def use_dtw(dataset_path):
dataset=seq_dataset(path)
test,train=split.person_dataset(dataset)
y_pred=wrap(train,test)
print(utils.data.find_errors(y_pred,test))
seq.check_prediction(y_pred,test['y'])
y_i=y[i]
mask_i=mask[i]
loss_i=model.train(x_i,y_i,mask_i)
cost.append(loss_i)
sum_j=sum(cost)/float(len(cost))
print(str(j) + ' ' + str(sum_j))
return model
def check_prediction(y_pred,y_true):
print(classification_report(y_true, y_pred))
print(confusion_matrix(y_true,y_pred))
def get_batches(x,batch_size=6):
n_batches=len(x)/batch_size
if((len(x) % batch_size)!=0):
n_batches+=1
return [x[i*batch_size:(i+1)*batch_size]
for i in range(n_batches)]
if __name__ == "__main__":
path='../dane5/seq/'
nn_path='../dane5/lstm_'
dataset=to_dataset.seq_dataset(path)
new_dataset=to_dataset.masked_dataset(dataset)
test,train=split.person_dataset(new_dataset)
print(train.keys())
model=make_model(train,False)
model.get_model().save(nn_path)
check_model(model,test)