MATLAB: Workaround for Parallel Computing with LSTM

Question

At the moment, I am unable to train LSTMs on MATLAB with a parallel pool. MATLAB says it does not seem to support a parallel environment execution when training LSTM networks. Are there any suggestions as to what can be done to workaround this? Perhaps a very similar type of network that does support parallel computing?

My team is deciding on whether or not to switch to Tensorflow and leave MATLAB due to this limitation. Is there any APIs or Toolboxes that can be used to use parallel pools with RNNs + LSTMs?

Thanks.

Answer 1

Hi Aditya,

Using multiple workers are not yet supported for LSTMs or recurrent networks. Unfortunately, there is no workaround for training recurrent neural networks like LSTM in parallel. The developers are aware of this behaviour and will try to provide a workaround in future releases.

Hope this helps.