Questions

[BasitAlawode]

Hi,

I must say this is a good implementation, simple and easy to follow. Nice work.

I have some questions though.

1. In the denoising example, why did you use sigma*1.15 as the error goal?
2. What do you think is the best ratio of the blocksize^2 to the number of words. e.g 64 ratio 256?
3. How do one know, in advance, the maximum number of blocks to train?
4. Can the algorithm work for pixel values in the range -1 to 1?

Thank you very much.

Regards.