Hello
I'm doing a project in computer vision for my course, and I have some problems, and I wonder if you can help me.
I read ANN article + lot of staff on the net, and my project is basically finished, except 3 main problems that I don't know how to solve:
1. If I have 1 output neuron, the total error decrease for the pattern ('a'), but when I run the network with another character, the result is the same as if I entered the pattern's character.
2. With 2+ neurons in the output layer, the total error of the layer is always increases :/
3. When a value of an error is too close to 1 to 0, the functions return 1 or 0 (like int value), and it is affect the whole weight calculation (and maybe 1 & 2 above)
I'm looking for someone who can solve these problems..I have no one in college who knows how to program an ANN (just mathematical theory)
I can send my project if you want to..it is very clear
Thank you for your time! :thumbsup: