Question about the size of Y in DARNN code written in Keras

kaelzhang / DA-RNN-in-Tensorflow-2-and-PyTorch

A Tensorflow 2 (Keras) implementation of DA-RNN (A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction, arXiv:1704.02971)

MIT License

30 stars 12 forks source link

Closed meticulousdev closed 1 year ago

meticulousdev commented 1 year ago

I have a question regarding the DARNN code written in Keras.

In line 439 of model.py file (da_rnn/keras/model.py), Y is extracted from the inputs.

However, using Y = inputs[:, :, -self.y_dim:], the size of Y becomes (batch_size, T, y_dim) instead of (batch_size, T - 1, y_dim).

Could you please check this code?

kaelzhang commented 1 year ago

Sorry, this repo is only for research purpose. Actually, the performance of DARNN is really poor, so I did not spend too much time on it.

The keras implementation is buggy, but pytorch one works. You might as well check the pytorch implementation.

Don't use this model in production, it will lose money.

meticulousdev commented 1 year ago

Thank you for replying to my question.

I was also studying DARNN for the purpose of the study.

I will check the pytorch version as well.

Thank you again!