-
Notifications
You must be signed in to change notification settings - Fork 553
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
A bit puzzled about the results of using this FFNN #11
Comments
Forget about it, just noob initialization of the weights. |
Glad you got it figured out. Initializing the weights as It sounds like you're using the MNIST training set. I've had good results training |
Ah, that's kind of good to know (that means we're working on the same thing at the same time, so not so good :D ), interested to see what will come out! Will check the iOS example project in a few days |
I'm only writing it for iOS right now. If you're writing an OS X version, that will be great to have. I'll move the MNIST data to the Examples directory so it's accessible for both apps (so we don't package two copies). I'm adding Softmax activation to |
Hi,
I finally tried your FFNN (since until now I just lurked reading the code and the future developments), and I stumbled across something I don't fully understand:
I'm trying to train a FFNN with
784
inputs and10
outputs, with100
hidden nodes. Momentum is0.5
and learning rate0.7
, error functionDefault
and activationSigmoid
.This is the output of the first iteration of the training (first input/output), I have
50000
items in my training set but all the items give the same result (so imagine this output pasted 50000 times).The training ends immediately since the error is "0" and every test I make with a validation set "succeeds" with output
[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
(that of course is wrong).Do you have an idea of why this happens?
The text was updated successfully, but these errors were encountered: