Public Domain Diagrams



Brain-shaped silhouette in cellular automata patterns.



This page is a collection of images and diagrams I’ve made to illustrate various machine learning topics. The images on this page are all released into the public domain unless otherwise noted, so feel free to use them in your own projects. Credit is appreciated but not required.

Attention



Illustation of dot-product attention.






Example of attention used on the edges of a graph. png





Illustation of attention encoder layer.



Deep Learning



Illustration of the role of model and dataset size in deep learning.




Convolution. png




Cartoon of translation invariance in convolution.




Cartoon diagram of a fully convolutional network and a conv-net with atrous convolutions




Cartoon diagram of a U-Net style convolutional neural network.




Convolution as element-wise multiplication in the Fourier domain. png




Shallow multilayer perceptron.




Illustation of “deep double descent.”




Overfitting. png


Graphs and Graph Neural Networks



A graph and its adjacency matrix. png




Arranging graph convolutions as matrix multiplication. png


Word vectors



Cartoon illustrating word relationships in vector space (e.g. word vectors, word2vec).


Transformers



Illustation of a switch encoder layer. png




illustation of dot-product attention, used in transformers. png


Protein Folding and Protein Structure



Cas9 protein with guide RNA, based on PDB structure 4OO8



Cartoon illustration of Levinthal’s paradox



Cartoon representation of multiple sequence alignment in bioinformatics.