def _checkpointGen(filePath,orCall,force,unpack,useNpy,*args,**kwargs):
# XXX assume pickling now, ends with 'npz'
# if the file from 'filePath' exists and 'force' is false, loads the file
# otherwise, calls 'orCall' and saves the result. *args and **kwargs
# are passed to 'orCall'.
# 'Unpack' unpacks the array upon a load. This makes it 'look' like a
# simple function call (returns the args, or a tuple list of args)
# use unpack if you aren't dealing with dictionaries or things like that
if pGenUtil.isfile(filePath) and not force:
if (useNpy):
return _npyLoad(filePath,unpack)
else:
# assume we pickle in binary
fh = open(filePath,'rb')
data = pickle.load(fh)
fh.close()
return data
else:
# couldn't find the file.
# make sure it exists
path = pGenUtil.getBasePath(filePath)
pGenUtil.ensureDirExists(path)
# POST: we can put our file here
dataToSave = orCall(*args,**kwargs)
# need to figure out if we need to unpack all the arguments..
if (useNpy):
_npySave(filePath,dataToSave)
else:
# open the file in binary format for writing
with open(filePath, 'wb') as fh:
pickle.dump(dataToSave,fh)
return dataToSave
def GetDemoInOut(demoName, baseDir=DemoDir(), raiseOnError=True):
"""
Returns the demo input and output directories, given a path baseDir and
name demoName. Recquires files to exist at "<baseDir><demoName>". If
encountering an error (e.g. permissions, something isn't mounted), raises
an error.
Args:
demoName: The name of the demo. Assumed to be the subdir under "basedir"
we want to use
baseDir: the base directory. Input and output directories are
"<baseDir><demoName>Input/" and "<baseDir><demoName>Output/", resp.
raiseOnError : if true, raises an error on an OS. otherwise, just
prints a warning that something went wrong.
Returns:
tuple of <inputDir>,<outputDir>
"""
fullBase = baseDir + demoName
inputV = pGenUtil.getSanitaryPath(fullBase + "/Input/")
outputV = pGenUtil.getSanitaryPath(fullBase + "/Output/")
try:
pGenUtil.ensureDirExists(inputV)
pGenUtil.ensureDirExists(outputV)
except OSError as e:
if (raiseOnError):
raise (e)
print("Warning, couldn't open demo directories based in " + fullBase +
". Most likely, not connected to JILA network")
return inputV, outputV
def _checkpointGen(filePath, orCall, force, unpack, useNpy, *args, **kwargs):
# XXX assume pickling now, ends with 'npz'
# if the file from 'filePath' exists and 'force' is false, loads the file
# otherwise, calls 'orCall' and saves the result. *args and **kwargs
# are passed to 'orCall'.
# 'Unpack' unpacks the array upon a load. This makes it 'look' like a
# simple function call (returns the args, or a tuple list of args)
# use unpack if you aren't dealing with dictionaries or things like that
if pGenUtil.isfile(filePath) and not force:
if (useNpy):
return _npyLoad(filePath, unpack)
else:
# assume we pickle in binary
fh = open(filePath, 'rb')
data = pickle.load(fh)
fh.close()
return data
else:
# couldn't find the file.
# make sure it exists
path = pGenUtil.getBasePath(filePath)
pGenUtil.ensureDirExists(path)
# POST: we can put our file here
dataToSave = orCall(*args, **kwargs)
# need to figure out if we need to unpack all the arguments..
if (useNpy):
_npySave(filePath, dataToSave)
else:
# open the file in binary format for writing
with open(filePath, 'wb') as fh:
pickle.dump(dataToSave, fh)
return dataToSave
def DemoJilaOrLocal(demoName,localPath):
"""
Looks for the demo dir in the default (jila-hosted) space. If nothing is
found, looks in the paths specified by localpath (where it puts input
and output directories according to its name)
Args:
demoName: see GetDemoInOut
localPath: equivalent of baseDir in GetDemoInOut. Where we put the input and Output directories for the unit test if JILA can't be found.
Returns:
tuple of <inputDir>,<outputDir>
"""
inDir,outDir = GetDemoInOut(demoName,raiseOnError=False)
if (not pGenUtil.dirExists(inDir)):
print("Warning: Couldn't connect to JILA's Network. Using local data.")
