Anna Go -

Bounding generalization error with input compression: An empirical study with infinite-width networks

paper

We study an information-theoretic generalization bound on neural networks in the infinite-width regime. Published in TMLR.

On interpretability in AI

opinion

In a recent issue of nature reviews physics, Prof. Matthew D. Schwartz from Harvard, whom I know through IAIFI, shared an opinion piece titled “Should artificial…

Understanding (NT)Kernels

tutorial

talk

This is the tutorial I gave at the IAIFI summer school 2022 as a supplement to Yasaman Bahri’s lecture on “Deep learning in the large-width regime”.

Condensing sparse layers

paper

A short paper that we presented at the workshop Sparsity in Neural Networks 2022, containing some perliminary ideas that we are currently expanding into a larger work.

Pruning a restricted Boltzmann machine for quantum state reconstruction

paper

A follow-up to the RBM scaling study with a focus on RBM overparametrization and pruning…

Are wider nets better given the same number of parameters?

paper

talk

This work was done during my internship at Blueshift with Guy Gur-Ari and Behnam Neyshabur.

Learnability scaling of quantum states: Restricted Boltzmann machines

paper

In 2022 we would have named this paper “RBM scaling laws”.

QuCumber: wavefunction reconstruction with neural networks

paper

package

QuCumber stands for “Quantum Calculator Used for Many-Body Eigenstate Reconstruction” (if you suspect we bent the full title to match the abbreviation, you are not…

Batch Normalization is a Cause of Adversarial Vulnerability

paper

We reveal that the use of Batch Normalization makes the network more vulnerable to adversarial attacks. Work presented at the ICML 2019 workshop Identifying and…

Adversarial Examples as an Input-Fault Tolerance Problem

paper

This work is part of my long-term collaboration with Angus Galloway from the Vector Institute and the Uof Guelph. We were interested in…

Restricted Boltzmann Machines

tutorial

These notes are meant to serve as a practical tutorial providing the minimum sufficient knowledge required to build and train a Restricted Boltzmann Machine (RBM).

Machine learning vortices at the Kosterlitz-Thouless transition

paper

This work was my project in the “Theoretical Physics Bootcamp” program at Perimeter Institute and my very fist acquaintance with ML. Published as Editors’ Suggestion in PRB.