def test_transform_boxcox(self):
x = np.hstack([np.random.standard_cauchy(size=(1000, 2))**2, np.exp(np.random.normal(size=(1000, 2)))])
out = g.Gaussianize(strategy='boxcox') # Only works on positive data
out.fit(x)
y = out.transform(x)
x_prime = out.inverse_transform(y)
assert np.allclose(x_prime, x)
out.qqplot(x, 'boxcox')
def test_transform_lambert(self):
x = np.hstack([np.random.standard_cauchy(size=(1000, 2)), np.random.normal(size=(1000, 2))])
out = g.Gaussianize()
out.fit(x)
y = out.transform(x)
x_prime = out.inverse_transform(y)
assert np.allclose(x_prime, x)
out.qqplot(x, 'lambert')
def to_gaussian(score_list):
x = np.array(score_list)
out = g.Gaussianize(strategy='brute')
out.fit(x)
y = out.transform(x)
return y.flatten()
def single_process(filename_train,filename_test,filename_save):
#filename=filename_list[idx]
#print filename
name_list_train,score_list_train=read_RNAcompete_data_2(filename_train)
score_list_train,name_list_train=shuffle_RNAcompete(score_list_train,name_list_train)
name_list_test,score_list_test=read_RNAcompete_data_2(filename_test)
score_list_test,name_list_test=shuffle_RNAcompete(score_list_test,name_list_test)
#score_list_train,name_list_train,score_list_test,name_list_test=leave_out(score_list,name_list)
preprocess_scaler=gaussianize.Gaussianize(strategy='brute')
preprocess_scaler.fit(score_list_train)
score_list_train=preprocess_scaler.transform(score_list_train)
score_list_train/=2.5
'''
plt.hist(score_list_train)#,plt.show()
plt.savefig('input_'+filename+'.png')
plt.close()
'''
filename_positives=os.path.join(output_pathname,'tmp_'+filename_save+'.positives.fa')
filename_negatives=os.path.join(output_pathname,'tmp_'+filename_save+'.negatives.fa')
#print filename_positives,filename_negatives
generate_RNAcompete_training_data(name_list_train,score_list_train,filename_positives,filename_negatives)
filename_test_tmp=os.path.join(output_pathname,'tmp_'+filename_save+'.test.fa')
fout_test=open(filename_test_tmp,'w')
for name in name_list_test:
print >>fout_test,'>',name
print >>fout_test,name
fout_test.close()
scaler,_model=training(output_pathname,'tmp_'+filename_save.replace('.txt',''))
pred=predicting2(filename_test_tmp,scaler,_model,None)
#pred=preprocess_scaler.invert(pred)
#print pred-score_list_test
score_list_test=preprocess_scaler.transform(score_list_test)
score_list_test/=2.5
score_list_test=score_list_test.ravel()
#print 'ave:',np.abs(pred-score_list_test).mean()
coef=stats.pearsonr(pred,score_list_test)
print filename_test,'correlation:',coef
plt.scatter(score_list_test,pred),plt.plot([vmin,vmax],[vmin,vmax],'r'),plt.xlim(vmin,vmax),plt.ylim(vmin,vmax)#,plt.show()
plt.savefig(os.path.join(output_pathname,filename_save+'.png'))
plt.close()
#save = np.array([pred,score_list_test])
filename_save=os.path.join(output_pathname,filename_save+'.pred')
fout_save = open(filename_save,'w')
for i in xrange(len(name_list_test)):
print >>fout_save,name_list_test[i],pred[i]
fout_save.close()
#print filename_positives
#print filename_test
os.remove(filename_positives)
os.remove(filename_negatives)
os.remove(filename_test_tmp)
return coef