What does MU in trainbr actually mean and can I EEG matrix data as inputs?

I am new to using ANNs practically so please be patient! I have two main questions 1. what does MU actually mean in the trainbr function? I have looked at documentation but the math is above my level of understanding.

2. I have matrices of EEG data. Each column is an ERP, or waveform that I want the network to learn. I am using patternnet. Can I do this? Meaning is the net 'understanding' that my columns are my units?

Thanks in advance very much!

Kim

Answer 1

[Greg Heath]

Before you post a question

1. Use the MATLAB commands

       lookfor, help, and doc

2. Check out Google and Wikipedia

3. For source code use the MATLAB command type

       type trainbr

 >>lookfor trainbr

 >> help trainbr
 ... Levenberg-Marquardt optimization.
 ... Bayesian Regularization
 ...  mu 0.005  Marquardt adjustment parameter
     mu_dec 0.1  Decrease factor for mu
     mu_inc 10  Increase factor for mu
     mu_max 1e10  Maximum value for mu

>> doc trainbr

... LIMITATIONS ...
... MORE ABOUT ... expand all
... Algorithms ...
... References ...

4. FINALLY: GOOGLE LEVENBERG-MARQUARDT

All inputs and target outputs in the NN Toolbox are column vectors

If you are classifying waveforms into c classes, the target columns should be {0,1} unit vector columns of eye(c). However, PATTERNNET is the go to classification algorithm with the default TRAINSCG.

So why are you considering TRAINBR?

Hope this helps.

Thank you for formally accepting my answer

Greg

[Kim]

Thank you! That is very helpful. Just using trainbr because I am new to this and unsure what function is best. I will try patternet!