Gradient, CNN, Evolution of Deep Learning
What's New? (5/2)
Gradient
I finished the last chapter of the India Edition of our popular Deep Learning Math Workbook. Here’s the preview. This chapter teaches you what happens if you change the weight of a connection and how to calculate “gradient” to quantify the change.
Please download the PDF and give it a try. Let me know what you think.
CNN
How can you calculate a Convolutional Neural Network (CNN) by hand? ✍️ I posted two lecture videos to show you Forward pass and Backward pass. How do you like my creative thumbnail design? 😉 I feel a bit dizzy upside down.
Evolution of Deep Learning
As my tribute to Geoff Hinton's Nobel Prize, I drew this animation to illustrate the key idea behind Hinton's major contributions to deep learning over the years, with artistic liberty.
A few comments on Chapter 12 PDF:
* In the diagram for boxes 21/22/23, in the right hand side, the fraction says dZ/dW2, while it should say dZ/dW3 because W3 is changing. In the same fraction, the value for the denominator of the fraction is already filled in, while it should be blank and should be labeled as box 22.
* In the diagram for boxes 26/27/28, the top dotted horizontal arrow starts from W1 and ends at W1, while it should start at W3 and end at W3.