Initial commit (6930a164) · Commits · Publications / Relating plasticity to dislocation properties by data analysis - scaling vs machine learning approaches

.gitignore

0 → 100644

+3 −0

Original line number	Diff line number	Diff line
		*csv
		*txt
		models/

DisDataExp.dat

0 → 100644

+140 −0

Original line number	Diff line number	Diff line
		1.7E-5 20500000000 0.635 0.4
		1.7E-5 29400000000 1.75 0.7
		1.7E-5 40000000000 1.73 0.7
		1.7E-5 72000000000 0.981 1.1
		1.7E-5 114000000000 1.32 1.5
		1.7E-5 122000000000 1.56 1.2
		1.7E-5 75600000000 3.23 2
		1.7E-5 75200000000 3.59 2.6
		1.7E-5 119000000000 3.3 2.2
		1.7E-5 136000000000 2.87 3.3
		1.7E-5 199000000000 2.7 4.3
		1.7E-5 203000000000 5.29 5.1
		1.7E-5 224000000000 4.59 5.2
		1.7E-5 313000000000 4.4 6.2
		1.7E-5 356000000000 5.16 6.3
		1.7E-5 480000000000 7.44 6.1
		1.7E-5 613000000000 4.82 6.9
		1.7E-5 1540000000000 7.63 8.2
		6.3E-5 956000 0.999 0
		6.3E-5 8950000 0.545 0
		6.3E-5 15800000 0.72 0
		6.3E-5 16100000 0.545 0
		6.3E-5 28000000 0.519 0
		6.3E-5 101000000 0.644 0
		6.3E-5 1820000000 1.18 0
		6.3E-5 2290000000 1.31 0
		6.3E-5 2110000000 1.39 0
		6.3E-5 3080000000 1.5 0
		6.3E-5 5560000000 1.62 0
		6.3E-5 18500000000 1.97 0
		6.3E-5 1410000000 0.649 0
		6.3E-5 1590000000 0.595 0
		6.3E-5 1920000000 0.616 0
		6.3E-5 2030000000 0.719 0
		6.3E-5 2670000000 0.743 0
		6.3E-5 3130000000 0.74 0
		6.3E-5 4130000000 0.722 0
		6.3E-5 5270000000 0.831 0
		6.3E-5 6730000000 0.849 0
		6.3E-5 8430000000 0.936 0
		6.3E-5 17800000000 0.897 0
		6.3E-5 14200000000 1.02 0
		6.3E-5 76400000000 1.32 0
		6.3E-5 79200000000 1.38 0
		1E-4 19600000000 1.02 0.5
		1E-4 33200000000 1.03 0.7
		1E-4 59200000000 1.58 1.5
		1E-4 88200000000 2.51 0.0871982863335726
		1E-4 287000000000 4.52 0.223486601498623
		1E-4 391000000000 5.54 0.309489286113777
		1E-4 456000000000 3.86 0.173607979688386
		1E-4 492000000000 5.54 0.309489286113777
		1E-4 1220000000000 7.92 0.548264701837827
		1E-4 3960000000000 15.4 1.58877774954313
		1E-4 6250000000000 19.5 2.31785024944096
		1E-4 8420000000000 26.1 3.69534344070533
		1E-4 12400000000000 28.4 4.229989990906
		1E-4 14100000000000 27.8 4.08791323224191
		1E-4 28700000000000 31.7 5.04341627088995
		1E-4 34900000000000 40 7.31688083083722
		1E-4 221000000000000 113 38.5438866283516
		1E-4 100000000 0.66 0
