• Mojtaba

    Dear Eric,

    Thank you for your very organized and useful blog. I have worked on sparse image representations for my MS. I wanted to use your deep learning code to classify Persian digits (for my research only). Since I’m not well familiar with deep learning, could you please tell me if the codes you provided here and (codes for previous exercises which I have also found on your website) are the only things that I need? That is if I put all of your files in one folder I can run the code for MNIST database without any need to change the code?

    I really appreciate your help. Thank you in advance.

  • BJJ

    Hi Eric,

    Were you able to visualize the wights for both layers?

    Thank you.

  • Ather

    Thankyou for sharing it a nice piece of code helpful of Stacked Autoencoders. I want to use this code for color images feature learning is it possible to use this code for that purpose.

  • Pingback: ()

  • zjq

    Dear Eric,
    Thank you for your sharing this blog.
    I didn’t know much about deep learning before. Recently, I tried to use deep learning technique to predict network traffic. Here I have a question to ask you. If I use the above code provided by you, then I want to know whether I don’t need to change the code to directly run the code for traffic prediction or not?
    I’m looking forward to your reply.
    Best wishes!

  • lxg

    there is something wrong in your stackedAECost.m,
    delta{depth+1} = -(softmaxTheta’ * (groundTruth – p)) .* dsigmoid(a{depth+1});
    delta{layer} = (stack{layer}.w’ * delta{layer+1}) .* dsigmoid(a{layer});

    it should not be disigmoid(a{depth+1}), should be dsigmoid(z{depth+1})