– Sorry to say, but this is a terrible video to follow. I know how it works, and I had to rewind every minute or so because the information you present is much too dense. Names dropping all around, lack of animations, … There is a very real possibility that someone who has never been in contact with deep learning and/or a less technical background, will be scared off. I would propose (yes I know, it takes time) to make more custom animations (e.g. when you’re discussing activation functions), avoid certain parts of the explanation such as the regularization by weights, add text on the slide for key concepts (perceptron, sigmoid, logistic regression, pooling layer, batch layer, drop-out layer, …), …
– The first half of the presentation is about artificial neural networks, but is labelled “Intro to deep learning”. Explaining artificial networks as such is an interesting topic and deserves a separate video. I would propose to split the video in 2 halves: one for general NN and one for CNN.
– I don’t really like that you keep referring to microstructure as an example. Once can be okay, but it is presented too prominently and focuses the attention of the viewer too much towards that topic.