		1E-4 100000000 0.37 0
		1E-4 200000000 0.37 0
		1E-4 200000000 0.35 0
		1E-4 1500000000 0.34 0
		1E-4 100000000 0.38 0
		1E-4 2000000000 0.37 0
		1E-4 10000000000 0.94 0
		6E-4 157000000 59 0
		6E-4 157000000 50.3 0
		6E-4 1420000000 38.5 0
		6E-4 3960000000 28.8 0
		6E-4 9340000000 20.7 0
		6E-4 17800000000 16.7 0
		6E-4 43300000000 16.1 0
		400 22000000000 9.73 0.5
		400 47000000000 8.78 1
		400 120000000000 7.51 1.5
		400 140000000000 8.1 1.8
		400 150000000000 8.03 1.8
		800 22000000000 12.1 0.5
		800 47000000000 10 1
		800 120000000000 9.1 1.5
		800 140000000000 9.32 1.8
		4.37E-5 47000000000 1.63 0.5
		1.76E-4 47000000000 1.73 0.5
		687 47000000000 10.4 0.5
		4.28E-5 140000000000 3.19 1.5
		1.62E-4 140000000000 3.33 1.5
		414 140000000000 9.69 1.5
		719 140000000000 12.3 1.5
		4.29E-5 4160000000000 2.12 0.7
		5.47E-4 4160000000000 2.38 0.7
		509 4160000000000 2.93 0.7
		799 4160000000000 3.21 0.7
		960 4160000000000 3.5 0.7
		1160 4160000000000 4.05 0.7
		1400 4160000000000 4.44 0.7
		850 4960000000000 2.85 0.681818181818182
		850 5770000000000 4.1 1.36363636363636
		850 8290000000000 6.23 2.72727272727273
		850 11900000000000 7.82 4.54545454545455
		850 17800000000000 8.88 6.36363636363636
		4.55E-4 1880000000000 10 2
		0.0439 1880000000000 8.53 2
		2120 1880000000000 13.3 2
		5830 1880000000000 26.1 2
		8970 1880000000000 39 2
		3830 6200000000000 6.94 0.4
		4160 6200000000000 7.78 0.4
		12400 6200000000000 17.9 0.4
		18800 6200000000000 23.6 0.4
		18200 6200000000000 23.9 0.4
		26700 6200000000000 32.7 0.4
		3830 6200000000000 6.94 0.4
		4.09E-4 6050000000000 6.62 0.1
		0.0042 6050000000000 6.28 0.1
		0.0441 6050000000000 6.62 0.1
		1.99 6050000000000 7.25 0.1
		10.1 6050000000000 7.88 0.1
		13.1 6050000000000 7.59 0.1
		79.5 6050000000000 7.88 0.1
		315 6050000000000 13.9 0.1
		379 6050000000000 11.4 0.1
		1920 6050000000000 23.6 0.1
		1920 6050000000000 24 0.1
		3450 6050000000000 55.1 0.1
		4490 6050000000000 57.9 0.1
		4870 6050000000000 59.1 0.1
		7010 6050000000000 73.6 0.1
		5830 6050000000000 74.2 0.1
		7010 6050000000000 74.9 0.1
		7300 6050000000000 89.3 0.1
		1000000 1500000000000 1290 0
		100 6680000000000 15.4 0.1
		1000 6680000000000 17.3 0.1
		10000 6680000000000 23.7 0.1
		100000 6680000000000 75.5 0.1
		1000000 6680000000000 274 0.1

