Numer of samples in deep learning problem

Marcos888

Hello.

I am solving a deep learning prediction problem and when I try to train more than 100k sample the Neural Network just catch around the 20% of the training samples set. Anyone knows why?

Thanks

Marcos888

This is the file.

Im sorry to answer so late, but I hadn´t seen the answer up to now.

The file have 50000 rows and 1001 atributes ( the last is the target)

Thank you in advance

varunm1

Yes.

varunm1

Hello @Marcos888

Can you post your process here? You can use FILE --> Export Process in Rapidminer and attach the .rmp file here so that we can take a look at the process and get back to you.

If you need us to debug or test the process please attach your data here or you can send in private message.

Marcos888

This is the file.

Im sorry to answer so late, but I hadn´t seen the answer up to now.

The file have 50000 rows and 1001 atributes ( the last is the target)

Thank you in advance

varunm1

Hello @Marcos888

I checked your process. I have a question regarding "Loop Batches" operator. Are you trying to use that to set Mini batch size to train a deep learning algorithm? If so, then that is not needed and I dont think this way works. If you are looking to set a mini batch size of "1000" then you need to go to "expert parameters"  setting in Deep learning operator parameters. There you can find "Mini_batch_size" parameter as shown below. Please let us know if this is what you are looking for.

Marcos888

So, in RapidMiner "Mini_batch_size" are the same as usually "batch_size" parameter?

varunm1

Yes.