Title: **On the binary perceptron**** **

**Abstract**: We consider the binary perceptron model, a simple model of neural networks that has gathered significant attention in the statistical physics, information theory and probability theory communities. For the symmetric binary perceptron model, we establish that the partition function of this model, normalized by its expected value, converges to a lognormal distribution. As a consequence, this allows us to establish several conjectures for this model, including contiguity, sharp threshold and strong freezing conjectures. The strong freezing property further points to an interesting phenomenon that is not present in sparse constraint satisfaction problems: on the one hand, the perceptron model is conjectured to be typically hard; on the other hand, efficient algorithms have been shown empirically to succeed in finding solutions, suggesting that such algorithms find atypical solutions. We establish formally such a phenomenon for both the symmetric and asymmetric binary perceptrons. We show that at low constraint density, there exists a subdominant connected cluster of solutions with almost maximal diameter, and that an efficient multiscale majority algorithm can find solutions in such a cluster with high probability.

*This is based on joint works with Emmanuel Abbe and Allan Sly.*

**Shuangping Li **is a fifth-year Ph.D. student in PACM at Princeton University, advised by Professor Allan Sly and Professor Emmanuel Abbe. Prior to Princeton, she obtained her Bachelor of Science in Mathematics from University of Hong Kong. Her research focuses on probability theory and theoretical machine learning.