DisDataSim.dat

0 → 100644

+194 −0

Original line number	Diff line number	Diff line
		0.1 22600000 1.38 1.01E-5
		0.1 173000000 0.381 1E-5
		0.1 583000000 0.281 1E-5
		0.1 1980000000 0.305 1E-5
		0.1 14700000000 1.08 1E-5
		0.1 110000000000 1.19 1E-5
		1 142000000 3.86 1E-4
		1 1100000000 0.931 1E-4
		1 3030000000 0.747 1E-4
		1 13900000000 0.882 1E-4
		1 107000000000 1.95 1E-4
		1 719000000000 3.38 1E-4
		1 141000000 3 1.01E-4
		1 1030000000 1.06 8.09E-5
		1 3030000000 0.747 1E-4
		1 13800000000 0.747 1E-4
		1 108000000000 1.62 1E-4
		1 141000000 3.86 1E-4
		1 1550000000 0.943 1E-4
		1 2030000000 0.588 1E-4
		1 14500000000 1.11 1E-4
		1 104000000000 1.51 1E-4
		1 733000000000 1.92 1E-4
		10 141000000 36.8 5E-4
		10 566000000 11.6 5E-4
		10 1960000000 3.41 5.01E-4
		10 122000000000 3.28 5E-4
		10 661000000000 2.82 5E-4
		10 4110000000000 6.98 5E-4
		10 141000000 29.1 5.01E-4
		10 566000000 11.5 5.01E-4
		10 1960000000 3.38 5.01E-4
		10 25400000000 2.09 3.78E-4
		10 106000000000 1.64 5E-4
		10 763000000000 1.85 5E-4
		10 141000000 36.8 5E-4
		10 566000000 11.6 5E-4
		10 1960000000 3.41 5.01E-4
		10 116000000000 2.18 5E-4
		10 659000000000 1.95 5E-4
		10 3850000000000 4.08 5.02E-4
		100 141000000 657 0.002
		100 566000000 151 0.002
		100 2520000000 36.8 0.002
		100 132000000000 2.45 0.002
		100 676000000000 4.03 0.002
		100 3940000000000 8.37 0.002
		100 25300000000000 9.76 0.00201
		100 141000000 490 0.00202
		100 566000000 151 0.002
		100 2940000000 37.8 0.00189
		100 282000000000 4.53 0.002
		100 646000000000 4.61 0.002
		100 4010000000000 10.3 0.002
		100 25000000000000 11.6 0.00201
		100 141000000 657 0.002
		100 566000000 151 0.002
		100 2210000000 36.7 0.002
		100 173000000000 3.67 0.002
		100 672000000000 4.74 0.002
		100 718000000000 4.9 0.002
		100 3620000000000 8.21 0.002
		100 27000000000000 11.4 0.002
		1000 566000000 1710 0.002
		1000 15200000000 62.9 0.002
		1000 158000000000 8.12 0.002
		1000 1150000000000 7.43 0.002
		1000 4450000000000 12.5 0.002
		1000 25100000000000 20.9 0.002
		1000 124000000000000 43.3 0.002
		1000 566000000 1710 0.002
		1000 4330000000 387 0.002
		1000 17700000000 65.7 0.00156
		1000 193000000000 9.51 0.002
		1000 1040000000000 6.9 0.002
		1000 4470000000000 10.5 0.002
		1000 24200000000000 19.1 0.00201
		1000 101000000000000 45.3 0.002
		1000 566000000 1710 0.002
		1000 7790000000 381 0.002
		1000 39800000000 42.5 0.002
		1000 188000000000 8.53 0.002
		1000 824000000000 8 0.002
		1000 4810000000000 12.3 0.002
		1000 23800000000000 14.8 0.002
		1000 116000000000000 32 0.00201
		1000 566000000 1710 0.002
		1000 2010000000 453 0.002
		1000 2010000000 453 0.002
		1000 14000000000 61.6 0.0018
		1000 175000000000 8.37 0.002
		1000 809000000000 8.53 0.002
		1000 1270000000000 8.74 0.002
		1000 4860000000000 12.2 0.002
		1000 28600000000000 24.9 0.002
		1000 126000000000000 39.2 0.002
		10000 6270000000 1540 0.002
		10000 51000000000 166 0.00201
		10000 75700000000 140 0.00201
		10000 108000000000 102 0.00201
		10000 1500000000000 19.4 0.002
		10000 5400000000000 18 0.002
		10000 27800000000000 25.6 0.002
