Calculating dz2 in batch training

[jezek]

Hello. Investigating your blog, I've stumbled upon this line in training function:

xor-neural-network/XOR-Net-Notebook.ipynb

Line 433 in af6eda3

" dz2 = a2 - y[j] # 1x1\n",

If following the math background of your source, they compute the output layer delta (in your case dz2) as δx,L=∇aCx⊙σ′(zx,L), which should be translated to your code as:

 dz2 = np.multiply(a2 - y[j], sigmoid(a2, derv=True)) # 1x1 .* 1x1 = 1x1
 

Or am I missing something?

Thank you for your time (and repo) and I'm looking forward for your answer.

[morvin117]

Sorry? you have send your messege to wrong e-майл, re-send it to write man/ Goog luck !          

Воскресенье, 16 августа 2020, 18:40 +03:00 от jEzEk ***@***.***>:     Hello. Investigating your blog, I've stumbled upon this line in training function: https://github.com/Frixoe/xor-neural-network/blob/af6eda3f8bb25e5a5f2e6f062a3a0e7080af2491/XOR-Net-Notebook.ipynb#L433 If following the math background of your source , they compute the output layer delta (in your case dz2) as δx,L=∇aCx⊙σ′(zx,L), which should be translated to your code as: dz2 = np.multiply(a2 - y[j], sigmoid(a2, derv=True)) # 1x1 .* 1x1 = 1x1 Or am I missing something? Thank you for your time (and repo) and I'm looking forward for your answer. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub , or unsubscribe .

 

[jezek]

@morvin117

Sorry? you have send your messege to wrong e-майл, re-send it to write man/ Goog luck !          

I did not send email. I wrote an issue to @Frixoe github repo xor-neural-network. Maybe you are watching the repository and you got notification from github to your email.

Thank you for your concerns and have a nice life. 😉