Abstract:Graph Laplacians as well as related spectral inequalities and (co-)homology provide a foray into discrete analogues of Riemannian manifolds, providing a rich interplay between combinatorics, geometry and theoretical physics. We apply some of the latest techniques in data science such as supervised and unsupervised machine-learning and topological data analysis to the Wolfram database of some 8000 finite graphs in light of studying these correspondences. Encouragingly, we find that neural classifiers, regressor… Show more
“…In [48], MLPs, decision trees and graph NNs could distinguish table/non-table ideals to 100% accuracy, whereby suggesting the existence of a yet-unknown formula. Graphs & Combinatorics: Various properties of finite graphs, such as cyclicity, genus, existence of Euler or Hamilton cycles, etc., were explored by "looking" at the the adjacency matrix with MLPs and SVMs [49]. The algorithms determining some of these quantities are quite involved indeed.…”
We review the recent programme of using machine-learning to explore the landscape of mathematical problems. With this paradigm as a model for human intuition -complementary to and in contrast with the more formalistic approach of automated theorem proving -we highlight some experiments on how AI helps with conjecture formulation, pattern recognition and computation.
“…In [48], MLPs, decision trees and graph NNs could distinguish table/non-table ideals to 100% accuracy, whereby suggesting the existence of a yet-unknown formula. Graphs & Combinatorics: Various properties of finite graphs, such as cyclicity, genus, existence of Euler or Hamilton cycles, etc., were explored by "looking" at the the adjacency matrix with MLPs and SVMs [49]. The algorithms determining some of these quantities are quite involved indeed.…”
We review the recent programme of using machine-learning to explore the landscape of mathematical problems. With this paradigm as a model for human intuition -complementary to and in contrast with the more formalistic approach of automated theorem proving -we highlight some experiments on how AI helps with conjecture formulation, pattern recognition and computation.
“…The works [22,23] are in a sense part of a more general program of applying methods from data science and machine learning to study problems in mathematical physics, see [24,25] for a pedagogical treatment. By now, machine learning and data science have been used to learn various aspects of number theory [26,27,28], quiver gauge theories and cluster algebras [29], knot theory [30,31,32] as well as graph theory [33] and commutative algebra [34]. Importantly, and in hindsight somewhat as a precursor of [23], support vector machines and neural networks were used to learn the algebraic structures of discrete groups and rings [35].…”
“…In these, the effectiveness of ML regressor and classifier techniques in various branches of mathematics and mathematical physics has been investigated. Applications of ML include: finding bundle cohomology on varieties [22,24,25]; distinguishing elliptic fibrations [26] and invariants of Calabi-Yau threefolds [27]; the Donaldson algorithm for numerical Calabi-Yau metrics [28]; the algebraic structures of groups and rings [29]; arithmetic geometry and number theory [30][31][32]; quiver gauge theories and cluster algebras [33]; patterns in particle masses [34]; statistical predictions and model-building in string theory [35][36][37]; and classifying combinatorial properties of finite graphs [38].…”
We describe how simple machine learning methods successfully predict geometric properties from Hilbert series (HS). Regressors predict embedding weights in projective space to ∼1 mean absolute error, whilst classifiers predict dimension and Gorenstein index to > 90% accuracy with ∼0.5% standard error. Binary random forest classifiers managed to distinguish whether the underlying HS describes a complete intersection with high accuracies exceeding 95%. Neural networks (NNs) exhibited success identifying HS from a Gorenstein ring to the same order of accuracy, whilst generation of "fake" HS proved trivial for NNs to distinguish from those associated to the threedimensional Fano varieties considered.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.