Weekly report 6

Weekly report 6

Hours used: 10

I fixed backpropagation once more, since adding activations to the test networks revealed a bug I had with calculating the base case error. I also corrected the notes about backpropagation, which was rather inconsistent and unclear. I also added some text about backpropagating biases, which I then implemented and tested with Pytorch. Then I added some documentation about the forward pass. I was happy to see that adding the biases made the network a better learner.

This was a very documentation driven week.