Probability and Mathematical Physics Seminar
This seminar covers a wide range of topics in pure and applied probability and in mathematical physics. Unless otherwise noted, the talks take place on Fridays, 11:10am–12pm, in room WWH 1302. See directions to the Courant Institute.
The seminar is run by Yuri Bakhtin, Gérard Ben Arous, Paul Bourgade, Percy Deift, Ruojun Huang, Eyal Lubetzky, Henry P. McKean, Chuck Newman, JeanChristophe Mourrat, Michel Pain, Yuval Peled, S. R. Srinivasa Varadhan, Ofer Zeitouni. To be kept informed, you can subscribe to the Probability Seminar Mailing list.
Seminar Organizer(s): Probabilists
Past Events

Friday, December 14, 201811AM, Warren Weaver Hall 512
The Dynamics of Foraging and Starvation
Sid Redner, Santa Fe InstituteSynopsis:
What is the fate of a randomwalk forager that depletes its environment as it wanders? Whenever the forager lands on a foodcontaining site, all the food is consumed and the forager becomes fully sated. However, when the forager lands on an empty site, it moves one time unit closer to starvation. If the forager wanders S steps without encountering food, it starves to death. We show that the lifetime of this starving random walk forager scales linearly with S in one dimension by solving an underlying nonMarkovian firstpassage problem. In greater than two dimensions, we present evidence that the lifetime grows quasiexponentially in S.
We also investigate the role of greed, in which the forager preferentially moves towards food when faced with a choice of hopping to food or to an empty site in its local neighborhood. Paradoxically, the forager lifetime can have a nonmonotonic dependence on greed, with different senses to the nonmonotonicity in one and in two dimensions. In one dimension, the forager lifetime exhibits a huge peak when greed is negative, while in two dimensions the maximum lifetime occurs for positive, but not perfect, greed.
Finally, we briefly discuss the role of frugality and myopia on foraging dynamics. Frugality means that the forager does not eat until it is nutritionally depleted beyond a specified level. Myopia means that the forager sometime does not "see" food at its current site and leaves the food undisturbed.

Friday, December 7, 201811AM, Warren Weaver Hall 512
The smoothed KPZ equation in dimension three and higher: EdwardsWilkinson regime of its fluctuations and its localization properties
Chiranjib Mukherjee, Universität MünsterSynopsis:
We study the KardarParisiZhang equation in dimension \(d\geq 3\) with spacetime white noise which is smoothed in space. There is a natural disorder parameter attached to this equation which measures the intensity of the noise. We show that when the disorder is small, the approximating solution converges to a welldefined limit (with the limit depending on both the disorder and the mollification procedure), while the rescaled fluctuations converge to a Gaussian limit as predicted by the EdwardsWilkionson regime. We also study the associated stochastic heat equation with multiplicative noise, which carries a natural Gaussian mutiplicative noise (GMC) on the Wiener space. When the disorder is large, we also show that the total mass of the GMC converges to zero, while the endpoint distribution of a Brownian path under the (renormlaized) GMC measure is purely atomic. Based on joint works with, A. Shamov & O. Zeitouni, F. Comets & C. Cosco as well Y. Broeker.

Friday, November 30, 20189:30AM, Warren Weaver Hall 512
Columbia/Courant probability seminar
Various Speakers 
Friday, November 16, 20189:30AM, Warren Weaver Hall 109
Northeast Probability Seminar, Day 2
Various Speakers 
Thursday, November 15, 20189AM, Kimmel Center 912/914
Northeast Probability Seminar, Day 1
Various Speakers 
Friday, November 9, 201811AM, Warren Weaver Hall 512
The Archimedean limit of random sorting networks
Duncan Dauvergne, U. TorontoSynopsis:
Consider a list of n particles labelled in increasing order. A sorting network is a way of sorting this list into decreasing order by swapping adjacent particles, using as few swaps as possible. Simulations of largen uniform random sorting networks reveal a surprising and beautiful global structure involving sinusoidal particle trajectories, a semicircle law, and a theorem of Archimedes.
Based on these simulations, Angel, Holroyd, Romik, and Virag made a series of conjectures about the limiting behaviour of sorting networks. In this talk, I will discuss how to use the local structure of random sorting networks to prove these conjectures.

Friday, November 2, 201811AM, Warren Weaver Hall 512
The algorithmic hardness threshold for continuous random energy models
Pascal Maillard, CRM (Montréal) and Université ParisSudSynopsis:
I will report on recent work with Louigi AddarioBerry on algorithmic hardness for finding lowenergy states in the continuous random energy model of Bovier and Kurkova. This model can be regarded as a toy model for strongly correlated random energy landscapes such as the SherringtonKirkpatrick model. We exhibit a precise and explicit hardness threshold: finding states of energy above the threshold can be done in linear time, while below the threshold this takes exponential time for any algorithm with high probability. I further discuss what insights this yields for understanding algorithmic hardness thresholds for random instances of combinatorial optimization problems.

Friday, October 26, 201811AM, Warren Weaver Hall 512
Distributional approximation of the characteristic polynomial of a Gaussian betaensemble
Elliot Paquette, Ohio StateSynopsis:
The characteristic polynomial of the Gaussian betaensemble can be represented, via its tridiagonal model, as an entry in a product of independent random twobytwo matrices. For a point z in the complex plane, at which the transfer matrix is to be evaluated, this product of transfer matrices splits into three independent factors, each of which can be understood as a different dynamical system in the complex plane. Conjecturally, we show that the characteristic polynomial is always represented as product of at most three terms, an exponential of a Gaussian field, the stochastic Airy function, and a diffusion similar to the stochastic sine equation.
We explain the origins of this decomposition, and we show partial progress in establishing part of it.
Joint work with Diane Holcomb and Gaultier Lambert.

