def main():
useSimpleShapes = True
if useSimpleShapes:
labels = None
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_500.pkl.gz')
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_5000.pkl.gz')
data = loadFromPklGz(
'../data/simple3DShapes/poisson_train_50000.pkl.gz')
data = data[:25000, :]
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_50000.pkl.gz')
else:
#labels, data = loadFromPklGz('../data/endlessforms/train_bool_50_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_bool_50000_0.pkl.gz')
labels, data = loadFromPklGz(
'../data/endlessforms/train_real_5000_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_real_50000_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_real_50_0.pkl.gz')
size = (10, 10, 20)
FAST_HACK = False
if FAST_HACK:
data = data[:, :400]
size = (10, 10, 4)
def main():
resman.start('junk', diary = False)
#l1tica = loadFromPklGz('results/130402_033310_44cc757_master_psearchTica_UP/00022_data/tica.pkl.gz')
l1tica = loadFromPklGz('results/130406_184751_3d90386_rapidhacks_upson3_1c_l1/tica.pkl.gz') # 1c Upson3
#layer1Whitener = loadFromPklGz('../data/upson_rovio_2/white/train_10_50000_1c.whitener.pkl.gz')
layer1Whitener = loadFromPklGz('../data/upson_rovio_3/white/train_10_50000_1c.whitener.pkl.gz')
layerSizePlan = [10, 15, 23, 35, 53, 80, 120, 180]
stackedTica = StackedTICA(l1tica, layer1Whitener, '../data/upson_rovio_3/imgfiles/',
layerSizePlan = layerSizePlan,
isColor = False,
saveDir = resman.rundir)
if False:
print 'JUST DEBUG...'
pdb.set_trace()
tica = stackedTica.ticas[0]
pdb.set_trace()
return
#data,labels,strings = loadUpsonData3('../data/upson_rovio_3/train_10_50000_1c.pkl.gz')
#data = loadFromPklGz('../data/upson_rovio_3/white/train_10_50000_1c.white.pkl.gz')
#stackedTica.plotResults(resman.rundir, stackedTica.ticas[0], data, (10,10), (15,15))
#pdb.set_trace()
params = {}
params['hiddenISize'] = 15
params['hiddenJSize'] = params['hiddenISize']
params['neighborhoodSize'] = 1.0
params['lambd'] = .026
params['randSeed'] = 0
#params['dataWidth'] = 10
#params['nColors'] = 1
#params['isColor'] = (params['nColors'] == 3)
#params['imgShape'] = ((params['dataWidth'], params['dataWidth'], 3)
# if params['isColor'] else
# (params['dataWidth'], params['dataWidth']))
params['maxFuncCalls'] = 3
#params['whiten'] = True # Just false for Space Invaders dataset...
params['dataCrop'] = None # Set to None to not crop data...
#params['dataCrop'] = 10000 # Set to None to not crop data...
stackedTica.learnNextLayer(params)
#print 'HACK FOR DEBUG'
saveToFile(os.path.join(resman.rundir, 'stackedTica.pkl.gz'), stackedTica) # save learned model
resman.stop()
def testTica():
data = loadFromPklGz('../data/rica_hyv_patches_16.pkl.gz')
data = data[:25, :500] # take just a small slice of data
#pdb.set_trace()
random.seed(0)
tica = TICA(imgShape=(5, 5),
hiddenLayerShape=(4, 5),
shrink=0,
lambd=.05,
epsilon=1e-5)
normData = True
if normData:
# Project each patch to the unit ball
patchNorms = sqrt(sum(data**2, 0) + (1e-8))
data = data / patchNorms
# Initialize weights WW
WW = random.randn(4 * 5, 5 * 5)
WW = (WW.T / sqrt(sum(WW**2, 1))).T
#loadedWW = loadtxt('../octave-randTheta')
#WW = loadedWW
WW = WW.flatten()
cost, sGrad = tica.cost(WW, data)
print 'Current cost is:', cost
print 'Gradient (symbolic) is:', sGrad
nGradFn = Gradient(lambda w: tica.cost(w, data)[0])
nGrad = nGradFn(WW)
print 'Gradient (finite diff) is:', nGrad
print 'diff is', sGrad - nGrad
pdb.set_trace()
def main():
parser = argparse.ArgumentParser(description='Visualizes a trained StackedLayers model.')
