## Seminars and Colloquia by Series

### Spatial mixing and the Swendsen-Wang dynamics

Series
ACO Seminar
Time
Friday, September 18, 2020 - 15:00 for 1 hour (actually 50 minutes)
Location
Speaker
Antonio Blanca Pennsylvania State University

The Swendsen-Wang dynamics is a popular algorithm for sampling from the Gibbs distribution for the ferromagnetic Ising and Potts models. The dynamics is a global Markov chain that is conjectured to converge quickly to equilibrium even at low temperatures, where the correlations in the system are strong and local chains converge slowly. In this talk, we present new results concerning the speed of convergence of the Swendsen-Wang dynamics under spatial mixing (i.e., decay of correlations) conditions. In particular, we provide tight results for three distinct geometries: the integer d-dimensional integer lattice graph Z^d, regular trees, and random d-regular graphs. Our approaches crucially exploit the underlying geometry in different ways in each case.

### CANCELLED for now: TBA by Prasad Tetali

Series
ACO Seminar
Time
Thursday, April 2, 2020 - 14:00 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Peter WinklerDartmouth College, NH

TBA

### An isoperimetric inequality for the Hamming cube and some consequences

Series
ACO Seminar
Time
Thursday, December 5, 2019 - 13:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Jinyoung ParkRutgers University

I will introduce an isoperimetric inequality for the Hamming cube and some of its applications. The applications include a “stability” version of Harper’s edge-isoperimetric inequality, which was first proved by Friedgut, Kalai and Naor for half cubes, and later by Ellis for subsets of any size. Our inequality also plays a key role in a recent result on the asymptotic number of maximal independent sets in the cube.

This is joint work with Jeff Kahn.

### Online algorithms for knapsack and generalized assignment problem under random-order arrival

Series
ACO Seminar
Time
Tuesday, September 24, 2019 - 13:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Arindam KhanComputer Science and Automation, Indian Institute of Science, Bangalore

For online optimization, the input instance is revealed in a sequence of steps and, after each step, the algorithm has to take an immediate and irrevocable decision based on the previous inputs. Online algorithms produce a sequence of decisions for such problems without the complete information of the future. In the worst-case analysis of online optimization problems, sometimes, it is impossible to achieve any bounded competitive ratio. An interesting way to bypass these impossibility results is to incorporate a stochastic component into the input model. In the random-order arrival model, the adversary specifies an input instance in advance but the input appears according to a random permutation. The knapsack problem is one of the classical problems in combinatorial optimization: Given a set of items, each specified by its size and profit, the goal is to find a maximum profit packing into a knapsack of bounded capacity. The generalized assignment problem (GAP) includes, besides the knapsack problem, several important problems related to scheduling and matching. In this talk, we will present improved competitive algorithms under random-order arrival for these two problems. This is joint work with Susanne Albers and Leon Ladewig.

### The cap set problem

Series
ACO Seminar
Time
Friday, January 20, 2017 - 15:05 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Dion GijswijtTU Delft
A subset of $\mathbb{F}_3^n$ is called a \emph{cap set} if it does not contain three vectors that sum to zero. In dimension four, this relates to the famous card game SET: a cap set corresponds to a collection of cards without a SET. The cap set problem is concerned with upper bounds on the size of cap sets. The central question raised by Frankl, Graham and R\”odl is: do cap sets have exponentially small density? May last year, this question was (unexpectedly) resolved in a pair of papers by Croot, Lev, and Pach and by Ellenberg and myself in a new application of the polynomial method. The proof is surprisingly short and simple.

