime), DATE(pat.dob), YEAR) AS age,
RANK() OVER (PARTITION BY icu.subject_id ORDER BY icu.intime) AS icustay_id_o... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
In the last column, we see that for some patients (e.g. Subject 40177 and 42281), there are multiple ICU stays. In research studies, we usually filter out follow-up ICU stays, and only keep the first ICU stay so as to minimize unwanted data correlation. For this purpose, we create an exclusion label column based on icu... | run_query('''
WITH co AS (
SELECT
icu.subject_id,
icu.hadm_id,
icu.icustay_id,
pat.dob,
TIMESTAMP_DIFF(icu.outtime, icu.intime, DAY) AS icu_length_of_stay,
DATE_DIFF(DATE(icu.intime), DATE(pat.dob), YEAR) AS age,
RANK() OVER (PARTITION BY icu.subject_id ORDER BY icu.intime) AS icustay_id_o... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Another filter that we often use is the current service that ICU patients are undergoing. This could be done by joining with the services table using the hadm_id column. We can use the BigQuery preview tab to gain some visual understanding of data in this table as usual. We could also find out the number of each servic... | run_query('''
SELECT
curr_service,
IF(curr_service like '%SURG' OR curr_service = 'ORTHO', 1, 0) AS surgical,
COUNT(DISTINCT hadm_id) num_hadm
FROM `datathon-datasets.mimic_demo.services`
GROUP BY 1, 2
ORDER BY 2, 1
''') | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
With this understanding of service types, we are now ready to join the icustays table with the services table to identify what serives ICU patients are undergoing. | run_query('''
SELECT
icu.hadm_id,
icu.icustay_id,
curr_service,
IF(curr_service like '%SURG' OR curr_service = 'ORTHO', 1, 0) AS surgical
FROM `datathon-datasets.mimic_demo.icustays` AS icu
LEFT JOIN `datathon-datasets.mimic_demo.services` AS se
ON icu.hadm_id = se.hadm_id
ORDER BY icustay_id
LIMIT 10
''')
| datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Notice that for a single ICU stay, there may be multiple services. The following query finds the first service from ICU stays from patients, and indicates whether the last service before ICU admission was surgical. | run_query('''
WITH serv AS (
SELECT
icu.hadm_id,
icu.icustay_id,
se.curr_service,
IF(curr_service like '%SURG' OR curr_service = 'ORTHO', 1, 0) AS surgical,
RANK() OVER (PARTITION BY icu.hadm_id ORDER BY se.transfertime DESC) as rank
FROM `datathon-datasets.mimic_demo.icustays` AS icu
LEFT JOI... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Finally, we are ready to add this surgical exclusion label to the cohort generation table we had before by joining the two tables. For the convenience of later analysis, we rename some columns, and filter out patients more than 100 years old. | df = run_query('''
WITH co AS (
SELECT
icu.subject_id,
icu.hadm_id,
icu.icustay_id,
pat.dob,
TIMESTAMP_DIFF(icu.outtime, icu.intime, DAY) AS icu_length_of_stay,
DATE_DIFF(DATE(icu.intime), DATE(pat.dob), YEAR) AS age,
RANK() OVER (PARTITION BY icu.subject_id ORDER BY icu.intime) AS icustay... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
The original MIMIC tutorial, also showed why the first_careunit field of the icustays table is not always the same as the surgical bit derived from the services table. It also demonstrated how the pandas dataframes returned from the BigQuery queries can be further processed in Python for summarization. We will not redo... | data = df[['age', 'first_stay', 'short_stay']]
data.reindex(np.random.permutation(data.index))
training_df=data.head(100)
validation_df=data.tail(27)
print "Training data summary:"
display(training_df.describe())
print "Validation data summary:"
display(validation_df.describe()) | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
And let's quickly check the label distribution for the features. | display(training_df.groupby(['short_stay', 'first_stay']).count())
fig, ax = plt.subplots()
shorts = training_df[training_df.short_stay==1].age
longs = training_df[training_df.short_stay==0].age
colors = ['b', 'g']
ax.hist([shorts, longs], bins=10, color=colors, label=['short_stay=1', 'short_stay=0'])
ax.set_xlabel(... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Let's first build a linear regression model to predict the numeric value of "short_stay" based on age and first_stay features. You can tune the parameters on the right-hand side and observe differences in the evaluation result. | #@title Linear Regression Parameters {display-mode:"both"}
BATCH_SIZE = 5 # @param
NUM_EPOCHS = 100 # @param
first_stay = tf.feature_column.numeric_column('first_stay')
age = tf.feature_column.numeric_column('age')
# Build linear regressor
linear_regressor = tf.estimator.LinearRegressor(feature_columns=[first_stay, a... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Remember that the label short_stay is actually categorical, with the value 1 for an ICU stay of 1 day or less and value 0 for stays of length 2 days or more. So a classification model better fits this task. Here we try a deep neural networks model using the DNNClassifier estimator. Notice the little changes from the re... | #@title ML Training example {display-mode:"both"}
BATCH_SIZE = 5 # @param
NUM_EPOCHS = 100 # @param
HIDDEN_UNITS=[10, 10] # @param
# Build linear regressor
classifier = tf.estimator.DNNClassifier(
feature_columns=[first_stay, age],
hidden_units=HIDDEN_UNITS)
# Train the Model.
model = classifier.... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
The ANZICS dataset also has some demo tables that you can explore. For example, in the APD and CCR tables, we can check the number of elements of certain types among the 100 demo records. | run_query('''
SELECT
hospitalclassification AS Type,
COUNT(*) AS NumHospitals
FROM `datathon-datasets.anzics_demo.apd`
GROUP BY 1
ORDER BY 2 DESC
''')
run_query('''
SELECT
iculevelname AS ICU_Level,
COUNT(*) AS NumHospitals,
SUM(hospitalbeds) AS NumBeds
FROM `datathon-datasets.anzics_demo.ccr`
GROUP BY 1
ORD... | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Congratulations! Now you have finished this datathon tutorial, and ready to explore the real data by querying Google BigQuery. To do so, simply remove the _demo suffix in the dataset names.
