Batch or online training?

In the patternnet network, what is the type of training, online or in batch?

Answer 1

[Abhipsa]

Hello @Lorena Andrade, the training type used in “patternnet” network is “batch”. This means that the network weights and biases are updated after processing the entire dataset, as opposed to online training, where updates occur after each individual training example or the given batch size.

As per MATLAB R2021b, “patternnet” uses “batch” processing. You can refer to the link below for a related MATLAB answer:

https://www.mathworks.com/matlabcentral/answers/1470511

Moreover, according to the MATLAB documentation for “patternnet”, no updates regarding this have been mentioned in the version history up to R2024b.

You can refer to the documentation of “patternnet” below:

https://www.mathworks.com/help/releases/R2024b/deeplearning/ref/patternnet.html#f8-3415715_vh

Furthermore, you can check the trainable parameters of the network using the command “>>net.trainParam” in MATLAB command window, where “net” being the name of your “patternnet” instance.

Hope this helps!