search

shobrook / sequitur

Library of autoencoders for sequential data
MIT License
410 stars 56 forks source link

about LSTM ae result less than 1 #8

Open guowhite opened 2 years ago

guowhite commented 2 years ago

hi, i was use the trail dataset like [[x1,y1], [x2,y2]] and change the shape to [[x1,x2,x3], [y1,y2,y3]] length about 150 and use LSTM AE to train it . i want to use this model to Implement representation learning.but the loss could not reduce. i did not know how to solve it. could you give me some suggestion? I will be looking forward to your rep

guowhite commented 2 years ago

hi, i was use the trail dataset like [[x1,y1], [x2,y2]] and change the shape to [[x1,x2,x3], [y1,y2,y3]] length about 150 and use LSTM AE to train it . i want to use this model to Implement representation learning.but the loss could not reduce. i did not know how to solve it. could you give me some suggestion? I will be looking forward to your reply