PCA meets RG

DOI10.1007/S10955-017-1770-6MaRDI QIDQ2401447zbMATH OpenOpenAlexWikidataFDO

Authors Serena Bradde, William Bialek

Publication date 8 September 2017

Published in Journal of Statistical Physics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1610.09733, http://europepmc.org/articles/pmc6054449

zbMATH Keywords

financial marketsneural networksprincipal component analysisrenormalization group

Mathematics Subject Classification ID

Renormalization group methods in equilibrium statistical mechanics (82B28) Neural nets applied to problems in time-dependent statistical mechanics (82C32) Macroeconomic theory (monetary models, models of taxation) (91B64) Neural networks for/in biological studies, artificial life and related topics (92B20)

Abstract: A system with many degrees of freedom can be characterized by a covariance matrix; principal components analysis (PCA) focuses on the eigenvalues of this matrix, hoping to find a lower dimensional description. But when the spectrum is nearly continuous, any distinction between components that we keep and those that we ignore becomes arbitrary; it then is natural to ask what happens as we vary this arbitrary cutoff. We argue that this problem is analogous to the momentum shell renormalization group (RG). Following this analogy, we can define relevant and irrelevant operators, where the role of dimensionality is played by properties of the eigenvalue density. These results also suggest an approach to the analysis of real data. As an example, we study neural activity in the vertebrate retina as it responds to naturalistic movies, and find evidence of behavior controlled by a nontrivial fixed point. Applied to financial data, our analysis separates modes dominated by sampling noise from a smaller but still macroscopic number of modes described by a non--Gaussian distribution.

Review text

There is an investigation on methods to find a lower-dimensional description for systems with many degrees of freedom. Actually this means to consider somehow a linear projection on a lower-dimensional space of higher-dimensional data. In this paper, the connection between two known concepts, the principal component analysis method (PAC), and the renormalization group method (RG), is largely discussed. After a theoretical description of the methods, the analysis of data in two practical examples from neural networks and financial systems are performed.

Recommendations

• Some implications of renormalization group theoretical ideas to statistics
• Principal Components and Principal Clusters
• scientific article; zbMATH DE number 558458
• Reduced principal component analysis
• Inferring relevant features: from QFT to PCA

Cites work

• Analysis of a population genetics model with mutation, selection, and pleiotropy
• Are biological systems poised at criticality?
• Cleaning large correlation matrices: tools from random matrix theory
• Dissecting financial markets: sectors and states
• Finite Euclidean graphs and Ramanujan graphs
• Fractal measures and their singularities: The characterization of strange sets
• scientific article; zbMATH DE number 596309 (Why is no real title available?)
• scientific article; zbMATH DE number 2232122 (Why is no real title available?)
• More is the same; phase transitions and mean field theories
• Neural networks and physical systems with emergent collective computational abilities
• Reflections on Gibbs: from statistical physics to the amistad V3.0
• Relating theories via renormalization
• Rotational Invariant Estimator for General Noisy Matrices
• Singularities in complex interfaces
• Theory of Financial Risk and Derivative Pricing
• Viscous flows in two dimensions

Cited in

(12)

• Principal component analysis of \(1/f^{\alpha}\) noise
• Constructive expansion for vector field theories I. Quartic models in low dimensions
• Inferring relevant features: from QFT to PCA
• On Koopman mode decomposition and tensor component analysis
• Generalized scale behavior and renormalization group for data analysis
• Correlations in the shear flow of athermal amorphous solids: a principal component analysis
• The higher-order spectrum of simplicial complexes: a renormalization group approach
• An intriguing connection between Pisarski's fixed point and (2 + 3)-spin glasses
• Interpreting deep learning by establishing a rigorous corresponding relationship with the renormalization group on the Ising model
• Large time effective kinetics -functions for quantum (2+p)-spin glass
• A functional renormalization group for signal detection and stochastic ergodicity breaking
• Functional renormalisation for signal detection: dimensional analysis and dimensional phase transition for nearly continuous spectra effective field theory

This page was built for publication: PCA meets RG

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2401447)