micro for neural network ? Ridiculous ? please enlighten me.

[dna_electro]

now i have one idea.

i'm often heard, character recognition with blablabla neural network. but how do i implement it ? i sometime think that i want to connect the graphic tablet and process its output with NN as character recognitio(remember CHARACTER, not WORD).

but one that stuck in my mind, is micro can process NN (let's assume worst case : 49 input (7x7 px),26 output (A-Z), 6-10 hidden layer, or probably perceptron NN), so we put the recognition program in micro. and learning data process in cpu.

recognition process in cpu is out of question, because if i process in cpu, it's a waste only using the graphic tablet output as NN input.

plese help, thanks.
eric

 

[3v0]

Not too many years ago the idea of a nural network on a micro controller would have been a joke.

The following paper talks about a nural net on a PIC18. I would use one of the 32 bit processors.
https://www.eng.auburn.edu/~wilambm/pap/2008/A%20Neural%20Network%20Implementation%20on%20an%20Inexpensive%20Eight%20Bit%20Microcontroller.pdf

 

[dna_electro]

thanks

thanks for the input
reading it, ang hoping you will help for any problems in advance.

btw, why choose picaxe ? in my country is harder to find picaxe micro.

 

[3v0]

My post and the article talked about PIC's not PICAXE.

A PICAXE is a PIC with a basic interperter programmed into flash memory and the user program in EEPROM.