Following a 2016 bachelor’s in neuroscience, mathematics and computer science, Sumner Magruder utilized his Boehringer Ingelheim Fonds grant to study neurodegenerative diseases in Germany. There, his work on graph theory, deep learning and visualization—with the support of Nvidia—simultaneously earned him a master’s equivalence in machine learning. He currently focuses on generative adversarial models and software development.
Project: A Deviation from “Try to Use Neural Nets for Graph Isomorphism”
Find an input form of graphs that is sufficient for neural networks to learn on.
Main Results in Detail
Despite sparsity, it seems that the adjacency matrix is sufficient for learning complex and even arbitrary properties of graphs.
- 1. Explore other input formats.
- 2. Export other graph properties.