parser.add_argument('--name', type = str, default = 'junk',
help = 'Name for GitResultsManager results directory (default: junk)')
#parser.add_argument('--quick', action='store_true', help = 'Enable quick mode (default: off)')
parser.add_argument('--nodiary', action='store_true', help = 'Disable diary (default: diary is on)')
parser.add_argument('stackedLayersFilename', type = str,
help = 'File where a StackedLayers model was stored, something like stackedLayers.pkl.gz')
parser.add_argument('command', type = str, default = 'embed', choices = ['visall', 'embed'], nargs='?',
help = 'What to do: one of {visall (save all plots), embed (drop into shell)}. Default: embed.')
args = parser.parse_args()
resman.start(args.name, diary = not args.nodiary)
saveDir = resman.rundir
print 'Loading StackedLayers from %s' % args.stackedLayersFilename
sl = loadFromPklGz(args.stackedLayersFilename)
print 'Loaded these StackedLayers:'
sl.printStatus()
if args.command == 'embed':
embed()
elif args.command == 'visall':
sl.visAll(saveDir)
else:
print 'Unknown command:', args.command
resman.stop()
def main():
resman.start('junk', diary = False)
stica = loadFromPklGz('results/130407_132841_76b6586_rapidhacks_upson3_1c_l2_first/stackedTica_mod.pkl.gz')
layer1Whitener = loadFromPklGz('../data/upson_rovio_3/white/train_10_50000_1c.whitener.pkl.gz')
layerSizePlan = [10, 15, 23, 35, 53, 80, 120, 180]
visLayer = 1
largeSampleMatrix, labelMatrix, labelStrings = cached(randomSampleMatrixWithLabels, trainFilter,
seed = 0, color = False,
Nw = layerSizePlan[visLayer], Nsamples = 50000)
seed = 0
Nw = layerSizePlan[visLayer-1] # e.g. 10
Nwbig = layerSizePlan[visLayer] # e.g. 15
Nwshift = Nwbig - Nw # e.g. 15 - 10 = 5
Nsamples = 1000
temp = getBigAndSmallerSamples(trainFilter, layer1Whitener, seed, False, Nw, Nwshift, Nsamples)
largeSampleMatrix, stackedSmall, stackedSmallWhite, labelMatrix, labelStrings = temp
pooled = stica.getRepresentation(largeSampleMatrix)
plotTopActivations(pooled, largeSampleMatrix, (Nwbig,Nwbig), resman.rundir, nActivations = 50, nSamples = 20)
pl = (pooled.T - pooled.mean(1)).T
for ii in range(len(labelStrings)):
print 'finding top for', labelStrings[ii]
if labelMatrix[ii,:].sum() == 0:
print ' skipping, no examples'
continue
avgActivationForClass = (pl * labelMatrix[ii,:]).mean(1)
sortIdx = argsort(avgActivationForClass)
topNeurons = sortIdx[-1:-(50+1):-1]
plotTopActivations(pooled[topNeurons,:], largeSampleMatrix, (Nwbig,Nwbig), resman.rundir,
nActivations = 50, nSamples = 20, prefix = 'topfor_%s' % labelStrings[ii])
resman.stop()
def main():
useSimpleShapes = True
if useSimpleShapes:
labels = None
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_500.pkl.gz')
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_5000.pkl.gz')
data = loadFromPklGz('../data/simple3DShapes/poisson_train_50000.pkl.gz'); data = data[:25000,:]
#data = loadFromPklGz('../data/simple3DShapes/poisson_train_50000.pkl.gz')
else:
#labels, data = loadFromPklGz('../data/endlessforms/train_bool_50_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_bool_50000_0.pkl.gz')
labels, data = loadFromPklGz('../data/endlessforms/train_real_5000_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_real_50000_0.pkl.gz')
#labels, data = loadFromPklGz('../data/endlessforms/train_real_50_0.pkl.gz')
size = (10,10,20)
FAST_HACK = False
if FAST_HACK:
data = data[:,:400]
size = (10,10,4)
def main():
parser = argparse.ArgumentParser(
description='Visualizes a trained StackedLayers model.')