For example, the table mimic_demo.icustays becomes mimic.icustays when you need the actual MIMIC data. Now, let's do the substitut... | run_query('''
SELECT
COUNT(*) AS num_rows
FROM
`datathon-datasets.mimic.icustays`
''')
run_query('''
SELECT
subject_id,
hadm_id,
icustay_id
FROM
`datathon-datasets.mimic.icustays`
LIMIT 10
''') | datathon/anzics18/tutorial.ipynb | GoogleCloudPlatform/healthcare | apache-2.0 |
Build the MLP
Now we use the provided layers from Lasagne to build our MLP | def build_mlp(n_in, n_hidden, n_out, input_var=None):
# Put your code here to build the MLP using Lasagne
| 2015-10_Lecture/Lecture2/code/3_Intro_Lasagne.ipynb | UKPLab/deeplearning4nlp-tutorial | apache-2.0 |
Create the Train Function
After loading the data and defining the MLP, we can now create the train function. | # Parameters
n_in = 28*28
n_hidden = 50
n_out = 10
# Create the necessary training and predict labels function
| 2015-10_Lecture/Lecture2/code/3_Intro_Lasagne.ipynb | UKPLab/deeplearning4nlp-tutorial | apache-2.0 |
Train the model
We run the training for some epochs and output the accurarcy of our network | #Put your code here to train the model using mini batches
| 2015-10_Lecture/Lecture2/code/3_Intro_Lasagne.ipynb | UKPLab/deeplearning4nlp-tutorial | apache-2.0 |
Let's determine the current through R1. There are many ways to solve this; the easiest is to combine the sources, combine the resistances, and then use Ohm's law. The result is a function of $V_x$: | Vx = V('V_x').Voc
I = (cct.V1.V - 4 * Vx) / (cct.R1.Z + cct.R2.Z) | doc/examples/notebooks/dependent_source_example1.ipynb | mph-/lcapy | lgpl-2.1 |
Now given the current, we can use Ohm's law to determine the voltage drop across R1. | I * cct.R1.Z
cct.V1.V - I * cct.R1.Z | doc/examples/notebooks/dependent_source_example1.ipynb | mph-/lcapy | lgpl-2.1 |
Thus we know that $V_x = 3 V_x + 2$ or $V_x = -1$. Of course, Lcapy can determine this directly. Here Ox is the name of the open circuit over which we wish to determine the voltage difference: | cct.Ox.V | doc/examples/notebooks/dependent_source_example1.ipynb | mph-/lcapy | lgpl-2.1 |
Alternatively, we can query Lcapy for the voltage at node 'x' with respect to ground. This gives the same result. | cct['x'].V | doc/examples/notebooks/dependent_source_example1.ipynb | mph-/lcapy | lgpl-2.1 |
Let's check the current with Lcapy: | cct.R1.I | doc/examples/notebooks/dependent_source_example1.ipynb | mph-/lcapy | lgpl-2.1 |
Simulation setup | cutoff = 50*unit.angstrom
useMinimize = True
epsilon_r = 80.
temperature = 300*unit.kelvin
kT = unit.BOLTZMANN_CONSTANT_kB*temperature
timestep = 10*unit.femtoseconds;
steps_eq = 5000
steps_production = 2e4
steps_total = steps_eq + steps_production | yukawa/yukawa.ipynb | mlund/openmm-examples | mit |
Convenience functions
A set of independent functions, useful for setting up OpenMM. | def findForce(system, forcetype, add=True):
""" Finds a specific force in the system force list - added if not found."""
for force in system.getForces():
if isinstance(force, forcetype):
return force
if add==True:
system.addForce(forcetype())
return findForce(system, forc... | yukawa/yukawa.ipynb | mlund/openmm-examples | mit |
Setup simulation | pdb = app.PDBFile('squares.pdb')
forcefield = app.ForceField('yukawa.xml')
system = forcefield.createSystem(pdb.topology, nonbondedMethod=app.CutoffPeriodic, nonbondedCutoff=cutoff )
box = np.array(pdb.topology.getPeriodicBoxVectors()).diagonal()
harmonic = findForce(system, mm.HarmonicBondForce)
n... | yukawa/yukawa.ipynb | mlund/openmm-examples | mit |
Run simulation | simulation = app.Simulation(pdb.topology, system, integrator)
simulation.context.setPositions(pdb.positions)
if useMinimize:
print('Minimizing...')
simulation.minimizeEnergy()
print('Equilibrating...')
