mmdNetLayerSizes = [25, 25]
l2_penalty = 1e-2
#init = lambda shape, name:initializations.normal(shape, scale=.1e-4, name=name)
#def my_init (shape):
# return initializers.normal(stddev=.1e-4)
#my_init = 'glorot_normal'
#######################
###### read data ######
#######################
# we load two CyTOF samples
data = 'person1_3month'
if data == 'person1_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),
'Data/Person1Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),
'Data/Person1Day2_baseline.csv')
if data == 'person2_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),
'Data/Person2Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),
'Data/Person2Day2_baseline.csv')
if data == 'person1_3month':
sourcePath = os.path.join(io.DeepLearningRoot(),
'Data/Person1Day1_3month.csv')
targetPath = os.path.join(io.DeepLearningRoot(),
'Data/Person1Day2_3month.csv')
if data == 'person2_3month':
sourcePath = os.path.join(io.DeepLearningRoot(),
from keras import initializations from keras.layers.normalization import BatchNormalization from keras.layers import Input, Dense, merge, Activation from keras.regularizers import l2 from keras.models import Model # configuration hyper parameters denoise = True # whether or not to train a denoising autoencoder to remover the zeros ###################### ###### get data ###### ###################### # we load two CyTOF samples p1d1Path = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_baseline.csv') p1d2Path = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_baseline.csv') p2d1Path = os.path.join(io.DeepLearningRoot(),'Data/Person2Day1_baseline.csv') p2d2Path = os.path.join(io.DeepLearningRoot(),'Data/Person2Day2_baseline.csv') p1Dae = load_model(os.path.join(io.DeepLearningRoot(),'savedModels/person1_baseline_DAE.h5')) p2Dae = load_model(os.path.join(io.DeepLearningRoot(),'savedModels/person2_baseline_DAE.h5')) p1d1 = genfromtxt(p1d1Path, delimiter=',', skip_header=0) p1d2 = genfromtxt(p1d2Path, delimiter=',', skip_header=0) p2d1 = genfromtxt(p2d1Path, delimiter=',', skip_header=0) p2d2 = genfromtxt(p2d2Path, delimiter=',', skip_header=0)
l2_penalty_ae = 1e-2
#MMD net configuration
mmdNetLayerSizes = [25, 25]
l2_penalty = 1e-2
#init = lambda shape, name:initializations.normal(shape, scale=.1e-4, name=name)
#def my_init (shape):
# return initializers.normal(stddev=.1e-4)
#my_init = 'glorot_normal'
#######################
###### read data ######
#######################
# we load two CyTOF samples
sourcePath = os.path.join(io.DeepLearningRoot(), 'Data/' + args.files[0])
targetPath = os.path.join(io.DeepLearningRoot(), 'Data/' + args.files[1])
'''
data = 'person2_baseline'
if data =='person1_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_baseline.csv')
if data =='person2_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day2_baseline.csv')
if data =='person1_3month':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_3month.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_3month.csv')
if data =='person2_3month':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day1_3month.csv')
from Calibration_Util import DataHandler as dh
import argparse
#detect display
havedisplay = "DISPLAY" in os.environ
# havedisplay = False
#if we have a display use a plotting backend
if havedisplay:
matplotlib.use('TkAgg')
else:
matplotlib.use('Agg')
parse = argparse.ArgumentParser(
description='command line interface for mmd net')
parse.add_argument('--source_path',
type=str,
default=os.path.join(io.DeepLearningRoot(),
'Data/Person1Day1_3month.csv'),
help='Path to the source dataset')
parse.add_argument('--target_path',
type=str,
default=os.path.join(io.DeepLearningRoot(),
'Data/Person1Day2_3month.csv'),
help='Path to the source dataset')
parse.add_argument('--epochs',
type=int,
default=500,
help='Number of epochs to run for')
parse.add_argument('--denoise',
type=bool,
default=False,
help='Whether or not to denoise the datasets')
from keras.models import Model
# configuration hyper parameters
denoise = True # whether or not to use a denoising autoencoder to remove the zeros
######################
###### get data ######
######################
# we load two CyTOF samples
data = 'person2_3month'
if data =='person1_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_baseline.csv')
sourceLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_baseline_label.csv')
targetLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_baseline_label.csv')
autoencoder = load_model(os.path.join(io.DeepLearningRoot(),'savedModels/person1_baseline_DAE.h5'))
if data =='person2_baseline':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day1_baseline.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day2_baseline.csv')
sourceLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day1_baseline_label.csv')
targetLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person2Day2_baseline_label.csv')
autoencoder = load_model(os.path.join(io.DeepLearningRoot(),'savedModels/person2_baseline_DAE.h5'))
if data =='person1_3month':
sourcePath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_3month.csv')
targetPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_3month.csv')
sourceLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day1_3month_label.csv')
targetLabelPath = os.path.join(io.DeepLearningRoot(),'Data/Person1Day2_3month_label.csv')