How is batch learning executed with the trainnet function ?

Hello dear Community,

i have done a lot of reading up on how to implement batch-learning with the trainnet function.

I come to the conclusion that the correct input and target data format is vital.

However, i do not get the correct format.

If someone has already done batch learning with the train net function and has some tips regarding the

input and target data format, i would highly appreciate any input.

Thank you

Best wishes

Answer 1

[Ayush Aniket]

Hi Tobias,

Refer to the following documentation which shows the process of training a deep learning network on out-of-memory sequence data using a custom mini-batch datastore in MATLAB:

https://www.mathworks.com/help/deeplearning/ug/train-network-using-custom-mini-batch-datastore-for-sequence-data.html

The important steps are:

1. Read your sequence data using sequenceDatastore function.
2. Set the mini-batch size of the datastore object.
3. Ensure to set the same mini-batch size in the trainingOptions function.

The format for input data for a vector sequence is s-by-c matrices, where s and c are the numbers of time steps and channels (features) of the sequences, respectively. Refer to the following documentation to read about sequence data input in the trainnet function: https://www.mathworks.com/help/deeplearning/ref/trainnet.html#mw_8b6c966a-f840-4426-9693-279222b8372f

[tobias]

Thank you very much, this might be the right track!

I noticed, that the example is for a classification task, i would want to perform a regression task.

Do you know, if it is possible to define the target values for the regression inside the sequence datastore so that the correct batch of input data is compared with the its corresponding batch of target data ?

Thank you !

[Ayush Aniket]

Please refer the following documentation: https://www.mathworks.com/help/deeplearning/ug/develop-custom-mini-batch-datastore.html#mw_1fbbfc62-d6e2-4e7c-843f-67b467135050