Conjectures hold a special status in mathematics. Good conjectures epitomise milestones in mathematical discovery, and have historically inspired new mathematics and shaped progress in theoretical physics. Hilbert’s list of 23 problems and André Weil’s conjectures oversaw major developments in mathematics for decades. Crafting conjectures can often be understood as a problem in pattern recognition, for which Machine Learning is tailor-made. In this talk, I will propose a framework that allows a principled study of a space of mathematical conjectures. Using this framework and exploiting domain knowledge and machine learning, we generate a number of conjectures in number theory and group theory. I will present evidence in support of some of the resulting conjectures and present a new theorem. I will lay out a vision for this endeavour, and conclude by posing some general questions about the pipeline.
Imagine three gamblers with respectively $A$, $B$, $C$ at the start. Each time, a pair of gamblers are chosen (uniformly at random) and a fair coin is flipped. Of course, eventually, one of the gamblers is eliminated and the game continues with the remaining two until one winds up with all $A+B+C$. In poker tournaments (really) it is of interest to know the chances of the six possible elimination orders (e.g. $3,1,2$ means gambler $3$ is eliminated first, then gambler 1, leaving 2 with all the cash). In particular, how do these depend on $A,B,C$? For small $A,B,C$, exact computation is possible, but for $A,B,C$ of practical interest, asymptotics are needed. The math involved is surprising; Whitney and John domains, Carlesson estimates. To test your intuition, recall that if there are two gamblers with $1$ and $N-1$ the chance that the first winds up with all $N$ is $1/N$. With three gamblers with $1,1$ and $N-2$ the chance that the third is eliminated first is $\frac{\operatorname{Const}}{N^3}$. We don't know the answer for four gamblers. This is a report of joint work with Stew Ethier, Kelsey Huston-Edwards and Laurent Saloff-Coste.
Spectral networks are a combinatorial tool consisting of labelled lines on a Riemann surface. They have a surprising amount of applications and are intimately linked to non-Abelianization of flat connections, Fock–Goncharov cluster coordinates, exact WKB theory, etc. After reviewing this story for the $SL(2)$ and $SL(3)$ case, I will describe this is in detail for the group $G_2$. Time permitting, I will give as an application a concrete parametrization of the nonabelian Hodge correspondence for the Hitchin component of the split real form of $G_2$. This is joint work with Andy Neitzke.
Superconformal ‘type B’ quantum mechanical sigma models arise in a variety of interesting contexts, such as the description of D-brane bound states in an $AdS_2$ decoupling limit. Focusing on $N= 2B$ models, we study superconformal indices which count short multiplets and provide an alternative to the standard Witten index, as the latter suffers from infrared issues. We show that the basic index receives contributions from lowest Landau level states in an effective magnetic field and that, due to the noncompactness of the target space, it is typically divergent. Fortunately, the models of interest possess an additional target space isometry which allows for the definition of a well-behaved refined index. We compute this index using localization of the functional integral and find that the result agrees with a naive application of the Atiyah-Bott fixed point formula outside of it’s starting assumptions. In the simplest examples, this formula can also be directly verified by explicitly computing the short multiplet spectrum.
This will be a survey talk about recent progress on norm and pointwise convergence problems for classical and multiple ergodic averages along polynomial orbits. A celebrated theorem of Szemeredi asserts that every subset of integers with nonvanishing upper Banach density contains arbitrarily long arithmetic progressions. We will discuss the significance of using ergodic theory and Fourier analysis in solving this problem. We will also explain how this problem led to the conjecture of Furstenberg-Bergelson-Leibman, which is a major open problem in pointwise ergodic theory. Relations with number theory and additive combinatorics will be also discussed.
I will describe a map that associates to every deformation of an object in a higher category a collection of generalized symmetries of the object. Building on work by Lurie, we will see that the failure of this map to be an equivalence can be quantified. Under favorable circumstances, the map is an equivalence, and this leads to an explicit description of the space of deformations in terms of solutions to certain equations. I will discuss applications of these results to topological field theory and holomorphic symplectic geometry. This talk is based on joint work with Bhanu Kiran.