parser.add_argument(
'--name',
type=str,
default='junk',
help='Name for GitResultsManager results directory (default: junk)')
#parser.add_argument('--quick', action='store_true', help = 'Enable quick mode (default: off)')
parser.add_argument('--nodiary',
action='store_true',
help='Disable diary (default: diary is on)')
parser.add_argument(
'stackedLayersFilename',
type=str,
help=
'File where a StackedLayers model was stored, something like stackedLayers.pkl.gz'
)
parser.add_argument(
'command',
type=str,
default='embed',
choices=['visall', 'embed'],
nargs='?',
help=
'What to do: one of {visall (save all plots), embed (drop into shell)}. Default: embed.'
)
args = parser.parse_args()
resman.start(args.name, diary=not args.nodiary)
saveDir = resman.rundir
print 'Loading StackedLayers from %s' % args.stackedLayersFilename
sl = loadFromPklGz(args.stackedLayersFilename)
print 'Loaded these StackedLayers:'
sl.printStatus()
if args.command == 'embed':
embed()
elif args.command == 'visall':
sl.visAll(saveDir)
else:
print 'Unknown command:', args.command
resman.stop()
def ricaRun(randSeed, resman):
testfn('rica')
data = loadFromPklGz('../data/atari/mspacmantrain_15_50000_3c.pkl.gz')
data = data.T # Make into one example per column
print 'randSeed is', randSeed
random.seed(randSeed)
print 'making dirs'
thisdir = os.path.join(resman.rundir, 'seed-%03d' % randSeed)
os.makedirs(thisdir)
rica = RICA(imgShape = (15, 15, 3),
nFeatures = 400,
lambd = .05,
epsilon = 1e-5,
saveDir = thisdir)
rica.run(data, maxFun = 3, whiten = True)
return 'finished return value'
def main():
parser = argparse.ArgumentParser(description='Makes features for the human activity deteciton dataset. Example usage:\n./humanObjectFeatures.py /path/to/data_obj_feats.txt')
parser.add_argument('--name', type = str, default = 'junk',
help = 'Name for GitResultsManager results directory (default: junk)')
parser.add_argument('--quick', action='store_true', help = 'Enable quick mode (default: off)')
parser.add_argument('--nodiary', action='store_true', help = 'Disable diary (default: diary is on)')
#parser.add_argument('stackedLayersFilename', type = str,
# help = 'File where a StackedLayers model was stored, something like stackedLayers.pkl.gz')
#parser.add_argument('command', type = str, default = 'embed', choices = ['visall', 'embed'], nargs='?',