simulation.context.setVelocitiesToTemperature(300*unit.kelvin)
simulation.step(steps_eq)
simulation.reporte... | yukawa/yukawa.ipynb | mlund/openmm-examples | mit |
通过上面的可视化过程，我们可以认为其数据的分布是近似线性的，所以我们可以用简单的线性回归模型来去拟合数据。 | from sklearn.linear_model import LinearRegression
linreg = LinearRegression()
trainx = waiting.values.reshape(272, 1)
trainy = eruptions.values
linreg.fit(trainx, trainy)
print "intercept b0 = %f" % linreg.intercept_
print "coefficiet b1 = %f" % linreg.coef_[0]
x = np.linspace(40, 100, 200)
x = x.reshape(200,1)
y = l... | 2.LINEAR_MODELS_FOR_REGRESSION/Old_Faithful_Geyser_Experiment.ipynb | jasonding1354/PRML_Notes | mit |
我们可以得到拟合的结果和参数：
$$\textbf{Y} = \beta_0+\beta_1\textbf{X}+\epsilon=-1.874016+0.075628\textbf{X}+\epsilon$$ | residues = trainy-linreg.predict(trainx)
fig = plt.figure(figsize=(8,8))
ax = fig.add_subplot()
plt.scatter(waiting.values, residues)
plt.xlabel("waiting")
plt.ylabel("residues")
plt.title("piecewise linear")
plt.show() | 2.LINEAR_MODELS_FOR_REGRESSION/Old_Faithful_Geyser_Experiment.ipynb | jasonding1354/PRML_Notes | mit |
上面对余量residues进行分析，发现可视化residues后，residues依然存在某种模式(pattern)，resudue不能很好的模拟随机的噪声，所以现在得到模型还不能很好地来描述数据，于是我们需要进一步来优化修正。
从residues图形我们可以看出其是分段线性的，在waiting=70这里点上将数据分成两部分，我们可以通过人为观察来构造几个非线性函数将原数据转化成新的特征，比如这里我们新加入两个特征，分别是$max(0, \textbf{$X_i$}-68)$和$max(0, \textbf{$X_i$}-72)$。
下面是加入非线性转换得到的线性模型的代码： | xi = waiting.values
feature2 = np.maximum(0, xi-68)
feature3 = np.maximum(0, xi-72)
train_data = {'f1': xi, 'f2': feature2, 'f3': feature3}
trainXFrame = pd.DataFrame(train_data)
linreg2 = LinearRegression()
linreg2.fit(trainXFrame.values, trainy)
x = np.linspace(40, 100, 200)
f2 = np.maximum(0, x-68)
f3 = np.maximum... | 2.LINEAR_MODELS_FOR_REGRESSION/Old_Faithful_Geyser_Experiment.ipynb | jasonding1354/PRML_Notes | mit |
$$\textbf{$Y_i$} = \beta_0+\beta_1\textbf{$X_i$}+\beta_2max(0, \textbf{$X_i$}-68)+\beta_3max(0, \textbf{$X_i$}-72)+\epsilon_i$$ | residues2 = trainy-linreg2.predict(trainXFrame.values)
fig = plt.figure(figsize=(8,8))
ax = fig.add_subplot()
plt.scatter(waiting.values, residues2)
plt.xlabel("waiting")
plt.ylabel("residues")
plt.show() | 2.LINEAR_MODELS_FOR_REGRESSION/Old_Faithful_Geyser_Experiment.ipynb | jasonding1354/PRML_Notes | mit |
Visualizating a scan of a male head
Included in ipyvolume, is a visualuzation of a scan of a human head, see the sourcecode for more details. | import ipyvolume as ipv
fig = ipv.figure()
vol_head = ipv.examples.head(max_shape=128);
vol_head.ray_steps = 400
ipv.view(90, 0) | docs/source/examples/volshow.ipynb | maartenbreddels/ipyvolume | mit |
Instead of "bagging" the CSV file we will use this create a metadata rich netCDF file.
We can convert the table to a DSG, Discrete Sampling Geometry, using pocean.dsg. The first thing we need to do is to create a mapping from the data column names to the netCDF axes. | axes = {"t": "time", "x": "lon", "y": "lat", "z": "depth"} | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
Now we can create a Orthogonal Multidimensional Timeseries Profile object... | import os
import tempfile
from pocean.dsg import OrthogonalMultidimensionalTimeseriesProfile as omtsp
output_fp, output = tempfile.mkstemp()
os.close(output_fp)
ncd = omtsp.from_dataframe(df.reset_index(), output=output, axes=axes, mode="a") | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
... And add some extra metadata before we close the file. | naming_authority = "ioos"
st_id = "Station1"
ncd.naming_authority = naming_authority
ncd.id = st_id
print(ncd)
ncd.close() | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
Time to create the archive for the file with BagIt. We have to create a folder for the bag. | temp_bagit_folder = tempfile.mkdtemp()
temp_data_folder = os.path.join(temp_bagit_folder, "data") | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
Now we can create the bag and copy the netCDF file to a data sub-folder. | import shutil
import bagit
bag = bagit.make_bag(temp_bagit_folder, checksum=["sha256"])
shutil.copy2(output, temp_data_folder + "/parameter1.nc") | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
Last, but not least, we have to set bag metadata and update the existing bag with it. | urn = "urn:ioos:station:{naming_authority}:{st_id}".format(
naming_authority=naming_authority, st_id=st_id
)
bag_meta = {
"Bag-Count": "1 of 1",
"Bag-Group-Identifier": "ioos_bagit_testing",
"Contact-Name": "Kyle Wilcox",
"Contact-Phone": "907-230-0304",
"Contact-Email": "axiom+ncei@axiomdatasc... | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
That is it! Simple and efficient!!
