I was wondering if anyone has any experience on this subject. The goal is to develop neural network capable of geting input from microphone and then recognising sound patterns. That will be replacement for fft. Detected sound ranges from 20Hz-20kHz. :o
This needs deep optimization, not only in C, but even in assembly. C# will not get you this.
How about visual c++?
Assembly is closest to the hardware. Everything else adds overhead.
You may need to use dedicated hardware. Such as RSC-4128 from Sensory the http://www.sensoryinc.com/products/RSC-4x_series.html
Has anyone actualy implemented neural network on a microcontroler ? If Ram is the problem I could, possibly, organise sd memory to act as a operating memory.