# help = 'What to do: one of {visall (save all plots), embed (drop into shell)}. Default: embed.')
parser.add_argument('--stackedlayers', type = str, default = '',
help = 'Path to a *.pkl.gz file containing a pickeled StackedLayers object to load (default: None)')
parser.add_argument('--fdata', type = str, default = '',
help = 'Path to a "..../formatted/data/method_name" directory containing files like 0510175829_1.txt. Needed only if adding truth labels. (default: None)')
parser.add_argument('--outfile', type = str, default = 'data_objs_feats_plus.txt',
help = 'What to name the output file (default: data_objs_feats_plus.txt)')
parser.add_argument('dataDir', type = str, default = 'data',
help = 'Where to look for the "by_id" directory')
parser.add_argument('data_obj_feats_file', type = str,
help = 'Which data_objs_feats.txt file to load')
args = parser.parse_args()
resman.start(args.name, diary = not args.nodiary)
saveDir = resman.rundir
#print 'Loading StackedLayers from %s' % args.stackedLayersFilename
#sl = loadFromPklGz(args.stackedLayersFilename)
#print 'Loaded these StackedLayers:'
#sl.printStatus()
stackedLayers = None
if args.stackedlayers:
stackedLayers = loadFromPklGz(args.stackedlayers)
makeFeats(dataDir = args.dataDir, featsFilename = args.data_obj_feats_file,
saveDir = saveDir, outputFilename = args.outfile, quick = args.quick,
stackedLayers = stackedLayers, formattedDataDir = args.fdata,
featsBase = True, featsConst = 0, featsRand = 0, featsSL = False, featsTruth = False)
resman.stop()
def ricaRun(randSeed, resman):
testfn('rica')
data = loadFromPklGz('../data/atari/mspacmantrain_15_50000_3c.pkl.gz')
data = data.T # Make into one example per column
print 'randSeed is', randSeed
random.seed(randSeed)
print 'making dirs'
thisdir = os.path.join(resman.rundir, 'seed-%03d' % randSeed)
os.makedirs(thisdir)
rica = RICA(imgShape=(15, 15, 3),
nFeatures=400,
lambd=.05,
epsilon=1e-5,
saveDir=thisdir)
rica.run(data, maxFun=3, whiten=True)
return 'finished return value'
def main():
parser = argparse.ArgumentParser(description='Trains a StackedLayers model.')
parser.add_argument('layerFilename', type = str,
help = 'File defining layers, something like tica-10-15.layers')
parser.add_argument('trainParamsFilename', type = str,
help = 'File defining training parameters, something like tica-10-15.trainparams')
parser.add_argument('--name', type = str, default = 'junk',
help = 'Name for GitResultsManager results directory (default: junk)')
parser.add_argument('--load', type = str, default = '',
help = ('Load a previously created StackedLayers object. This can ' +
'be used to resume training a previously checkpointed, ' +
'partially trained StackedLayers object (default: none)'))
parser.add_argument('--quick', action='store_true', help = 'Enable quick mode (default: off)')
parser.add_argument('--nodiary', action='store_true', help = 'Disable diary (default: diary is on)')
args = parser.parse_args()
resman.start(args.name, diary = not args.nodiary)
saveDir = resman.rundir
layerDefinitions = importFromFile(args.layerFilename, 'layers')
trainParams = importFromFile(args.trainParamsFilename, 'trainParams')
shutil.copyfile(args.layerFilename, os.path.join(saveDir, 'params.layers'))
shutil.copyfile(args.trainParamsFilename, os.path.join(saveDir, 'params.trainparams'))
if args.load:
print 'Loading StackedLayers object from %s' % args.load
sl = loadFromPklGz(args.load)
else:
print 'Creating new StackedLayers object'
sl = StackedLayers(layerDefinitions)
sl.printStatus()
sl.train(trainParams, saveDir = saveDir, quick = args.quick)
resman.stop()
'''
import pdb
import os, sys, time
from numpy import *
from rica import RICA
from GitResultsManager import resman
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary = False)
Nw = 4
dataXX, dataYY = loadFromPklGz('../data/simpleShapes/train_%02d_50.pkl.gz' % Nw)
#dataXX, dataYY = loadFromPklGz('../data/simpleShapes/train_%02d_50000.pkl.gz' % Nw)
data = dataXX.T # Make into one example per column
random.seed(0)
rica = RICA(imgShape = (Nw, Nw),
nFeatures = 400,
lambd = .05,
epsilon = 1e-5,
saveDir = resman.rundir)
rica.run(data, maxFun = 300)
#rica.run(data, maxFun = 300, whiten = True)
resman.stop()
from PIL import Image
from scipy.optimize.lbfgsb import fmin_l_bfgs_b
from tica import TICA
from GitResultsManager import resman, fmtSeconds
from util.plotting import tile_raster_images
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary = False)
print 'probably want simple instead... finish this file first'
sys.exit(1)
data = loadFromPklGz('../data/atari/mspacman_train_15_50000_3c.pkl.gz')
data = data.T # Make into one example per column
#data = data[:,:5000] # HACK!!!!!!!!!
