Neural Network Demo

Source code: Standalone .exe: (PS: Feel free to PM me Q’s about the sourc…
Video Rating: 4 / 5

11 thoughts on “Neural Network Demo

  1. Hi, i’m interested about your work. I’m doing quite the same, but with fish, with 6 different input and 4 output (one for forward, one backward and 2 for direction). I’m trying to make the network evolve himself and the connexion are done randomly. As you had some result I have some questions : How do you update the wheight of the connexions for each generations? How do you choose the hidden layer? Thanks and sorry, i’m frensh and my english isn’t realy good.

  2. Because simple perceptrons, without hidden layers, can only solve problems that are linearly separable. Complex problems quickly become non-linearly separable. Driving is an example of such problem. Hidden layers make it possible for a neural network to solve problems that are non-linearly separable. Or do you mean, why use 8 instead of X hidden nodes?

  3. That is not how it works. Neural networks can be trained within simulations. Perhaps they still need some tuning afterwards, but it isn’t as expensive as you think it is. Most of the training is done in a safe and relatively cheap environment.

Leave a Reply