		10000 117000000000000 33.6 0.002
		10000 711000000000000 84.5 0.00209
		10000 6270000000 1540 0.002
		10000 19600000000 584 0.002
		10000 106000000000 105 0.00201
		10000 611000000000 32 0.002
		10000 1370000000000 20 0.002
		10000 7720000000000 25.4 0.002
		10000 24000000000000 19.9 0.00201
		10000 110000000000000 48.6 0.00201
		10000 712000000000000 90.2 0.00201
		10000 6270000000 1540 0.002
		10000 18300000000 637 0.002
		10000 158000000000 100 0.002
		10000 298000000000 55.1 0.002
		10000 802000000000 21.4 0.002
		10000 4650000000000 16.7 0.002
		10000 25400000000000 21.6 0.002
		10000 132000000000000 43.7 0.002
		10000 746000000000000 71 0.002
		10000 13900000000 768 0.00201
		10000 109000000000 102 0.00204
		10000 232000000000 84.5 0.00194
		10000 553000000000 36.7 0.002
		10000 746000000000 27.1 0.002
		10000 1620000000000 17.4 0.002
		10000 5350000000000 20.1 0.002
		10000 18100000000000 24.5 0.00201
		10000 22900000000000 31.1 0.002
		10000 27400000000000 31.1 0.002
		10000 107000000000000 42.9 0.002
		10000 697000000000000 85.7 0.002
		10000 733000000000000 71.4 0.002
		100000 60100000000 1770 0.00201
		100000 334000000000 878 0.002
		100000 978000000000 134 0.00202
		100000 6490000000000 55.5 0.002
		100000 27900000000000 41.2 0.002
		100000 122000000000000 53.1 0.00204
		100000 745000000000000 86.5 0.00205
		100000 3.76E15 131 0.00204
		100000 5.48E15 110 0.00202
		100000 55400000000 1980 0.002
		100000 109000000000 1010 0.002
		100000 275000000000 425 0.002
		100000 1690000000000 100 0.00201
		100000 12100000000000 54.7 0.002
		100000 28200000000000 49.8 0.00202
		100000 109000000000000 53.5 0.002
		100000 729000000000000 94.3 0.00203
		100000 5.52E15 93.5 0.00201
		100000 55600000000 1650 0.002
		100000 109000000000 959 0.002
		100000 954000000000 147 0.00205
		100000 5420000000000 53.9 0.002
		100000 6720000000000 52.7 0.00201
		100000 6720000000000 52.7 0.00201
		100000 33200000000000 48.6 0.00202
		100000 126000000000000 53.5 0.00204
		100000 726000000000000 111 0.00203
		100000 3.77E15 136 0.00212
		100000 5.39E15 147 0.0019
		1000000 1460000000000 992 0.00201
		1000000 2320000000000 576 0.002
		1000000 11400000000000 234 0.002
		1000000 32700000000000 155 0.002
		1000000 149000000000000 114 0.002
		1000000 774000000000000 127 0.00205
		1000000 3.88E15 164 0.00201
		1000000 5.65E15 96.8 0.00202
		1000000 680000000000 1650 0.00221
		1000000 4120000000000 429 0.00228
		1000000 28100000000000 165 0.00209
		1000000 114000000000000 100 0.0022
		1000000 766000000000000 88.2 0.00213
		1000000 5.47E15 98.8 0.00225
		1000000 1500000000000 935 0.00201
		1000000 1990000000000 645 0.002
		1000000 2250000000000 645 0.002
		1000000 8430000000000 256 0.002
		1000000 43500000000000 160 0.002
		1000000 155000000000000 126 0.00201
		1000000 767000000000000 136 0.00203
		1000000 3.89E15 131 0.00222
		1000000 5.59E15 109 0.00201
		1000000 1.2E16 138 0.00204
		250000000 700000000000000 457 0.002
		250000000 1.26E15 249 0.002
		250000000 2.4E15 234 0.002
		250000000 7.65E15 158 0.002
		250000000 2.35E16 293 0.002