The cell below illustrates the bag directory tree.
(Note that the commands below will not work on Windows and some *nix systems may require the installation of the command tree, however, they are only need for this demonstration.) | !tree $temp_bagit_folder
!cat $temp_bagit_folder/manifest-sha256.txt | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
We can add more files to the bag as needed. | shutil.copy2(output, temp_data_folder + "/parameter2.nc")
shutil.copy2(output, temp_data_folder + "/parameter3.nc")
shutil.copy2(output, temp_data_folder + "/parameter4.nc")
bag.save(manifests=True, processes=4)
!tree $temp_bagit_folder
!cat $temp_bagit_folder/manifest-sha256.txt | notebooks/2017-11-01-Creating-Archives-Using-Bagit.ipynb | ioos/notebooks_demos | mit |
Weekly sentiment score analysis | score_data = pd.read_csv("../data/nyt_bitcoin_with_score.csv", index_col='time',
parse_dates=[0], date_parser=lambda x: datetime.datetime.strptime(x, time_format))
score_data.head() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Ratio of "negative", "neutral", "positive" | score_data.sentiment.unique()
score_data.groupby("sentiment").sentiment.count().plot(kind='bar',rot=0) | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Massively negative ratings!!!! Is this special to bitcoin news? To double check, run the same analysis on news with headline including "internet".
Alternate news analysis (digress) | internet_news = pd.read_csv("../data/nyt_internet_with_score.csv", index_col='time',
parse_dates=[0], date_parser=lambda x: datetime.datetime.strptime(x, time_format))
internet_news.head()
internet_news.groupby("sentiment").sentiment.count() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
So it seems most of news would be classified as negative by the Stanford classifier. How about other classifier?
Indico.io sentiment score
Here we analyze the score generated by Indico.io API on the same dataset. The score is between 0 and 1, and scores >0.5 are considered as positive. | indico_news = pd.read_csv("../data/indico_nyt_bitcoin.csv", index_col='time',
parse_dates=[0], date_parser=lambda x: datetime.datetime.strptime(x, time_format))
indico_news.head()
indico_news.indico_score.describe() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Distribution | indico_news.indico_score.plot(kind='hist') | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
The distribution of indico score looks quite like a normal distribution, which is better than the Stanford one of course. So maybe we should try using indico score? | indico_news.resample('w', how='mean').plot() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Let's try again with news about "internet". | indico_news = pd.read_csv("../data/indico_nyt_internet.csv", index_col='time',
parse_dates=[0], date_parser=lambda x: datetime.datetime.strptime(x, time_format))
indico_news.head()
indico_news.indico_score.plot(kind='hist', bins=20) | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Again, it is a normal distribution. I am not sure whether the reasonable distribution of sentiment about a thing should be like this? Because this is not a very neutral thing, and we should probably expect the distribution be positively skewed.
This needs to be further studied for validity. | indico_news.indico_score.resample('w', how='mean').plot()
indico_news.indico_score.describe() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Weekly distribution | weekly_news_count = score_data.resample('w', how='count').fillna(0)
weekly_news_count.sentiment.describe() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
News distribution by week | weekly_news_count.sentiment.plot() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
News distribution | weekly_news_count.sentiment.plot(kind='hist') | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Average weekly sentiment score | weekly_score = score_data.resample('d', how='mean').fillna(0)
weekly_score.head() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Score Distribution | weekly_score.sentimentValue.plot(kind='hist') | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
Score distribution by week | weekly_score.plot() | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
We miss news about bitcoin for about half of the all time. Therefore we try keyword "internet". | missing_news = 100*weekly_score[weekly_score.sentimentValue==0].count()/float(weekly_score.count())
print "Percentage of weeks without news: %f%%" % missing_news | notes/news_analysis.ipynb | yyl/btc-price-analysis | gpl-2.0 |
<center>Find Simulations and Load data | # Search for a simulation. Hopefully the results will be from different codes. NOTE that this could be done more manually so that we don't "hope" but know.
A = scsearch( q=[1,4], nonspinning=True, verbose=True )
# Select which of the search results we wish to keep
U,V = A[77],A[131]
# Load the modes
u = gwylm(U,lmax... | review/notebooks/compare_waveforms_from_two_codes.ipynb | Cyberface/nrutils_dev | mit |
<center>Recompose the Waveforms | #
theta,phi = 0,0
a,b = u.recompose(theta,phi,kind='strain'),v.recompose(theta,phi,kind='strain') | review/notebooks/compare_waveforms_from_two_codes.ipynb | Cyberface/nrutils_dev | mit |
<center>Plot the amplitudes to verify correct scaling between GT and SXS waveforms | figure( figsize=2*array([5,3]) )
plot( a.t - a.intrp_t_amp_max, a.amp )
plot( b.t - b.intrp_t_amp_max, b.amp )
gca().set_yscale("log", nonposy='clip')
ylim([1e-5,1e-1])
xlim([-400,100])
title('the amplitudes should be approx. the same')
a.plot();b.plot(); | review/notebooks/compare_waveforms_from_two_codes.ipynb | Cyberface/nrutils_dev | mit |
When you get home, you ask your roommate, who conveniently happens to work in a nuclear chemistry lab, to measure some of the properties of this odd material that you've been given. When she comes home that night, she says that she managed to purify the sample and measure its radioactive decay rate (which is to say, t... | # this block of code reads in the data files. Don't worry too much about how they