It has been recently suggested that the strong Emergence Proposal is realized in equi-dimensional M-theory limits by integrating out all light towers of states with a typical mass scale not larger than the species scale, i.e the eleventh dimensional Planck mass. Within the BPS sector, these are transverse M2- and M5-branes, that can be wrapped and particle-like, carrying Kaluza-Klein momentum along the compact directions. We provide additional evidence for this picture by revisiting and investigating further the computation of $R^4$-interactions in M-theory à la Green-Gutperle-Vanhove. A central aspect is a novel UV-regularization of Schwinger-like integrals, whose actual meaning and power we clarify by first applying it to string perturbation theory. We consider then toroidal compactifications of M-theory and provide evidence that integrating out all light towers of states via Schwinger-like integrals thus regularized yields the complete result for $R^4$-interactions. In particular, this includes terms that are tree-level, one-loop and space-time instanton corrections from the weakly coupled point of view. Finally, we comment on the conceptual difference of our approach to earlier closely related work by Kiritsis-Pioline and Obers-Pioline, and conjecture a correspondence between two types of constrained Eisenstein series.
The classical Kantorovich-Rubinstein duality theorem established a significant connection between Monge optimal transport and the maximization of a linear form on 1-Lipschitz functions. This result has been widely used in various research areas, particularly to demonstrate a bridge between Monge transport theory and some class of optimal design problems in mechanics.
The aim of this talk is to present a similar theory when the linear form is maximized over all real $C^{1,1}$ functions with a Hessian matrix spectral norm not exceeding one. It turns out that this new maximization problem can be viewed as the dual of a specific optimal transport problem. The task is to find a minimal three-point plan with given first two marginals, where the third is assigned to be larger than both in the sense of convex order. The existence of optimal plans allows to express solutions of the underlying Beckman problem as a combination of rank-one tensor measures supported by a graph. In the context of two-dimensional mechanics, this graph encodes the optimal location of a grillage to support a given bending load.
The maximum likelihood problem for Hidden Markov Models is usually numerically solved by the Baum-Welch algorithm, which uses the Expectation-Maximization algorithm to find the estimates of the parameters. This algorithm has a recursion depth equal to the data sample size and cannot be computed in parallel, which limits the use of modern GPUs to speed up computation time. A new algorithm is proposed that provides the same estimates as the Baum-Welch algorithm, requiring about the same number of iterations, but is designed in such a way that it can be parallelized. As a consequence, it leads to a significant reduction in the computation time. We illustrate this by means of numerical examples, where we consider simulated data as well as real datasets.
The Dyson Brownian motion (DMB) is a system of infinitely many interacting Brownian motions with logarithmic interaction potential, which was introduced by Freeman Dyson '62 in relation to the random matrix theory. In this talk, we reveal that an infinite-dimensional differential structure induced by the DBM has a Bakry-Émery lower Ricci curvature bound. As an application, we show that the DBM can be realised as the unique Wasserstein-type gradient flow of the Boltzmann-Shannon entropy associated with $\operatorname{Sine}_\beta$ ensemble.
In the last two decades, Craig interpolation has emerged as an essential tool in formal verification, where first-order theories are used to express constraints on the system, such as on the datatypes manipulated by programs. Interpolants for such theories are largely exploited as an efficient method to approximate the reachable states of the system and for invariant synthesis. In this talk, we report recent results on a stronger form of interpolation, called uniform interpolation, and its connection with the notion of model completion from model-theoretic algebra. We discuss how uniform interpolants can be used in the context of formal verification of infinite-state systems to develop effective techniques for computing the reachable states in an exact way. Finally, we present some results about the transfer of uniform interpolants to theory combinations. We argue that methods based on uniform interpolation are particularly effective and computationally efficient when applied to verification of the so-called data-aware processes: these are systems where the control flow of a process can interact with a data storage.
We study transitions from chaotic to integrable Hamiltonians in the double scaled SYK and $p$-spin systems. The dynamics of our models is described by chord diagrams with two species. We begin by developing a path integral formalism of coarse graining chord diagrams with a single species of chords, which has the same equations of motion as the bi-local Liouville action, yet appears otherwise to be different and in particular well defined. We then develop a similar formalism for two types of chords, allowing us to study different types of deformations of double scaled SYK and in particular a deformation by an integrable Hamiltonian. The system has two distinct thermodynamic phases: one is continuously connected to the chaotic SYK Hamiltonian, the other is continuously connected to the integrable Hamiltonian, separated at low temperature by a first order phase transition.
In recent years, there has been a growing number of applications of stable homotopy theory to condensed matter physics, many of which stem from a conjecture of Kitaev that gapped invertible phases of matter should be classified by the homotopy groups of a spectrum. This gives rise to a mathematical modeling question: how do we model quantum systems in such a way that this result can be better understood, perhaps even proved? In this talk, I will discuss some aspects of this modeling problem. This is based on joint work with Mike Hermele, Juan Moreno, Markus Pflaum, Marvin Qi and Daniel Spiegel, David Stephen, Xueda Wen.