connect-data.py

0 → 100644

+23 −0

Original line number	Diff line number	Diff line
		#!/usr/bin/env python3
		# -- coding: utf-8 --
		"""
		This file is used to postprocess the data by Bapst et al. It relies on several
		conventions used by Bapst to sample and format his data, so do not use this
		to process other data without checking that it behaves as intended.
		@author: Victor Bapst et al.
		modified by Stefan Hiemer
		"""

		import pandas as pd
		from glob import glob

		if __name__=="__main__":

		for f in glob("origin"):
		origin = pd.read_csv(f,sep=',',header=None)
		origin.columns=["data source"]
		data = f.split("_")
		data = "_".join([data[0].split("-")[0],data[1]])
		data = pd.read_csv(data,sep=',',header=0)
		print(data)
		print(pd.concat((data,origin)))

random-subset-selction.py

0 → 100644

+301 −0

Original line number	Diff line number	Diff line
		#!/usr/bin/env python3
		# -- coding: utf-8 --
		"""
		This file is used to postprocess the data by Bapst et al. It relies on several
		conventions used by Bapst to sample and format his data, so do not use this
		to process other data without checking that it behaves as intended.
		@author: Victor Bapst et al.
		modified by Stefan Hiemer
		"""

		import numpy as np

		from sklearn.preprocessing import StandardScaler
		from sklearn.model_selection import train_test_split
		from sklearn.linear_model import ElasticNet
		from sklearn.kernel_ridge import KernelRidge
		from sklearn.metrics import r2_score
		from sklearn.tree import DecisionTreeRegressor

		import tensorflow as tf
		from tensorflow import keras
		from tensorflow.keras.layers import Dense, BatchNormalization
		from tensorflow.keras.optimizers import Adam
		from tensorflow.keras.callbacks import ModelCheckpoint, TensorBoard
		from tensorflow.keras import backend as K

		def coeff_determination(y_true, y_pred):

		SS_res = K.sum(K.square( y_true-y_pred ))
		SS_tot = K.sum(K.square( y_true - K.mean(y_true) ) )
		return ( 1 - SS_res/(SS_tot + K.epsilon()) )

		def build_perceptron(units, activation):
		"""
		Simple multilayer perceptron.

		units: list
		activation: activation function
		"""

		model = keras.Sequential()

		for unit in units:
		model.add(Dense(unit, activation=activation))
		model.add(BatchNormalization())

		model.add(Dense(1))

		minimizer = Adam(learning_rate = 0.001,
		beta_1 = 0.9, beta_2 = 0.999,
		amsgrad = False)
		model.compile(loss = 'mse', optimizer = minimizer,
		metrics = ['mse',coeff_determination])

		return model

		if __name__ == '__main__':

		# load data
		exp = np.loadtxt('DisDataExp.dat', dtype=float, delimiter='\t')
		sim = np.loadtxt('DisDataSim.dat', dtype=float, delimiter='\t')

		# reorder the data as currently it is impractical
		exp = np.concatenate((exp[:,[0,1,3]], exp[:,2:3]),axis=1)
		sim = np.concatenate((sim[:,[0,1,3]], sim[:,2:3]),axis=1)

		# split in training and test set
		exp_train, exp_test = train_test_split(exp, train_size=120, random_state=0)
		sim_train, sim_test = train_test_split(sim, train_size=174, random_state=0)

		# unite both datasets
		train = np.concatenate((exp_train,sim_train),axis=0)
		test = np.concatenate((exp_test,sim_test),axis=0)

		# logarithm of columns 0,1,3
		trainlog = train.copy()
		trainlog[:,[0,1,3]] = np.log10(train[:,[0,1,3]])

		testlog = test.copy()
		testlog[:,[0,1,3]] = np.log10(test[:,[0,1,3]])

		# scale to unit variance and zero mean
		scaler = StandardScaler().fit(trainlog)
		trainlog_scaled = scaler.transform(trainlog)
		testlog_scaled = scaler.transform(testlog)