# work right now -- we'll talk about that in a couple of weeks weeks!
import numpy as np
'''
count_times = the time since the start of data-taking when the data was
taken (in seconds)
count_rate = the number of counts ... | past-semesters/fall_2016/day-by-day/day06-modeling-radioactivity-day1/radioactivity_modeling.ipynb | ComputationalModeling/spring-2017-danielak | agpl-3.0 |
Using the four numpy arrays created in the cell above, plot the measured count rates as a function of time and, on a separate plot, plot the measured sample amounts as a function of time. What do you notice about these plots, compared to the ones from your analytic estimate? Also, if you inspect the data, approximate... | # put your code here! add additional cells if necessary.
| past-semesters/fall_2016/day-by-day/day06-modeling-radioactivity-day1/radioactivity_modeling.ipynb | ComputationalModeling/spring-2017-danielak | agpl-3.0 |
Based on the plot in the previous cell, what do you think is going on?
put your answer here!
How might you modify your model to emulate this behavior? In other words, how might you modify the equation for decay rate to get something that matches the observed decay rate?
put your answer here!
More complex data manipul... | # put your code here! add additional cells if necessary.
| past-semesters/fall_2016/day-by-day/day06-modeling-radioactivity-day1/radioactivity_modeling.ipynb | ComputationalModeling/spring-2017-danielak | agpl-3.0 |
Feedback on this assignment
Please fill out the form that appears when you run the code below. You must completely fill this out in order for your group to receive credit for the assignment! | from IPython.display import HTML
HTML(
"""
<iframe
src="https://goo.gl/forms/F1MvFMDpIWPScchr2?embedded=true"
width="80%"
height="1200px"
frameborder="0"
marginheight="0"
marginwidth="0">
Loading...
</iframe>
"""
) | past-semesters/fall_2016/day-by-day/day06-modeling-radioactivity-day1/radioactivity_modeling.ipynb | ComputationalModeling/spring-2017-danielak | agpl-3.0 |
Con un dataframe | z = pd.DataFrame(np.array([parse_latex(r'x+y=6'),parse_latex(r'x-y=0')]))
z
z.columns
for i in z.index:
print(z.loc[i][0])
latex(solve(z.loc[i][0] for i in z.index))
for i in z.index: print(z.loc[i][0]) | tmp/.ipynb_checkpoints/Probando_parse_latex-checkpoint.ipynb | crdguez/mat4ac | gpl-3.0 |
Build a simple keras DNN model
We will use feature columns to connect our raw data to our keras DNN model. Feature columns make it easy to perform common types of feature engineering on your raw data. For example, you can one-hot encode categorical data, create feature crosses, embeddings and more. We'll cover these in... | INPUT_COLS = [
"pickup_longitude",
"pickup_latitude",
"dropoff_longitude",
"dropoff_latitude",
"passenger_count",
]
# Create input layer of feature columns
# TODO 1
feature_columns = {
colname: tf.feature_column.numeric_column(colname) for colname in INPUT_COLS
} | courses/machine_learning/deepdive2/introduction_to_tensorflow/solutions/3_keras_sequential_api.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Next, we create the DNN model. The Sequential model is a linear stack of layers and when building a model using the Sequential API, you configure each layer of the model in turn. Once all the layers have been added, you compile the model. | # Build a keras DNN model using Sequential API
# TODO 2a
model = Sequential(
[
DenseFeatures(feature_columns=feature_columns.values()),
Dense(units=32, activation="relu", name="h1"),
Dense(units=8, activation="relu", name="h2"),
Dense(units=1, activation="linear", name="output"),
... | courses/machine_learning/deepdive2/introduction_to_tensorflow/solutions/3_keras_sequential_api.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Next, to prepare the model for training, you must configure the learning process. This is done using the compile method. The compile method takes three arguments:
An optimizer. This could be the string identifier of an existing optimizer (such as rmsprop or adagrad), or an instance of the Optimizer class.
A loss funct... | # TODO 2b
# Create a custom evalution metric
def rmse(y_true, y_pred):
return tf.sqrt(tf.reduce_mean(tf.square(y_pred - y_true)))
# Compile the keras model
model.compile(optimizer="adam", loss="mse", metrics=[rmse, "mse"]) | courses/machine_learning/deepdive2/introduction_to_tensorflow/solutions/3_keras_sequential_api.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
There are various arguments you can set when calling the .fit method. Here x specifies the input data which in our case is a tf.data dataset returning a tuple of (inputs, targets). The steps_per_epoch parameter is used to mark the end of training for a single epoch. Here we are training for NUM_EVALS epochs. Lastly, fo... | # TODO 3
%%time
steps_per_epoch = NUM_TRAIN_EXAMPLES // (TRAIN_BATCH_SIZE * NUM_EVALS)
LOGDIR = "./taxi_trained"
history = model.fit(
x=trainds,
steps_per_epoch=steps_per_epoch,
epochs=NUM_EVALS,
validation_data=evalds,
callbacks=[TensorBoard(LOGDIR)],
) | courses/machine_learning/deepdive2/introduction_to_tensorflow/solutions/3_keras_sequential_api.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
The kinematics of local F stars
Load the TGAS and 2MASS data that we use, among other things, select F-type stars: | # Load TGAS and 2MASS
tgas= gaia_tools.load.tgas()
twomass= gaia_tools.load.twomass()
jk= twomass['j_mag']-twomass['k_mag']
dm= -5.*numpy.log10(tgas['parallax'])+10.
