def linear_objective(space):
model = get_model(args.model, nfeat=feat_dict["train"].size(1),
nclass=nclass,
nhid=0, dropout=0, cuda=args.cuda)
val_acc, _, _ = train_linear(model, feat_dict, space['weight_decay'], args.dataset=="mr")
print( 'weight decay ' + str(space['weight_decay']) + '\n' + \
'overall accuracy: ' + str(val_acc))
return {'loss': -val_acc, 'status': STATUS_OK}
def run_model(config, fold, fold_base=None):
if config['model'] == 'LSTM':
return train_lstm(config, fold)
if config['model'] == 'BiLSTM':
return train_bilstm(config, fold)
if config['model'] == 'NN':
return train_net(config, fold)
if config['model'] == 'linear':
return train_linear(config, fold)
if config['model'] == 'svm':
return train_svm(config, fold)
if config['model'] == 'random_forest':
return train_random_forrest(config, fold)
if config['model'] == 'baseline':
return train_baseline(config, fold, fold_base)
# Get training and validation sets
(train_fea, test_fea) = getTrainValidSets(feature, 2)
(train_tar, test_tar) = getTrainValidSets(target, 2)
train_fea = feature
train_tar = target
# Feature normalization
(train_fea, m, s) = normalize(train_fea, axis=0)
print(m.shape)
test_fea = normalizePara(test_fea, m, s)
# print ('Training')
# Training
beta = train_linear(train_fea,
train_tar,
test_fea,
test_tar,
eta=1e-5,
lamb=0,
maxIter=1000000,
debug=0)
# Write beta
modFile = open(sys.argv[2], 'w')
beta = beta.tolist()
m = m.tolist()
s = s.tolist()
modFile.write(','.join(str(b[0]) for b in beta) + '\n')
modFile.write(','.join(str(a) for a in m) + '\n')
modFile.write(','.join(str(a) for a in s) + '\n')
from train import train_linear
from parseData import parseData, parseData_sliding
from eva import evaluate
from util import getTrainValidSets, normalize
import numpy as np
import sys
[history, target, factors, trainLen] = parseData_sliding('./data/train.csv')
# Get training and validation sets
(trainData, testTrain) = getTrainValidSets(history, 1.1)
(gndData, testGnd) = getTrainValidSets(target, 1.1)
#
trainData = history
gndData = target
# Feature normalization
# trainData = normalize(trainData, axis=0)
# testTrain = normalize(testTrain, axis=0)
# trainData = np.divide(trainData, 2)
# testTrain = np.divide(testTrain, 2)
beta = train_linear(trainData, gndData, testTrain, testGnd, 5e-9, 1e-10, 1e-6,
500000, 0)
print(factors)
evaluate('./data/test_X.csv', beta, factors, trainLen, sys.argv[1])