		# split into features and target
		xtrain = trainlog_scaled[:,:3]
		ytrain = trainlog_scaled[:,3]

		xtest = testlog_scaled[:,:3]
		ytest = testlog_scaled[:,3]

		# feed to linear models
		lin = None
		for alpha in np.arange(0.1,10.1,0.1):
		for l1_ratio in np.arange(0,1.1,0.1):
		lin_regressor = ElasticNet(alpha=alpha,
		l1_ratio=l1_ratio,
		max_iter=100000,
		random_state=0).fit(xtrain,ytrain)

		if not lin:
		lin = lin_regressor
		r2_train = r2_score(ytrain,lin_regressor.predict(xtrain))
		r2_test = r2_score(ytest,lin_regressor.predict(xtest))

		print()
		print(alpha, l1_ratio, r2_train, r2_test)
		print()

		else:
		if r2_score(ytest,lin_regressor.predict(xtest)) > r2_test:
		lin = lin_regressor
		r2_train = r2_score(ytrain,lin_regressor.predict(xtrain))
		r2_test = r2_score(ytest,lin_regressor.predict(xtest))

		print()
		print(alpha, l1_ratio, r2_train, r2_test)
		print()

		print(lin.get_params())
		print(r2_train, r2_test)

		# feed to kernel ridge regression
		kridge = None
		for alpha in np.logspace(-5,5,101):
		for gamma in np.logspace(-5,5,101):
		kridge_regressor = KernelRidge(alpha=alpha,
		gamma=gamma,
		kernel='rbf').fit(xtrain,ytrain)

		if not kridge:
		kridge = kridge_regressor
		r2_train = r2_score(ytrain,kridge_regressor.predict(xtrain))
		r2_test = r2_score(ytest,kridge_regressor.predict(xtest))

		print()
		print(alpha, gamma, r2_train, r2_test)
		print()

		else:
		if r2_score(ytest,kridge_regressor.predict(xtest)) > r2_test:
		kridge = kridge_regressor
		r2_train = r2_score(ytrain,kridge_regressor.predict(xtrain))
		r2_test = r2_score(ytest,kridge_regressor.predict(xtest))

		print()
		print(alpha, gamma, r2_train, r2_test)
		print()

		print(kridge.get_params())
		print(r2_train, r2_test)

		# feed to tree model
		tree = None
		for max_depth in [2,4,8,16,32,64,None]:
		for min_samples_split in [2,4,8,16,32,64]:
		for min_samples_leaf in [2,4,8,16,32,64]:
		tree_regressor = DecisionTreeRegressor(max_depth=max_depth,
		min_samples_split=min_samples_split,
		min_samples_leaf=min_samples_leaf)\
		.fit(xtrain,ytrain)

		if not tree:
		tree = tree_regressor
		r2_train = r2_score(ytrain, tree_regressor.predict(xtrain))
		r2_test = r2_score(ytest, tree_regressor.predict(xtest))

		print()
		print(max_depth, min_samples_split, min_samples_leaf,
		r2_train, r2_test)
		print()

		else:
		if r2_score(ytest,tree_regressor.predict(xtest)) > r2_test:
		tree = tree_regressor
		r2_train = r2_score(ytrain,
		tree_regressor.predict(xtrain))
		r2_test = r2_score(ytest,
		tree_regressor.predict(xtest))

		print()
		print(max_depth, min_samples_split, min_samples_leaf,
		r2_train, r2_test)
		print()

		print(tree.get_params())
		print(r2_train, r2_test)

