arXiv subjects

F. Petrov

Publications and source records attributed to F. Petrov.

11 recordsLinked to original sources

arXiv2026

CayleyPy-4: AI-Holography. Towards analogs of holographic string dualities for AI tasks

This is the fourth paper in the CayleyPy project, which applies AI methods to the exploration of large graphs. In this work, we suggest the existence of a new discrete version of holographic string dualities for this setup, and discuss their relevance to AI systems and mathematics. Many modern AI tasks -- such as those addressed by GPT-style language models or RL systems -- can be viewed as direct analogues of predicting particle trajectories on graphs. We investigate this problem for a large family of Cayley graphs, for which we show that surprisingly it admits a dual description in terms of discrete strings. We hypothesize that such dualities may extend to a range of AI systems where they can lead to more efficient computational approaches. In particular, string holographic images of states are proposed as natural candidates for data embeddings, motivated by the "complexity = volume" principle in AdS/CFT. For Cayley graphs of the symmetric group S_n, our results indicate that the corresponding dual objects are flat, planar polygons. The diameter of the graph is equal to the number of integer points inside the polygon scaled by n. Vertices of the graph can be mapped holographically to paths inside the polygon, and the usual graph distances correspond to the area under the paths, thus directly realising the "complexity = volume" paradigm. We also find evidence for continuous CFTs and dual strings in the large n limit. We confirm this picture and other aspects of the duality in a large initial set of examples. We also present new datasets (obtained by a combination of ML and conventional tools) which should be instrumental in establishing the duality for more general cases.

hep-th↗

arXiv2025

CayleyPy RL: Pathfinding and Reinforcement Learning on Cayley Graphs

This paper is the second in a series of studies on developing efficient artificial intelligence-based approaches to pathfinding on extremely large graphs (e.g. $10^{70}$ nodes) with a focus on Cayley graphs and mathematical applications. The open-source CayleyPy project is a central component of our research. The present paper proposes a novel combination of a reinforcement learning approach with a more direct diffusion distance approach from the first paper. Our analysis includes benchmarking various choices for the key building blocks of the approach: architectures of the neural network, generators for the random walks and beam search pathfinding. We compared these methods against the classical computer algebra system GAP, demonstrating that they "overcome the GAP" for the considered examples. As a particular mathematical application we examine the Cayley graph of the symmetric group with cyclic shift and transposition generators. We provide strong support for the OEIS-A186783 conjecture that the diameter is equal to n(n-1)/2 by machine learning and mathematical methods. We identify the conjectured longest element and generate its decomposition of the desired length. We prove a diameter lower bound of n(n-1)/2-n/2 and an upper bound of n(n-1)/2+ 3n by presenting the algorithm with given complexity. We also present several conjectures motivated by numerical experiments, including observations on the central limit phenomenon (with growth approximated by a Gumbel distribution), the uniform distribution for the spectrum of the graph, and a numerical study of sorting networks. To stimulate crowdsourcing activity, we create challenges on the Kaggle platform and invite contributions to improve and benchmark approaches on Cayley graph pathfinding and other tasks.

cs.LG↗

arXiv2023

Tight lower bound on $|A+λA|$ for algebraic integer $λ$

We prove an asymptotically tight lower bound on $|A+λA|$ for $A\subset \mathbb{C}$ and algebraic integer $λ$. The proof combines strong version of Freiman's theorem, structural theorem on dense subsets of a hypercubic lattice and a generalisation of the continuous result on tight bound for the measure of $K+τK$ for a compact subset $K\subset \mathbb{R}^d$ of unit Lebesgue measure and a fixed linear operator $τ\colon\mathbb{R}^d\to \mathbb{R}^d$, obtained in our previous work.

math.CO↗

arXiv2023

Limit spectral measures of matrix distributions of metric triples

A notion of the limit spectral measure of a metric triple (i.e., a metric measure space) is defined. If the metric is square integrable, then the limit spectral measure is deterministic and coinsides with the spectrum of the integral operator in $L^2(μ)$ with kernel $ρ$. We construct an example in which there is no deterministic spectral measure.