# get "sanitary paths" which as OS-indepdent (in theory..)
localPath = pGenUtil.ensureEnds(localPath,"/")
inDir = pGenUtil.getSanitaryPath(localPath)
outDir = pGenUtil.getSanitaryPath(localPath + "Output" + demoName +"/")
pGenUtil.ensureDirExists(outDir)
if (not pGenUtil.dirExists(inDir)):
# whoops...
raise IOError("Demo Directory {:s} not found anywhere.".\
format(inDir))
return inDir,outDir
def GetDemoInOut(demoName,baseDir=DemoDir(),raiseOnError=True):
"""
Returns the demo input and output directories, given a path baseDir and
name demoName. Recquires files to exist at "<baseDir><demoName>". If
encountering an error (e.g. permissions, something isn't mounted), raises
an error.
Args:
demoName: The name of the demo. Assumed to be the subdir under "basedir"
we want to use
baseDir: the base directory. Input and output directories are
"<baseDir><demoName>Input/" and "<baseDir><demoName>Output/", resp.
raiseOnError : if true, raises an error on an OS. otherwise, just
prints a warning that something went wrong.
Returns:
tuple of <inputDir>,<outputDir>
"""
fullBase = baseDir + demoName
inputV = pGenUtil.getSanitaryPath(fullBase + "/Input/")
outputV = pGenUtil.getSanitaryPath(fullBase + "/Output/")
try:
pGenUtil.ensureDirExists(inputV)
pGenUtil.ensureDirExists(outputV)
except OSError as e:
if (raiseOnError):
raise(e)
print("Warning, couldn't open demo directories based in " + fullBase +
". Most likely, not connected to JILA network")
return inputV,outputV
def saveFile(filePath, dataToSave, useNpy):
path = pGenUtil.getBasePath(filePath)
pGenUtil.ensureDirExists(path)
# need to figure out if we need to unpack all the arguments..
if (useNpy):
_npySave(filePath, dataToSave)
else:
# open the file in binary format for writing
with open(filePath, 'wb') as fh:
# XXX make protocol specifiable?
pickle.dump(dataToSave, fh, pickle.HIGHEST_PROTOCOL)
def saveFile(filePath,dataToSave,useNpy):
path = pGenUtil.getBasePath(filePath)
pGenUtil.ensureDirExists(path)
# need to figure out if we need to unpack all the arguments..
if (useNpy):
_npySave(filePath,dataToSave)
else:
# open the file in binary format for writing
with open(filePath, 'wb') as fh:
# XXX make protocol specifiable?
pickle.dump(dataToSave,fh,pickle.HIGHEST_PROTOCOL)
def run(fitter,fitterParams,fitterCoeff,dataClass,label,valid=0.05,
train="train.csv",test="test.csv",profile=False,nTrials=1,
force=True,forceFeat=True,plot=False):
trainFile = train
testFile = test
inDir,cacheDir,outDir = getDirsFromCmdLine()
# add the label for this run (ie: SVM/Boost/LogisticRegression)
outDir = pGenUtil.ensureDirExists(outDir + label +"/")
# get the directories we want
predictDir = pGenUtil.ensureDirExists(outDir + "predictions")
if (profile and plot):
profileDir = pGenUtil.ensureDirExists(outDir + "profile")
else:
profileDir = None
# get the data object, by cache or otherwise
dataObj = \
pCheckUtil.pipeline([[cacheDir+'data.pkl',getData,dataClass,outDir,
inDir+trainFile,valid,False,profileDir,]],forceFeat)
return analyze(dataObj,inDir,outDir,testFile,fitter,fitterParams,
fitterCoeff,label,dataClass,nTrials,force,plot)
def analyze(dataObj,dataDir,outDir,testFile,createFitter,fitterParams,
fitterCoeff,label,dataClass,nTrials,force,plot):
# 'createfitter' takes in the current iteration 'i', and returns a fitter
# e.g. "return LogisticRegression(C=[10,30,40][i])"
# 'fitterParams' gives the value of the parameters used at each iter.
predictDir = pGenUtil.ensureDirExists(outDir + "predictions/")
testDat = getData(dataClass,outDir,dataDir + testFile,test=True)
params = fitterParams()
fName = outDir+"accuracyTrials_{:d}repeats_{:d}params.pkl".format(nTrials,
len(params))
means,std=pCheckUtil.getCheckpoint(fName,getAllTrials,
force,params,outDir,predictDir,dataObj,testDat,nTrials,
fitterCoeff,createFitter,plot,dataClass)