		# feed to keras perceptron
		perceptron = None
		for units in [[10,10],[10,10,10],[10,10,10,10]]:
		for activation in ["relu","sigmoid"]:
		model_callback = ModelCheckpoint(filepath = './models/tensorboard-logs/perceptron',
		save_weights_only = True,
		monitor = 'mse',
		mode = 'min',
		save_best_only = True,
		save_freq = 'epoch')
		tensorboard_callback = TensorBoard(log_dir= './models/tensorboard-logs/perceptron',
		histogram_freq = 0,
		write_graph = True,
		write_images = False,
		update_freq = 'epoch',
		profile_batch = 2,
		embeddings_freq = 0,
		embeddings_metadata = None)

		network = build_perceptron(units, activation)

		network.fit(xtrain, ytrain,
		epochs = 1000,
		batch_size = ytrain.shape[0],
		shuffle = True,
		validation_data = (xtest, ytest),
		callbacks = [model_callback, tensorboard_callback])

		if not perceptron:
		perceptron = network
		r2_train = r2_score(ytrain,perceptron.predict(xtrain))
		r2_test = r2_score(ytest,perceptron.predict(xtest))
		_units = units
		_act = activation
		print()
		print(units, activation, r2_train, r2_test)
		print()

		else:
		if r2_score(ytest,network.predict(xtest)) > r2_test:
		perceptron = network
		r2_train = r2_score(ytrain,perceptron.predict(xtrain))
		r2_test = r2_score(ytest,perceptron.predict(xtest))
		_units = units
		_act = activation
		print()
		print(units, activation, r2_train, r2_test)
		print()

		print(_units, _act)
		print(r2_train, r2_test)

		# make final predictions
		train_size = np.shape(ytrain)[0]
		test_size = np.shape(ytest)[0]

		ytrain_lin = lin.predict(xtrain).reshape((train_size,1))
		ytest_lin = lin.predict(xtest).reshape((test_size,1))

		ytrain_kridge = kridge.predict(xtrain).reshape((train_size,1))
		ytest_kridge = kridge.predict(xtest).reshape((test_size,1))

		ytrain_tree = tree.predict(xtrain).reshape((train_size,1))
		ytest_tree = tree.predict(xtest).reshape((test_size,1))

		ytrain_perceptron = perceptron.predict(xtrain).reshape((train_size,1))
		ytest_perceptron = perceptron.predict(xtest).reshape((test_size,1))

		# undo the standardization and the logarithm
		backscaler = StandardScaler().fit(np.reshape(trainlog[:,3],(train_size,1)))

		ytrain_lin = 10**backscaler.inverse_transform(ytrain_lin)
		ytest_lin = 10**backscaler.inverse_transform(ytest_lin)

		ytrain_kridge = 10**backscaler.inverse_transform(ytrain_kridge)
		ytest_kridge = 10**backscaler.inverse_transform(ytest_kridge)

		ytrain_tree = 10**backscaler.inverse_transform(ytrain_tree)
		ytest_tree = 10**backscaler.inverse_transform(ytest_tree)

		ytrain_perceptron = 10**backscaler.inverse_transform(ytrain_perceptron)
		ytest_perceptron = 10**backscaler.inverse_transform(ytest_perceptron)

		# create labels which mark experimental and simulated values
		train_label = np.concatenate((np.full((120,1),'exp'),np.full((174,1),'sim'))
		,axis=0)
		test_label = np.concatenate((np.full((20,1),'exp'),np.full((20,1),'sim'))
		,axis=0)

		# save to files
		np.savetxt("train_random-choice.txt",
		np.concatenate((train,
		ytrain_lin,
		ytrain_kridge,
		ytrain_tree,
		ytrain_perceptron),axis=1),
		header = 'origin,strain rate,dislocation density, strain, yield stress, elastic net model, kernel ridge regression, decision tree, perceptron',
		delimiter=',')
		np.savetxt("train-origin_random-choice.txt",
		train_label, fmt = '%s')

		np.savetxt("test_random-choice.txt",
		np.concatenate((test,
		ytest_lin,
		ytest_kridge,
		ytest_tree,
		ytest_perceptron),axis=1),
		header = 'origin,strain rate,dislocation density, strain, yield stress, elastic net model, kernel ridge regression, decision tree, perceptron',
		delimiter=',')
		np.savetxt("test-origin_random-choice.txt",
		test_label, fmt = '%s')