Aaron Zweig

I’ve recently completed my PhD in the CILVR group at NYU, advised by Joan Bruna.

My thesis primarily concerned the theoretical properties of symmetric and anti-symmetric neural networks. During my PhD I studyied quantatitive approximation bounds based on pairwise attention in symmetric architectures, proved which symmetric functions are provably learnable with gradient methods, and analyzed the capacity to approximate symmetric functions defined over infinitely large sets.

Previously, I’ve worked on graph neural network autoencoders, reinforcement learning theory, and differentiable approximations to discrete operators.

My full CV is available here, feel free to reach me at az831@nyu.edu