#hiddenISize = 20
#hiddenJSize = 20
#lambd = .05
#neighborhoodSize = 1
#print '\nChosen TICA parameters'
hiddenISize = random.randint(4, 25+1)
hiddenJSize = random.randint(10, 30+1)
lambd = .1 * 2 ** random.randint(-5, 5+1)
neighborhoodSize = random.uniform(.1,3)
print '\nRandomly selected TICA parameters'
self.WW = self.runOptimization(data, maxFun, plotEvery)
#if self.saveDir:
# saveToFile(os.path.join(self.saveDir, 'WW+pca.pkl.gz'), (WW, self.pca))
# Make and save some plots
self.plotCostLog()
#if self.doPlots:
# self.plotWW(WW)
#self.plotActivations(WW, data)
#if self.doPlots:
# self.plotReconstructions(WW, data)
if __name__ == '__main__':
resman.start('junk', diary=False)
data = loadFromPklGz('../data/rica_hyv_patches_16.pkl.gz')
#data = data[:,:5000] #HACK
random.seed(0)
rica = RICA(imgShape=(16, 16),
nOutputs=50,
lambd=.05,
epsilon=1e-5,
float32=False,
saveDir=resman.rundir)
rica.learn(data, plotEvery=None, maxFun=5, whiten=True)
resman.stop()
import os, sys, time
from numpy import *
from PIL import Image
from tica import TICA
from GitResultsManager import resman
from util.plotting import tile_raster_images
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary = False)
#data = loadFromPklGz('../data/atari/mspacman_train_15_50000_3c.pkl.gz')
data = loadFromPklGz('../data/atari/space_invaders_train_15_50000_3c.pkl.gz')
data = data.T # Make into one example per column
#data = data[:,:5000] # HACK!!!!!!!!!
hiddenISize = 8
hiddenJSize = 8
lambd = .05
neighborhoodSize = 1.5
print '\nChosen TICA parameters'
#hiddenISize = random.randint(4, 25+1)
#hiddenJSize = random.randint(10, 30+1)
#lambd = .1 * 2 ** random.randint(-5, 5+1)
#neighborhoodSize = random.uniform(.1,3)
#print '\nRandomly selected TICA parameters'
import os, sys, time
from numpy import *
from PIL import Image
from scipy.optimize.lbfgsb import fmin_l_bfgs_b
from tica import TICA
from GitResultsManager import resman, fmtSeconds
from util.plotting import tile_raster_images
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary=False)
print 'probably want simple instead... finish this file first'
sys.exit(1)
data = loadFromPklGz('../data/atari/mspacman_train_15_50000_3c.pkl.gz')
data = data.T # Make into one example per column
#data = data[:,:5000] # HACK!!!!!!!!!
#hiddenISize = 20
#hiddenJSize = 20
#lambd = .05
#neighborhoodSize = 1
#print '\nChosen TICA parameters'
hiddenISize = random.randint(4, 25 + 1)
hiddenJSize = random.randint(10, 30 + 1)
lambd = .1 * 2**random.randint(-5, 5 + 1)
neighborhoodSize = random.uniform(.1, 3)
print '\nRandomly selected TICA parameters'
def main():
parser = argparse.ArgumentParser(
description=
'Makes features for the human activity deteciton dataset. Example usage:\n./humanObjectFeatures.py /path/to/data_obj_feats.txt'
)
parser.add_argument(
'--name',
type=str,
default='junk',
help='Name for GitResultsManager results directory (default: junk)')
parser.add_argument('--quick',
action='store_true',
help='Enable quick mode (default: off)')
parser.add_argument('--nodiary',
action='store_true',
help='Disable diary (default: diary is on)')
#parser.add_argument('stackedLayersFilename', type = str,
# help = 'File where a StackedLayers model was stored, something like stackedLayers.pkl.gz')
#parser.add_argument('command', type = str, default = 'embed', choices = ['visall', 'embed'], nargs='?',
