Factorizing Multivariate Function Classes
نویسنده
چکیده
The mathematical framework for factorizing equivalence classes of multivariate functions is formulated in this paper. Independent component analysis is shown to be a special case of this decomposition. Using only the local geometric structure of a class representative, we derive an analytic solution for the factorization. We demonstrate the factorization solution with numerical experiments and present a preliminary tie to decorrelation.
منابع مشابه
Making problems tractable on big data via preprocessing with polylog-size output
To provide a dichotomy between those queries that can be made feasible on big data after appropriate preprocessing and those for which preprocessing does not help, Fan et al. developed the u-tractability theory. This theory provides a formal foundation for understanding the tractability of query classes in the context of big data. Along this line, we introduce a novel notion of u′-tractability ...
متن کاملThe Variational Gaussian Approximation Revisited
The variational approximation of posterior distributions by multivariate gaussians has been much less popular in the machine learning community compared to the corresponding approximation by factorizing distributions. This is for a good reason: the gaussian approximation is in general plagued by an Omicron(N)(2) number of variational parameters to be optimized, N being the number of random vari...
متن کاملOn some factorization problems
We consider three notions of factorization arising in different frameworks: factorizing languages, factorization of the natural numbers, factorizing codes. A language X ⊆ A∗ is called factorizing if there exists a language Y ⊆ A∗ such that XY = A∗ and the product is unambiguous. This is a decidable property for recognizable languages X . If we consider the particular case of unary alphabets, we...
متن کاملBayesian factorization of joint categorical distributions for relational data and classical conditioning models
We explore the problem of infering latent structure in the joint probability of categorical variables by factorizing them into a latent representation for categories and a weight matrix that encodes a PMF, mapping these latent representations to the mass associated with seeing categories appear together. The prior for latent category representations is either the Chinese Restaurant Process (CRP...
متن کامل