What is the difference between training, adapting, and learning in the neural network?

I know that we can use different training functions. I do not know what is the difference between [net,tr] = adapt(net,inputs,targets) or net.trainFcn = 'traingdx';[net,tr] = train(net,inputs,targets) or net.adaptFcn = 'trains'; net.inputWeights{1,1}.learnFcn = 'learngd'; net.biases{1}.learnFcn = 'learngd';

I was wondering if some one has any experience about them?

Answer 1

[Greg Heath]

NN learning is the process of modifying net parameters to try to achieve a goal.

I interpret adaptation as learning from a sequence of I/O examples where parameters are updated after each presentation of a single example. I equate this with the term "sequential learning"

I interpret training as learning from a group of I/O examples where parameters are updated after each presentation of the whole group. I equate this with the term "batch learning"

The official definitions may differ. If you find some, please post the references.

For any function:

1.Use the help and doc commands to obtain the online documentation and a simple example

2. Use the type command to see the source code.

3. Use the lookfor command to find ALL instances of a term in the help section of ANY function

For example try

 lookfor learn

and

 lookfor train

Hope this helps.

• Thank you for formally accepting my answer*

Greg

[Greg Heath]

See the glossary in the NNTBX documentation

doc nnet

[Sanaz]

I appreciate your attention and response. I agree about train and adapt but what about learn. I found the example for learning but I don't know how I can use learn? Is it batch or incremental training? or if there is any specific condition that any of those work better?

[Greg Heath]

There are two ways to learn:

1. Batch

2. Incremental

Batch tends to be the default because it is orders of magnitude faster for large problems.