# help = 'What to do: one of {visall (save all plots), embed (drop into shell)}. Default: embed.')
parser.add_argument(
'--stackedlayers',
type=str,
default='',
help=
'Path to a *.pkl.gz file containing a pickeled StackedLayers object to load (default: None)'
)
parser.add_argument(
'--fdata',
type=str,
default='',
help=
'Path to a "..../formatted/data/method_name" directory containing files like 0510175829_1.txt. Needed only if adding truth labels. (default: None)'
)
parser.add_argument(
'--outfile',
type=str,
default='data_objs_feats_plus.txt',
help='What to name the output file (default: data_objs_feats_plus.txt)'
)
parser.add_argument('dataDir',
type=str,
default='data',
help='Where to look for the "by_id" directory')
parser.add_argument('data_obj_feats_file',
type=str,
help='Which data_objs_feats.txt file to load')
args = parser.parse_args()
resman.start(args.name, diary=not args.nodiary)
saveDir = resman.rundir
#print 'Loading StackedLayers from %s' % args.stackedLayersFilename
#sl = loadFromPklGz(args.stackedLayersFilename)
#print 'Loaded these StackedLayers:'
#sl.printStatus()
stackedLayers = None
if args.stackedlayers:
stackedLayers = loadFromPklGz(args.stackedlayers)
makeFeats(dataDir=args.dataDir,
featsFilename=args.data_obj_feats_file,
saveDir=saveDir,
outputFilename=args.outfile,
quick=args.quick,
stackedLayers=stackedLayers,
formattedDataDir=args.fdata,
featsBase=True,
featsConst=0,
featsRand=0,
featsSL=False,
featsTruth=False)
resman.stop()
Jason Yosinski
'''
import pdb
import os, sys, time
from numpy import *
from rica import RICA
from GitResultsManager import resman
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary=False)
Nw = 4
dataXX, dataYY = loadFromPklGz(
'../data/simpleShapes/train_%02d_50.pkl.gz' % Nw)
#dataXX, dataYY = loadFromPklGz('../data/simpleShapes/train_%02d_50000.pkl.gz' % Nw)
data = dataXX.T # Make into one example per column
random.seed(0)
rica = RICA(imgShape=(Nw, Nw),
nFeatures=400,
lambd=.05,
epsilon=1e-5,
saveDir=resman.rundir)
rica.run(data, maxFun=300)
#rica.run(data, maxFun = 300, whiten = True)
resman.stop()
import os, sys, time
from numpy import *
from PIL import Image
from scipy.optimize.lbfgsb import fmin_l_bfgs_b
from rica import RICA
from GitResultsManager import resman, fmtSeconds
from util.plotting import tile_raster_images
from util.dataLoaders import loadFromPklGz, saveToFile
if __name__ == '__main__':
resman.start('junk', diary = False)
data = loadFromPklGz('../data/upson_rovio_2/train_10_50000_1c.pkl.gz')
data = data.T # Make into one example per column
#data = data[:,:5000] # HACK!!!!!!!!!
nFeatures = 100
lambd = .05
neighborhoodSize = 1.5
print '\nChosen TICA parameters'
for key in ['nFeatures', 'lambd']:
print ' %20s: %s' % (key, locals()[key])
random.seed(0)
rica = RICA(imgShape = (10, 10),
nFeatures = nFeatures,
lambd = lambd,
#if self.saveDir:
# saveToFile(os.path.join(self.saveDir, 'WW+pca.pkl.gz'), (WW, self.pca))
# Make and save some plots
self.plotCostLog()
#if self.doPlots:
# self.plotWW(WW)
#self.plotActivations(WW, data)
#if self.doPlots:
# self.plotReconstructions(WW, data)
if __name__ == '__main__':
resman.start('junk', diary = False)
data = loadFromPklGz('../data/rica_hyv_patches_16.pkl.gz')
#data = data[:,:5000] #HACK
random.seed(0)
rica = RICA(imgShape = (16, 16),
nOutputs = 50,
lambd = .05,
epsilon = 1e-5,
float32 = False,
saveDir = resman.rundir)
rica.learn(data, plotEvery = None, maxFun = 5, whiten = True)
resman.stop()