Chemistry aware lifting of molecular substructures using TNN

Title : Chemistry aware lifting of molecular substructures using TNN

Abstract:

Topological Neural Networks (TNNs) are a generalization of Graph Neural Networks (GNNs), which are able to model connections between nodes at higher ranks. Organic molecules naturally fit these models as there are interactions which occur in groups larger than 2 nodes, which can be better modeled through the interactions of TNNs. This work applies 6 popular TNNs, spanning Cellular Complexes, Combinatorial Complexes and Hypergraphs to three separate datasets, including the widely used Zinc-500k dataset, and compares their performances.

Biography:

Noga Bregman is an BSc student in Computer Science at Tel Aviv University, specializing in Deep Learning, Graph Neural Networks, and Algorithmic Game Theory. With a professional experience in the AI horizon team at PayPal, Noga has worked on cutting-edge technologies, including graph-based solutions for fraud detection and recommendation systems. Passionate about innovation and problem-solving, Noga integrates her academic expertise and industry skills to drive impactful projects in data science and machine learning.