def mlp_predict(model_file_name, test_file_name, pred_file_name, n_dim):
logger.info('loading test file')
(test_set_x, test_set_y) = svm2numpy(test_file_name, n_dim)
logger.info('predicting')
f = open(pred_file_name, 'w')
for label in get_y_pred(model_file_name, test_set_x):
f.write(str(label)+'\n')
def featdel(input_file_name, n_features, model_file_name):
logger.info('loading input file: ' + input_file_name)
x, y = svm2numpy(input_file_name, n_features)
logger.info('deleting features')
x_sum = x.sum(axis=0)
model = []
for idx in xrange(len(x_sum)):
if x_sum[idx] == 0.0:
model.append(idx)
logger.info(('before deletion: %i, after deletion: %i') % (n_features, n_features - len(model)))
logger.info('outputing model to file: ' + model_file_name)
np.save(model_file_name, np.array(model))
def gen_feat(input_file_name, n_features, model_file_name, output_file_name):
logger.info('loading input file: ' + input_file_name)
x, y = svm2numpy(input_file_name, n_features)
logger.info('loading model file: ' + model_file_name)
model = np.load(model_file_name)
logger.info('outputing file: ' + output_file_name)
x = np.delete(x, model, 1)
f = open(output_file_name, 'w')
for n in xrange(len(x)):
output_str = str(y[n])
for idx in xrange(len(x[n])):
if x[n, idx] != 0:
output_str = ('%s %i:%f') % (output_str, idx, x[n, idx])
f.write(output_str + '\n')
def gen_feat(input_file_name, n_features, model_file_name, output_file_name):
logger.info('loading input file: ' + input_file_name)
x, y = svm2numpy(input_file_name, n_features)
logger.info('loading model file: ' + model_file_name)
mask_idx = np.load(model_file_name)
logger.info('len after selection: %i' % (len(mask_idx)))
logger.info('outputing file: ' + output_file_name)
f = open(output_file_name, 'w')
for n in xrange(x.shape[0]):
output_str = str(y[n])
ptr = 1
for idx in mask_idx:
if x[n, int(idx)-1] != 0.0:
output_str = ('%s %i:%f') % (output_str, ptr, x[n, int(idx)-1])
ptr += 1
f.write(output_str + '\n')