Friday, October 19, 201811AM, Warren Weaver Hall 512
Sharp transition of invertibility of sparse random matrices
Anirban Basak, ICTSTIFRSynopsis:
Consider an \(n \times n\) matrix with i.i.d.~Bernoulli(\(p\)) entries. It is well known that for \(p= \Omega(1)\), i.e.~\(p\) is bounded below by some positive constant, the matrix is invertible with high probability. If \(p \ll \frac{\log n}{n}\) then the matrix contains zero rows and columns with high probability and hence it is singular with high probability. In this talk, we will discuss the sharp transition of the invertibility of this matrix at \(p =\frac{\log n}{n}\). This phenomenon extends to the adjacency matrices of directed and undirected Erdős–Rényi graphs, and random bipartite graph.
Joint work with Mark Rudelson.

Friday, October 12, 201810:15AM, Warren Weaver Hall 512
Estimating graph parameters via multiple random walks
Roberto Oliveira, IMPASynopsis:
What can one say about a graph from multiple (short) random walk trajectories on it? In this talk we consider algorithms that only "see" walk trajectories and the degrees along the way.
We will show that the number of vertices, edges and mixing time can be all estimated with a number of RW steps that is sublinear in the size of the graph and in its mixing or relaxation time. Our bounds on the number of RW steps are optimal up to constant or polylog factors. We also argue that such algorithms cannot "know when to stop", and discuss additional conditions that circumvent this limitation. To prove our results, we rely on bounds for random walk intersections by Peres, Sauerwald, Sousi and Stauffer.
Joint with Anna BenHamou (Sorbonne) and Yuval Peres (MSR).

Friday, October 5, 201811AM, Warren Weaver Hall 512
Short time asymptotics for Liouviille heat kernel and graph distances
Ofer Zeitouni, Courant Institute/Weizmann InstituteSynopsis:
The Liouville Brownian motion is Brownian motion in 2D, time changed by the exponential of the Gaussian free field along its trajectory. In recent work with Jian Ding and Fuxi Zhang, we established the existence of scaling exponent for the (logarithmic) short time asymptotics of the heat kernel, and linked the exponent to the scaling exponent of the graph distance for Liouville quantum gravity., which we also proved exists. I will describe this and some related developments.

Friday, September 28, 201811AM, Warren Weaver Hall 512
Rigidity and tolerance for perturbed lattices
Yuval Peres, Microsoft ResearchSynopsis:
Consider a perturbed lattice \(v+Y_v\) obtained by adding IID \(d\)dimensional Gaussian variables \(Y_v\) to the lattice points in \(Z^d\). Suppose that one point, say \(Y_0\), is removed from this perturbed lattice; is it possible for an observer, who sees just the remaining points, to detect that a point is missing? In one and two dimensions, the answer is positive: the two point processes (before and after \(Y_0\) is removed) can be distinguished using smooth statistics, analogously to work of Sodin and Tsirelson (2004) on zeros of Gaussian analytic functions. (cf. Holroyd and Soo (2011)). The situation in higher dimensions is more delicate; our solution depends on a gametheoretic idea, in one direction, and on the unpredictable paths constructed by Benjamini, Pemantle and the speaker (1998), in the other. (Joint work with Allan Sly).

Friday, September 21, 201811AM, Warren Weaver Hall 512
Optimization via Langevin dynamics in complex landscapes
Reza Gheissari, NYUSynopsis:
Consider the problem of recovering \(v \in \mathbb S^{N1}\) from an i.i.d. Gaussian \(k\)tensor spiked by the rankone tensor \(v^{\otimes k}\). The likelihood function for this is an example of a complex highdimensional landscape. It is informationtheoretically possible to detect the spike at order 1 signaltonoise ratios, but if \(k>2\) it is expected that the signaltonoise ratio needs to diverge in \(N\) to efficiently recover \(v\). We seek to understand the mechanisms for, and obstructions to, such planted signal recovery in high dimensions, via locally optimizing algorithms. We study recovery thresholds for a family of "vanilla" algorithms known as Langevin dynamics, on general landscapes consisting of a planted nonlinear signal function perturbed by isotropic Gaussian noise. We propose a mechanism for success/failure of recovery via such algorithms in terms of the curvature of the planted signal.
Joint work with G. Ben Arous and A. Jagannath.

Friday, September 14, 201811AM, Warren Weaver Hall 512
First order phase transition for the Potts and randomcluster model with q>4
Matan Harel, Tel Aviv UniversitySynopsis:
The randomcluster model is a dependent percolation model where the weight of a configuration is proportional to q to the power of the number of connected components. It is highly related to the ferromagnetic qPotts model, where every vertex is assigned one of q colors, and monochromatic neighbors are encouraged. Through nonrigorous means, Baxter showed that the phase transition is firstorder whenever \(q>4\)  i.e. there are multiple Gibbs measures at criticality. We provide a rigorous proof of this claim. Like Baxter, our proof uses the correspondence between the above models and the sixvertex model, which we analyze using the Bethe ansatz and transfer matrix techniques. We also prove Baxter's formula for the correlation length of the models at criticality. This is joint work with Hugo DuminilCopin, Maxime Gangebin, Ioan Manolescu, and Vincent Tassion.