[theory students] Fw: Probability seminar Apr. 13: Brice Huang (MIT)

[ILIAS DIAKONIKOLAS <ilias@xxxxxxxxxxx>]

A seminar of interest this week.

---

From: SEBASTIEN ROCH <roch@xxxxxxxxxxxxx>
Sent: Wednesday, April 12, 2023 4:35 PM
To: ILIAS DIAKONIKOLAS <ilias@xxxxxxxxxxx>
Subject: FW: Probability seminar Apr. 13: Brice Huang (MIT)

 

Hi Ilias:

 

This week’s probability seminar may be of interest to the theory group.

 

-Sbt

 

From: 'Erik Bates' via probsem <probsem@xxxxxxxxxxxxxxxxx>
Date: Wednesday, April 12, 2023 at 4:33 PM
To: probsem@xxxxxxxxxxxxxxxxx <probsem@xxxxxxxxxxxxxxxxx>
Subject: Probability seminar Apr. 13: Brice Huang (MIT)

Hi, Everyone,

 

This week’s seminar will be given by Brice Huang, who is visiting through Friday.  Join us at the usual time and place.

 

Time: 2:30pm, Thursday April 13

Place: 901 Van Vleck

 

Best,

 

Erik

 

April 13, 2023, in person: Brice Huang (MIT)

Algorithmic Threshold for Multi-Species Spherical Spin Glasses

This talk focuses on optimizing the random and non-convex Hamiltonians of spherical spin glasses with multiple species. Our main result identifies the best possible value ALG achievable by class of Lipschitz algorithms and gives a matching algorithm in this class based on approximate message passing. The threshold ALG is given by a certain variational problem, which surprisingly may possess multiple optimizers.

Our hardness result is proved using the Branching OGP introduced in our previous work [H-Sellke 21] to identify ALG for single-species spin glasses. This and all other OGPs for spin glasses have been proved using Guerra's interpolation method. We introduce a new method to prove the Branching OGP which is both simpler and more robust. It works even for models in which the true maximum value of the objective function remains unknown.

Based on joint work with Mark Sellke.