How can i change the number of iterations per epoch in LSTM RNN time series forecasting model?

Question

Yara Hassan Moustafa 2023-4-2

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1939854-how-can-i-change-the-number-of-iterations-per-epoch-in-lstm-rnn-time-series-forecasting-model

编辑： Sourabh 2023-4-21

I have specified the number of epochs and the min batch size; however in the training process the number of iterations per epoch is still 1!. I have checked the previously asked questions and someone sugggested that i should use squence length option. However, my observations for all sequences are with the same length or have the same number of samples.

So my question is how to change the number of iterations per single epoch in LSTM RNN time series forecasting model?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Sourabh 2023-4-21

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1939854-how-can-i-change-the-number-of-iterations-per-epoch-in-lstm-rnn-time-series-forecasting-model#answer_1220338

编辑：Sourabh 2023-4-21

Greetings,

As far as I am aware, there is no option to manually set the number of iterations per epoch in the LSTM RNN time series forecasting model.

The number of iterations per epoch in a LSTM RNN time series forecasting model is determined by the batch size and the sequence length.

After ensuring that there is no issue with your training data, I would suggest varying the mini-batch size parameter. You could also try and increase your sequence length either by padding the sequences or by concatenating multiple sequences to see if it gives you your desired change in iterations.

It's also worth noting that the number of iterations per epoch is not necessarily a good indicator of your model’s performance. Instead, looking at the loss and/or accuracy metrics gives a better evaluation of your model. You can try adjusting the learning rate or optimizer to see if this improves the training process as well.

You might find the following links useful:

All the best.