mj= twomass['j_mag']-dm | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
Load the dwarf locus and select TGAS F-type stars with good parallaxes: | # Select F stars
sp= effsel.load_spectral_types()
sptype= 'F'
jkmin= (sp['JH']+sp['HK'])[sp['SpT']=='%s0V' % sptype]
if sptype == 'M':
jkmax= (sp['JH']+sp['HK'])[sp['SpT']=='%s5V' % sptype]
else:
jkmax= (sp['JH']+sp['HK'])[sp['SpT']=='%s9V' % sptype]
jmin= main_sequence_cut_r(jkmax)
jmax= main_sequence_cut_r(jk... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
Using XD to get $\sigma_{v_z}$ from the TGAS data alone
We will determine $\sigma_{v_z}^2$ from the TGAS data alone by, that is, just using the TGAS parallaxes and proper motions. This only works close to the plane, because further away the $v_z$ component of the velocity only weakly projects onto the proper motions. F... | # Compute XYZ
lb= bovy_coords.radec_to_lb(tgas['ra'],tgas['dec'],degree=True,epoch=None)
XYZ= bovy_coords.lbd_to_XYZ(lb[:,0],lb[:,1],1./tgas['parallax'],degree=True)
# Generate vradec
vradec= numpy.array([bovy_coords._K/tgas['parallax']*tgas['pmra'],
bovy_coords._K/tgas['parallax']*tgas['pmdec']])... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
If we use the same bins as those in which we computed the density, we have the following number of stars in each bin: | zbins= numpy.arange(-0.4125,0.425,0.025)
indx= (numpy.fabs(XYZ[:,2]) > -0.4125)\
*(numpy.fabs(XYZ[:,2]) <= 0.4125)\
*(numpy.sqrt(XYZ[:,0]**2.+XYZ[:,1]**2.) < 0.2)
bovy_plot.bovy_print(axes_labelsize=17.,text_fontsize=12.,xtick_labelsize=15.,ytick_labelsize=15.)
_= hist(XYZ[indx,2],bins=zbins)
xlabel(r'$Z\,(... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
Now we perform an example fit: | def combined_sig2(amp,mean,covar):
indx= numpy.sqrt(covar) < 30.
tamp= amp[indx]/numpy.sum(amp[indx])
return (numpy.sum(tamp*(covar+mean**2.)[indx])-numpy.sum(tamp*mean[indx])**2.)
# Fit with mix of Gaussians
ii= 14
print("Z/pc:",500.*(numpy.roll(zbins,-1)+zbins)[:-1][ii])
indx= (XYZ[:,2] > zbins[ii])\
... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
The following function computes a bootstrap estimate of the uncertainty in $\sigma_{v_z}$: | def bootstrap(nboot,vrd,vrd_cov,proj,ngauss=2):
out= numpy.empty(nboot)
for ii in range(nboot):
# Draw w/ replacement
indx= numpy.floor(numpy.random.uniform(size=len(vrd))*len(vrd)).astype('int')
ydata= vrd[indx]
ycovar= vrd_cov[indx]
initamp= numpy.random.uniform(size=ng... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
A test case: | ii= 7
indx= (XYZ[:,2] > zbins[ii])\
*(XYZ[:,2] <= zbins[ii+1])\
*(numpy.sqrt(XYZ[:,0]**2.+XYZ[:,1]**2.) < 0.2)
b= bootstrap(20,vradec.T[indx],vradec_cov[indx],proj[indx],ngauss=2)
zbins[ii], numpy.mean(numpy.sqrt(b)), numpy.std(numpy.sqrt(b)) | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
Now we do the XD fit to each $Z$ bin and compute the uncertainty using bootstrap: | savefilename= 'Fstar-sigz.sav'
if not os.path.exists(savefilename):
zbins= numpy.arange(-0.4125,0.425,0.025)
nboot= 200
nstar= numpy.zeros(len(zbins)-1)-1
sig2z= numpy.zeros(len(zbins)-1)-1
sig2z_err= numpy.zeros(len(zbins)-1)-1
all_sam= numpy.zeros((len(zbins)-1,nboot))-1
ngauss= 2
star... | py/Fstar-kinematics.ipynb | jobovy/simple-m2m | mit |
Create a SparkSession and give it a name.
Note: This will start the spark client console -- there is no need to run spark-shell directly. | spark = SparkSession \
.builder \
.appName("PythonPi") \
.getOrCreate() | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
partitions is the number of spark workers to partition the work into. | partitions = 2 | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
n is the number of random samples to calculate | n = 100000000 | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
This is the sampling function. It generates numbers in the square from (-1, -1) to (1, 1), and returns 1 if it falls inside the unit circle, and 0 otherwise. | def f(_):
x = random() * 2 - 1
y = random() * 2 - 1
return 1 if x ** 2 + y ** 2 <= 1 else 0 | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
Here's where we farm the work out to Spark. | count = spark.sparkContext \
.parallelize(range(1, n + 1), partitions) \
.map(f) \
.reduce(add)
print("Pi is roughly %f" % (4.0 * count / n)) | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
Shut down the spark server. | spark.stop() | examples/Jupyter Spark example.ipynb | mreid-moz/jupyter-spark | mpl-2.0 |
Now that we see how the data is organized, let's use a MLP, with an architecture as the one taught in "Machine Learning" from Stanford in coursera.org to recognize the digits.