math.RT↗

arXiv2022

Central Measures of Continuous Graded Graphs:\\ the Case of Distinct Frequencies

We define a class of continuous graded graphs similar to the graph of Gelfand--Tsetlin patterns, and describe the set of all ergodic central measures of discrete type on the path spaces of such graphs. The main observation is that an ergodic central measure on a subgraph of a Pascal-type graph can often be obtained as the restriction of the standard Bernoulli measure to the path space of the subgraph. This observation dramatically changes the approach to finding central measures also on discrete graphs, such as the famous Young graph. The simplest example of this type is given by the theorem on the weak limits of normalized Lebesgue measures on simplices; these are the so-called Cesàro measures, which are concentrated on the sequences with prescribed Cesàro limits (this limit parametrizes the corresponding measure). More complicated examples are the graphs of continuous Young diagrams with fixed number of rows and the graphs of spectra of infinite Hermitian matrices of finite rank. We prove existence and uniqueness theorems for ergodic central measures and describe their structure. In particular, our results 1) give a new spectral description of the so-called infinite-dimensional Wishart measures~\cite{W}~ -- ergodic unitarily invariant measures of discrete type on the set of infinite Hermitian matrices; 2) describe the structure of continuous analogs of measures on discrete graded graphs. New problems and connections which appear are to be considered in new publications.

math.CO↗

arXiv2022

Asymptotics of Landau--Okhotin function

Landau function $g(n)$ is the maximal possible least common multiple of several positive integers with sum not exceeding $n$. Under additional assumptions that these numbers are the differences of disjoint bi-infinite arithmetic progressions the maximum is denoted $\tilde{g}(n)$, it was introduced by Okhotin. We find a sharp logarithmic asymptotics of $\tilde{g}(n)$.

math.NT↗

arXiv2018

On a question of Sidorenko

For a positive integer $n>1$ denote by $ω(n)$ the maximal possible number $k$ of different functions $f_1,\dots,f_k:\mathbb{Z}/n\mathbb{Z}\mapsto \mathbb{Z}/n\mathbb{Z}$ such that each function $f_i-f_j,i<j$, is bijective. Recently A. Sidorenko conjectured that $ω(n)$ equals to the minimal prime divisor of $n$. We disprove it for $n=15,21,27$ by several counterexamples found by computer.

math.CO↗

arXiv2016

Combinatorics of the Lipschitz polytope

Let $ρ$ be a metric on the set $X=\{1,2,\dots,n+1\}$. Consider the $n$-dimensional polytope of functions $f:X\rightarrow \mathbb{R}$, which satisfy the conditions $f(n+1)=0$, $|f(x)-f(y)|\leq ρ(x,y)$. The question on classifying metrics depending on the combinatorics of this polytope have been recently posed by A. M. Vershik \cite{V}. We prove that for any "generic" metric the number of $(n-m)$-dimensional faces, $0\leq m\leq n$, equals $\binom{n+m}{m,m,n-m}=(n+m)!/m!m!(n-m)!$. This fact is intimately related to regular triangulations of the root polytope (the convex hull of the roots of $A_n$ root system). Also we get two-sided estimates for the logarithm of the number of Vershik classes of metrics: $n^3\log n$ from above and $n^2$ from below.

math.CO↗

arXiv2014

Virtual Continuity of Measurable Functions and Its Applications

Classical theorem of Luzin states that a measurable function of one real variable is "almost" continuous. For measurable functions of several variables the analogous statement (continuity on the product of sets having almost full measure) does not hold in general. Searching for a right analogue of Luzin theorem leads to a notion of virtually continuous functions of several variables. This probably new notion implicitly appears in the statements of embedding theorems and trace theorems for Sobolev spaces. In fact it reveals the nature of such theorems as statements about virtual continuity. Our results imply that under conditions of Sobolev theorems there is a well-defined integration of a function over wide class of singular measures, including the measures concentrated on submanifolds. The notion of virtual continuity is used also for the classification of measurable functions of several variables and in some questions on dynamical systems, theory of polymorphisms and bistochastic measures. In this paper we recall necessary definitions and properties of admissible metrics, give several definitions of virtual continuity and discuss some applications. Revised version (without the proofs) is published in \cite{VZPFA}.

math.FA↗

arXiv2012

Geometry and Dynamics of Admissible Metrics in Measure Spaces

We study a wide class of metrics in a Lebesgue space with a standard measure, the class of so-called admissible metrics. We consider the cone of admissible metrics, introduce a special norm in it, prove compactness criteria, define the "-entropy of a measure space with an admissible metric, etc. These notions and related results are applied to the theory of transformations with invariant measure; namely, we study the asymptotic properties of orbits in the cone of admissible metrics with respect to a given transformation or a group of transformations. The main result of this paper is a new discreteness criterion for the spectrum of an ergodic transformation: we prove that the spectrum is discrete if and only if the "-entropy of the averages of some (and hence any) admissible metric over fragments of its trajectory is uniformly bounded.

math.DS↗

arXiv2002

How to sum up triangles

We prove configuration theorems that generalize the Desargues, Pascal, and Pappus theorems. Our generalization of the Desargues theorem allows us to introduce the structure of an Abelian group on the (properly extended) set of triangles which are perspective from a point. In barycentric coordinates, the corresponding group operation becomes the addition in R^3.

math.AG↗