We provide a probabilistic characterization of the class of probability measures that can be represented by the Matrix Product Ansatz (MPA). We describe a constructive procedure, based on a suitable enlargement of the state space, showing that a probability measure can be expressed in terms of non-negative matrices via the MPA if and only if it can be written as a mixture of inhomogeneous product measures, where the mixing law is given by a Markov bridge. We illustrate this construction by applying it to examples of interacting particle systems. Finally, we discuss how the resulting probabilistic structure can be exploited to obtain large deviation principles for this class of measures. Joint work with Davide Gabrielli.
On a background Minkowski spacetime, the Euler equations (both relativistic and not) are known to admit unstable homogeneous solutions with finite-time shock formation. Such shock formation can be suppressed on cosmological spacetimes whose spatial slices expand at an accelerated rate. However, situations with decelerated expansion, which are relevant in our early universe, are not as well understood. I will present some recent joint work in this direction, based on collaborations with David Fajman, Maciej Maliborski, Todd Oliynyk and Max Ofner.
In this talk, we consider the facilitated exclusion process on the one-dimensional discrete $N$-torus. Because of the facilitating mechanism, the process freezes in finite time if the particle density of the initial configuration is subcritical, i.e., if it is smaller than (or equal to) 1/2. We prove that, starting from any subcritical Bernoulli product measure, the correct scale of the transience/freezing time is of order $\log^3(N)$. Based on a joint work with Oriane Blondel, Clément Erignoux and Sanha Lee.
I will first introduce the bigraded cohomology for real algebraic varieties developed by Johannes Huisman and Dewi Gleuher. This is a cohomology theory that refines the equivariant cohomology "à la Kahn-Krasnov" of the complex points of a real variety, the latter often being preferred (by the algebraic geometers) in the cohomological study of real algebraic varieties. Since the construction of this bigraded cohomology and its associated characteristic classes relies on the sheaf exponential morphism, I will explain how to produce an arithmetic (or algebraic) variant of these cohomology groups, whose main advantage is toeliminate topological or transcendental conditions. I will conclude by comparing these two versions of bigraded cohomology.
Existing machine learning frameworks operate over the field of real numbers ($\mathbb{R}$ ) and learn representations in real (Euclidean or Hilbert) vector spaces (e.g., $\mathbb{R}^d$ ). Their underlying geometric properties align well with intuitive concepts such as linear separability, minimum enclosing balls, and subspace projection; and basic calculus provides a toolbox for learning through gradient-based optimization.
But is this the only possible choice? In this seminar, we study the suitability of a radically different field as an alternative to $\mathbb{R}$ — the ultrametric and non-archimedean space of $p$-adic numbers, $\mathbb{Q}_p$. The hierarchical structure of the $p$-adics and their interpretation as infinite strings make them an appealing tool for code theory and hierarchical representation learning. Our exploratory theoretical work establishes the building blocks for classification, regression, and representation learning with the $p$-adics, providing learning models and algorithms. We illustrate how simple Quillian semantic networks can be represented as a compact $p$-adic linear network, a construction which is not possible with the field of reals. We finish by discussing open problems and opportunities for future research enabled by this new framework.
Existing machine learning frameworks operate over the field of real numbers ($\mathbb{R}$ ) and learn representations in real (Euclidean or Hilbert) vector spaces (e.g., $\mathbb{R}^d$ ). Their underlying geometric properties align well with intuitive concepts such as linear separability, minimum enclosing balls, and subspace projection; and basic calculus provides a toolbox for learning through gradient-based optimization.
But is this the only possible choice? In this seminar, we study the suitability of a radically different field as an alternative to $\mathbb{R}$ — the ultrametric and non-archimedean space of $p$-adic numbers, $\mathbb{Q}_p$. The hierarchical structure of the $p$-adics and their interpretation as infinite strings make them an appealing tool for code theory and hierarchical representation learning. Our exploratory theoretical work establishes the building blocks for classification, regression, and representation learning with the $p$-adics, providing learning models and algorithms. We illustrate how simple Quillian semantic networks can be represented as a compact $p$-adic linear network, a construction which is not possible with the field of reals. We finish by discussing open problems and opportunities for future research enabled by this new framework.
The Malmquist-Takenaka (MT) system is a complete orthonormal system in H^2(T) generated by an arbitrary sequence of points in the unit disk that do not approach the boundary very fast. The nth point of the sequence is responsible for multiplying the nth and subsequent terms of the system by a Möbius transform taking the point to 0. One can recover the classical trigonometric system, its perturbations or conformal transformations, as particular examples of the MT system. However, many interesting choices of the sequence an, the MT system is less understood. We prove almost everywhere convergence of the MT series for three different classes of generating sequences.