### On the method of typical bounded differences

Series
ACO Seminar
Time
Tuesday, November 15, 2016 - 13:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Lutz WarnkeCambridge University and Georgia Tech
Concentration inequalities are fundamental tools in probabilistic combinatorics and theoretical computer science for proving that functions of random variables are typically near their means. Of particular importance is the case where f(X) is a function of independent random variables X=(X_1,...,X_n). Here the well-known bounded differences inequality (also called McDiarmid's or Hoeffding--Azuma inequality) establishes sharp concentration if the function f does not depend too much on any of the variables. One attractive feature is that it relies on a very simple Lipschitz condition (L): it suffices to show that |f(X)-f(X')| \leq c_k whenever X,X' differ only in X_k. While this is easy to check, the main disadvantage is that it considers worst-case changes c_k, which often makes the resulting bounds too weak to be useful. In this talk we discuss a variant of the bounded differences inequality which can be used to establish concentration of functions f(X) where (i) the typical changes are small although (ii) the worst case changes might be very large. One key aspect of this inequality is that it relies on a simple condition that (a) is easy to check and (b) coincides with heuristic considerations as to why concentration should hold. Indeed, given a good' event G that holds with very high probability, we essentially relax the Lipschitz condition (L) to situations where G occurs. The point is that the resulting typical changes c_k are often much smaller than the worst case ones. If time permits, we shall illustrate its application by considering the reverse H-free process, where H is 2-balanced. We prove that the final number of edges in this process is concentrated, and also determine its likely value up to constant factors. This answers a question of Bollobás and Erdös.

### Polynomials and (Finite) Free Probability

Series
ACO Seminar
Time
Tuesday, November 3, 2015 - 16:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Adam MarcusMathematics and PACM, Princeton University
Recent work of the speaker with Dan Spielman and Nikhil Srivastava introduced the method of interlacing polynomials'' (MOIP) for solving problems in combinatorial linear algebra. The goal of this talk is to provide insight into the inner workings of the MOIP by introducing a new theory that reveals an intimate connection between the use of polynomials in the manner of the MOIP and free probability, a theory developed by Dan Voiculescu as a tool in the study of von Neumann algebras. I will start with a brief introduction to free probability (for those, like me, who are not operator theorists). In particular, I will discuss the two basic operations in free probability theory (the free additive and free multiplicative convolutions), and how they relate to the asymptotic eigenvalue distributions of random matrices. I will then show how certain binary operations on polynomials act as finite analogues of the free convolutions and how the MOIP is effectively transferring the asymptotic bounds obtained in free probability to bounds in the new theory (which can then be applied to finite scenarios). If time permits, I will show how such a theory gives far better intuition as to how one might apply the MOIP in the future, using recent results on restricted invertibility and the existence of Ramanujan graphs as examples.

### Extremal Cuts of Sparse Random Graphs

Series
ACO Seminar
Time
Monday, October 5, 2015 - 13:05 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Amir DemboStanford University
The Max-Cut problem seeks to determine the maximal cut size in a given graph. With no polynomial-time efficient approximation for Max-Cut (unless P=NP), its asymptotic for a typical large sparse graph is of considerable interest. We prove that for uniformly random d-regular graph of N vertices, and for the uniformly chosen Erdos-Renyi graph of M=N d/2 edges, the leading correction to M/2 (the typical cut size), is P_* sqrt(N M/2). Here P_* is the ground state energy of the Sherrington-Kirkpatrick model, expressed analytically via Parisi's formula. This talk is based on a joint work with Subhabrata Sen and Andrea Montanari.

### CANCELLED: Greedy-like algorithms and a myopic model for the non-monotone submodular maximization problem

Series
ACO Seminar
Time
Monday, April 7, 2014 - 13:00 for 1 hour (actually 50 minutes)
Location
Klaus 1116
Speaker
Allan BorodinUniversity of Toronto
We are generally interested in the following ill-defined problem: What is a conceptually simple algorithm and what is the power and limitations of such algorithms? In particular, what is a greedy algorithm or more generally a myopic algorithm for a combinatorial optimization problem? And to be even more specific, motivated by the Buchbinder et al `online double sided greedy algorithm'' for the unconstrained non-monotone submodular maximization problem, what are (if any) the limitations of algorithms "of this genre" for the general unconstrained problem and for specific instances of the problem, such as Max-Di-Cut?Joint work with Norman Huang.

### Short Paths on the Voronoi Graph and the Closest Vector Problem with Preprocessing

Series
ACO Seminar
Time
Monday, March 31, 2014 - 16:05 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker