2 ms·My current hypothesis is that training data was too noisy.by sanxiyn 8y agoMy current hypothesis is that training data was too noisy.