The MLP will have 3 layers.
The input layer will have 784 units
The hidden layer will have 25 units
The output layer will have, obviously, 10 u... | from sklearn.neural_network import MLPClassifier
myX = myTrainDf[myTrainDf.columns[1:]]
myY = myTrainDf[myTrainDf.columns[0]]
# Use 'adam' solver for large datasets, alpha is the regularization term.
# Display the optimization by showing the cost.
myClf = MLPClassifier(hidden_layer_sizes=25, activation='logistic', so... | digit_recognition/notebook.ipynb | drublackberry/fantastic_demos | mit |
The results are quite good, as expected, now let's make a prediction for the test set. | myYtestPred = myClf.predict(myTestDf)
myOutDf = pd.DataFrame(index=myTestDf.index+1, data=myYtestPred)
myOutDf.reset_index().to_csv('submission.csv', header=['ImageId', 'Label'],index=False) | digit_recognition/notebook.ipynb | drublackberry/fantastic_demos | mit |
Cross-validation set and regularization parameter
The following code will split the test set into a training set and cross-validation set. | REG_ARRAY = [100, 1, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6, 0]
def splitDataset (aDf, aFrac):
aTrainDf = aDf.sample(frac=aFrac)
aXvalDf = aDf.iloc[[x for x in aDf.index if x not in aTrainDf.index]]
return aTrainDf, aXvalDf
mySampleTrainDf, mySampleXvalDf = splitDataset(myTrainDf, 0.8)
myAccuracyDf = pd.DataF... | digit_recognition/notebook.ipynb | drublackberry/fantastic_demos | mit |
From here one can tell that the default regularization parameter (around 1e-5)
Multiple layers | REG_ARRAY = [100, 1, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6, 0]
myAccuracyDf = pd.DataFrame(index=REG_ARRAY, columns=['Accuracy'])
for myAlpha in REG_ARRAY:
print ('Training with regularization param ', str(myAlpha))
myClf = MLPClassifier(hidden_layer_sizes=[400, 400, 100, 25], activation='logistic', solver='adam',... | digit_recognition/notebook.ipynb | drublackberry/fantastic_demos | mit |
Let's produce a new output file with no regularization and a complex MLP of 784x400x400x100x25x10 | myClf = MLPClassifier(hidden_layer_sizes=[400, 400, 100, 25], activation='logistic', solver='adam', alpha=0, verbose=True)
myClf.fit(myTrainDf[myTrainDf.columns[1:]], myTrainDf['label'])
myYtestPred = myClf.predict(myTestDf)
myOutDf = pd.DataFrame(index=myTestDf.index+1, data=myYtestPred)
myOutDf.reset_index().to_csv('... | digit_recognition/notebook.ipynb | drublackberry/fantastic_demos | mit |
Model Averaging
<table class="tfo-notebook-buttons" align="left">
<td>
<a target="_blank" href="https://www.tensorflow.org/addons/tutorials/average_optimizers_callback"><img src="https://www.tensorflow.org/images/tf_logo_32px.png" />View on TensorFlow.org</a>
</td>
<td>
<a target="_blank" href="https://co... | !pip install -U tensorflow-addons
import tensorflow as tf
import tensorflow_addons as tfa
import numpy as np
import os | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Build Model | def create_model(opt):
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(64, activation='relu'),
tf.keras.layers.Dense(64, activation='relu'),
tf.keras.layers.Dense(10, activation='softmax')
])
model.compile(optimi... | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Prepare Dataset | #Load Fashion MNIST dataset
train, test = tf.keras.datasets.fashion_mnist.load_data()
images, labels = train
images = images/255.0
labels = labels.astype(np.int32)
fmnist_train_ds = tf.data.Dataset.from_tensor_slices((images, labels))
fmnist_train_ds = fmnist_train_ds.shuffle(5000).batch(32)
test_images, test_labels... | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
We will be comparing three optimizers here:
Unwrapped SGD
SGD with Moving Average
SGD with Stochastic Weight Averaging
And see how they perform with the same model. | #Optimizers
sgd = tf.keras.optimizers.SGD(0.01)
moving_avg_sgd = tfa.optimizers.MovingAverage(sgd)
stocastic_avg_sgd = tfa.optimizers.SWA(sgd) | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Both MovingAverage and StochasticAverage optimizers use ModelAverageCheckpoint. | #Callback
checkpoint_path = "./training/cp-{epoch:04d}.ckpt"
checkpoint_dir = os.path.dirname(checkpoint_path)
cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_dir,
save_weights_only=True,
verbose=1)
... | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Train Model
Vanilla SGD Optimizer | #Build Model
model = create_model(sgd)
#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[cp_callback])
#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accuracy) | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Moving Average SGD | #Build Model
model = create_model(moving_avg_sgd)
#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[avg_callback])
#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accuracy... | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
Stocastic Weight Average SGD | #Build Model
model = create_model(stocastic_avg_sgd)
#Train the network
model.fit(fmnist_train_ds, epochs=5, callbacks=[avg_callback])
#Evalute results
model.load_weights(checkpoint_dir)
loss, accuracy = model.evaluate(test_images, test_labels, batch_size=32, verbose=2)
print("Loss :", loss)
