Deformed Statistics Formulation of the Information Bottleneck Method

11/19/2008
by   R. C. Venkatesan, et al.
0

The theoretical basis for a candidate variational principle for the information bottleneck (IB) method is formulated within the ambit of the generalized nonadditive statistics of Tsallis. Given a nonadditivity parameter q , the role of the additive duality of nonadditive statistics ( q^*=2-q ) in relating Tsallis entropies for ranges of the nonadditivity parameter q < 1 and q > 1 is described. Defining X , X̃ , and Y to be the source alphabet, the compressed reproduction alphabet, and, the relevance variable respectively, it is demonstrated that minimization of a generalized IB (gIB) Lagrangian defined in terms of the nonadditivity parameter q^* self-consistently yields the nonadditive effective distortion measure to be the q -deformed generalized Kullback-Leibler divergence: D_K-L^q[p(Y|X)||p(Y|X̃)] . This result is achieved without enforcing any a-priori assumptions. Next, it is proven that the q^*-deformed nonadditive free energy of the system is non-negative and convex. Finally, the update equations for the gIB method are derived. These results generalize critical features of the IB method to the case of Tsallis statistics.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
10/23/2019

Variational Predictive Information Bottleneck

In classic papers, Zellner demonstrated that Bayesian inference could be...
research
11/14/2018

Total Positivity of A Kind of Generalized Toric-Bernstein Basis

The normalized totally positive basis is a set of basis functions define...
research
05/17/2012

Free Energy and the Generalized Optimality Equations for Sequential Decision Making

The free energy functional has recently been proposed as a variational p...
research
04/22/2022

The Lack of Convexity of the Relevance-Compression Function

In this paper we investigate the convexity of the relevance-compression ...
research
12/14/2020

Disentangled Information Bottleneck

The information bottleneck (IB) method is a technique for extracting inf...
research
12/17/2021

Online Generalized Additive Model

Additive models and generalized additive models are effective semiparame...
research
05/04/2019

Projection Theorems, Estimating Equations, and Power-Law Distributions

Projection theorems of divergence functionals reduce certain estimation ...

Please sign up or login with your details

Forgot password? Click here to reset