Analog Coding, List Decoding, Bandwidth, and Mean Dimension

Uploaded May 2026 | Updated September 2026, 3 weeks ago

Computer Science/Discrete Mathematics Seminar I
10:30am|Simonyi Hall 101 and Remote Access
Topic: Analog Coding, List Decoding, Bandwidth, and Mean Dimension
Speaker: Elon Lindenstrauss
Affiliation: Institute for Advanced Study
Date: May 4, 2026

Analog Coding, List Decoding, Bandwidth, and Mean Dimension - Elon Lindenstrauss

Duality of Fluid Mechanics and Solution of Decaying Turbulence - Alexander Migdal

Asymptotic Symmetries and Detectors - Part 2 - Sabrina Pasterski

Hodge theory for non-Archimedean analytic spaces - Vladimir Berkovich

Studying Jade in a Peach Blossom Spring | Institute Instances – Yulian Wu

Lattice Packing of Spheres in High Dimensions Using a Stochastically Evolving Ellipsoid-Boaz Klartag

Numerical Modeling of Plasmas Across Fluid and Kinetic Regimes - Chuanfei Dong

Full-kinematics cosmological collider at higher masses and higher loop - Qianshu Lu

Cohen-Macaulayness of Local Models via Shellability of the Admissible Set - Xuhua He

Black hole non-linear stability: a mathematical overview - Elena Giorgi

The Black Hole Explorer: Tracing an Edge of the Visible Universe - Alex Lupsasca

$Algorithms for Overcomplete Tensor Decomposition - Pravesh Kothari Computer Science/Discrete Mathematics Seminar II 10:30am|Simonyi 101 and Remote Access Topic: Algorithms for Overcomplete Tensor Decomposition Speaker: Pravesh Kothari Affiliation: Princeton University Date: April 28, 2026 Tensor decomposition is the task of writing an n x n x n input tensor T as sum_{i = 1}^r a_i otimes b_i otimes c_i for the smallest possible r (called the rank of T). This problem is NP-hard in general. Jennrichs algorithm succeeds if a_is, b_is, and c_is are generic and r less than or equal to n. The overcomplete regime with r greater than n generic components is a major challenge in both algebraic complexity theory and algorithm design, with many applications in statistical estimation. In this talk, I will describe a recent joint work (with Moitra and Wein) that gives an efficient algorithm for overcomplete tensor decomposition whenever r less than or equal to (2-eps)n for every constant eps greater than 0. A key component is a new rank-detection gadget based on Koszul Young Flattenings that reduces tensor rank certification to matrix rank certification.$