print("Accuracy :", accur... | site/en-snapshot/addons/tutorials/average_optimizers_callback.ipynb | tensorflow/docs-l10n | apache-2.0 |
At first, we need to define the dataset names and temporal ranges. Please note that the datasets have different time ranges. So we will download the data from 1981, when CHRIPS starts (ARC2 is from 1983). | dh=datahub.datahub(server,version,API_key)
dataset1='noaa_arc2_africa_01'
variable_name1 = 'pr'
time_start = '1983-01-01T00:00:00'
time_end = '2018-01-01T00:00:00' | api-examples/ARC2_download_example.ipynb | planet-os/notebooks | mit |
Then we define spatial range. In this case we define all Africa. Keep in mind that it's a huge area and downloading all the data from Africa for 35 years is a lot and can take even several hours. | area_name = 'Africa'
latitude_north = 42.24; longitude_west = -24.64
latitude_south = -45.76; longitude_east = 60.28 | api-examples/ARC2_download_example.ipynb | planet-os/notebooks | mit |
Download the data with package API
Create package objects
Send commands for the package creation
Download the package files | package_arc2_africa_01 = package_api.package_api(dh,dataset1,variable_name1,longitude_west,longitude_east,latitude_south,latitude_north,time_start,time_end,area_name=area_name)
package_arc2_africa_01.make_package()
package_arc2_africa_01.download_package() | api-examples/ARC2_download_example.ipynb | planet-os/notebooks | mit |
The pandas library in Python is designed to make it easy to process tabular data, and we can use it display the CSV file so that it looks more like a table. In the next example, we import the library (and give it the short name pd), and then use its read_cv() method to slurp up the CSV file. | import pandas as pd
table = pd.read_csv("../data/formats/messages.csv")
table | notebooks/data_formats.ipynb | edinburghlivinglab/dds-notebooks | cc0-1.0 |
One issue to note is that if a value in the CSV file contains a comma, then we have to wrap that value with quote signs, as in "James, Ewan".
<a name="xml">XML</a>
XML (eXtensible Markup Language) is a W3C open standard, used widely for storing data and for transferring it between applications and services.
Here is a s... | from lxml import etree
tree = etree.parse("../data/formats/messages.xml")
print(etree.tostring(tree, pretty_print = True, encoding="unicode")) | notebooks/data_formats.ipynb | edinburghlivinglab/dds-notebooks | cc0-1.0 |
<a name="json">JSON</a>
JSON (JavaScript Object Notation) is intended as format for transferring data that is easier for humans to write and read than XML. Unlike CSV and XML, JSON lets us represent lists directly using [ and ]. For example, the list containing the two strings 'James' and 'Ewan' is ['James', 'Ewan'].
... | import json
fp = open("../data/formats/messages.json")
data = json.load(fp)
print(json.dumps(data, indent=2, sort_keys=True)) | notebooks/data_formats.ipynb | edinburghlivinglab/dds-notebooks | cc0-1.0 |
Load Data
As usual, let's start by loading some network data. This time round, we have a physician trust network, but slightly modified such that it is undirected rather than directed.
This directed network captures innovation spread among 246 physicians in for towns in Illinois, Peoria, Bloomington, Quincy and Galesb... | # Load the network.
G = cf.load_physicians_network()
# Make a Circos plot of the graph
import numpy as np
from circos import CircosPlot
nodes = sorted(G.nodes())
edges = G.edges()
edgeprops = dict(alpha=0.1)
nodecolor = plt.cm.viridis(np.arange(len(nodes)) / len(nodes))
fig = plt.figure(figsize=(6,6))
ax = fig.add_s... | 4. Cliques, Triangles and Graph Structures (Student).ipynb | SubhankarGhosh/NetworkX | mit |
Question
What can you infer about the structure of the graph from the Circos plot?
Structures in a Graph
We can leverage what we have learned in the previous notebook to identify special structures in a graph.
In a network, cliques are one of these special structures.
Cliques
In a social network, cliques are groups of... | # Example code.
def in_triangle(G, node):
"""
Returns whether a given node is present in a triangle relationship or not.
"""
# We first assume that the node is not present in a triangle.
is_in_triangle = False
# Then, iterate over every pair of the node's neighbors.
for nbr1, nbr2 in it... | 4. Cliques, Triangles and Graph Structures (Student).ipynb | SubhankarGhosh/NetworkX | mit |
Exercise
Can you write a function that takes in one node and its associated graph as an input, and returns a list or set of itself + all other nodes that it is in a triangle relationship with? Do not return the triplets, but the set/list of nodes.
Possible Implementation: If my neighbor's neighbor's neighbor includes m... | # Possible answer
def get_triangles(G, node):
neighbors = set(G.neighbors(node))
triangle_nodes = set()
"""
Fill in the rest of the code below.
"""
triangle_nodes.add(node)
is_in_triangle = False
# Then, iterate over every pair of the node's neighbors.
for nbr1, nbr2 in itertool... | 4. Cliques, Triangles and Graph Structures (Student).ipynb | SubhankarGhosh/NetworkX | mit |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.