# plot the accuracies versus the fit parameter.
if (plot):
plotAccuracies(outDir,label,means,std,params)
return means,std
fig = pPlotUtil.figure()
plt.subplot(1,2,1)
plt.title("Mean of g(xBar)-g(mu) approaches 0",fontsize=fontsize)
expMean = 0
plt.plot(nVals,means,'ko',label="Actual Mean")
plt.axhline(expMean,color='b',linestyle='--',
label="Expected Mean: {:.2g}".format(expMean))
plt.ylim(-min(means),max(means)*1.1)
plt.xlabel("Value of n for binomial",fontsize=fontsize)
plt.ylabel("Value of g(xBar)-g(mu)",fontsize=fontsize)
plt.legend(fontsize=fontsize)
pPlotUtil.tickAxisFont()
plt.subplot(1,2,2)
plt.semilogy(nVals,varReal,'ko',label="Actual Variance")
plt.semilogy(nVals,varDist,'b--',label="Expected Variance")
plt.title("Variance of g(xBar)-g(mu)\n approaches expected",
fontsize=fontsize)
plt.xlabel("Value of n for binomial",fontsize=fontsize)
plt.ylabel("Value of g(xBar) variance",fontsize=fontsize)
pPlotUtil.tickAxisFont()
plt.legend(fontsize=fontsize)
pPlotUtil.savefig(fig,outDir + "MeanVar")
if __name__ == '__main__':
_nVals = np.array([10,20,50,75,100,150,200,350,500,1000])
pGenUtil.ensureDirExists(outDir)
_p=1/3.
_nPoints = 1e5
dataMatrix = getBinomials(_nVals,_p,_nPoints)
plotBinomials(dataMatrix,_nVals,_p)
# force floating point division. Can still use integer with //
from __future__ import division
# This file is used for importing the common utilities classes.
import numpy as np
import matplotlib.pyplot as plt
# need to add the utilities class. Want 'home' to be platform independent
# import the patrick-specific utilities
import GenUtilities as pGenUtil
import PlotUtilities as pPlotUtil
import CheckpointUtilities as pCheckUtil
from scipy.stats import norm
outDir = "./out/"
pGenUtil.ensureDirExists(outDir)
mean = 0
stdev = 1
epsilon = stdev / 100
nPoints = 1000
normDist = norm(loc=mean, scale=stdev)
offsets = np.linspace(mean - 3 * stdev, mean + 3 * stdev, nPoints)
probability = 2 * (normDist.cdf(
(offsets + epsilon - mean) / stdev) - normDist.cdf(
(offsets - epsilon - mean) / stdev))
fig = pPlotUtil.figure()
plt.plot(offsets,probability,'r-',
label="mu = {:.1f}, sigma = {:.1f}, epsilon = {:.2f}".\
format(mean,stdev,epsilon))
plt.xlabel("offset for CDF, c0")
plt.ylabel("Probability (arbitrary units) to land within epsilon of c0")
if (generatePNGs):
X,Y,c1,c2 = getAllStages(fileDict,condition,trial,mWorking,
condNum,trialNum)
saveAsSubplot(X,Y,c1,c2,allStageDir,vizFileFormat)
# format the ffmpeg arguments as we want them
# POST: all videos saved for this trial. make the movie
generateMovie(allStageDir,condition,trialNum,vizFileFormat)
def saveConditions(condition,condNum,conditionKeys,workDir,outDir):
for j,trial in enumerate(conditionKeys):
saveSingleTrial(workDir,outDir,condition,condNum,trial,j)
if __name__ == '__main__':
inDir,workDir,outDir = parseCmdLine()
# next two must match, for the automatic video encoding to work
gUtil.ensureDirExists(outDir)
gUtil.ensureDirExists(workDir)
# get all the files. returns a dictionary of dictionaries.
# each key in the outer (first) dictionary is a condition
# each key in the innter (second) dictionary is a trial for that condition
fileDict = getCheckpointFileDict(inDir)
# loop through each condition and trial
conditionArr = fileDict.keys()
processes= []
for i,condition in enumerate(conditionArr):
print("Forking off a process for condition {:s}".format(condition))
func = saveConditions
funcArgs = (condition,i,fileDict[condition].keys(),workDir,outDir)
p = (Process(target=func, args=funcArgs))
processes.append(p)
p.start()
