Online Tutorials
Some of the resources I frequently used when I was learning Machine Learning and Deep Learning concepts. Don’t forget to reproduce the coding examples on your own computer. This speeds up your learning process and you can accumulate some good algorithms which you can use later.
-
Jeremy Howard’s fastai: Very nice and thorough tutorials on Machine Learning and Deep Learning. Howard uses PyTorch backbone for his fastai framework, but I suggest his material even if you prefer TensorFlow, because he explains the concepts really well.
-
DeepLizard: I watched Deep Learning with PyTorch video series in this blog when I was learning PyTorch and CNNs. You will find some good coding practices here as well.
Lectures
These lectures are helpful to anyone, but especially for those who already built a theoretical background, gained some hands-on experience and looking for some practical advice. Some algorithms look perfect on paper, but finds little to no usage in solving real world problems. These lectures will give you some insights on best practices and common pitfalls.
Talks on Similarity Learning
Similarity Learning is used mainly for person identification and face recognition problems. Its goal is to minimize a similarity metric, usually Euclidean distance or Cosine Similarity, between genuine (alike) pairs while maximizing it for imposter (unlike) pairs. Here are some talks that I found useful to understand the concepts:
-
Jacek Komorowski on Similarity Learning using Deep Neural Networks
-
Tomasz Trzciński on Siamese architecture
-
Andrew Ng on Triplet loss
-
Introduction to Deep Metric Learning. The audio quality is not good, but this is a good introduction to motivation behind the similarity learning approach.