I find that sometimes the best way to understand a topic is to see it in action. That can be tricky for some topics, like computer vision and deep learning, so I created this page to aggregate some of my favorite visualizations in these fields. I hope to continuously add to this page. Let me know if you have any suggestions for what made an idea click for you!
- Hands-on demonstration of how image kernels work
- Beautiful step-by-step process of matrix multiplication
- Possibly the best place to gain an intuition of how neural networks work is the Tensorflow Neural Network Playground
- The Standford CS231n webpage shows how convolutional neural networks work very clearly
- Distill is an online journal for clearly presenting machine learning research. There is a lot of great content here.
- Grant Sanderson’s Youtube channel, 3Blue1Brown, is amazing for gaining a deep understanding of many of the mathematical concepts behind deep learning as well as an appreciation for the beauty of it. He has a series called Essence of linear algebra that I recommend. He is the great math teacher I have ever known.
- There are a lot of great visualizations on this Medium post. I especially like the gif of the handwritten “8” flipping back and forth with the matrix values. Understanding that there are numbers behind every pixel is the first step in computer vision, so I always show this to people when I’m explaining it.
- I don’t know if this counts as a visualization but the Wikipedia List of mathematical symbols is a good place to look up any notation you don’t know. If you don’t know what a given set of squiggles means, it’s hard to Google it, so check to see if it’s on this list.
- Andrei Kaparthy made an excellent demo of how neural networks create separability in a toy dataset.
- Towards Data Science has an excellent blog post on the math behind convolutional neural networks
- dtreeviz makes beautiful visualizations of decision trees