The dually flat structure for singular models

نویسندگان

چکیده

The dually flat structure introduced by Amari–Nagaoka is highlighted in information geometry and related fields. In practical applications, however, the underlying pseudo-Riemannian metric may often be degenerate, such an excellent geometric rarely defined on entire space. To fix this trouble, present paper, we propose a novel generalization of for certain class singular models from viewpoint Lagrange Legendre singularity theory—we introduce quasi-Hessian manifold endowed with possibly degenerate particular symmetric cubic tensor, which exceeds concept statistical manifolds adapted to theory (weak) contrast functions. particular, establish Amari–Nagaoka’s extended Pythagorean theorem projection general setup, consequently, most applications these theorems are suitably justified even cases. This work motivated various interests different backgrounds Frobenius mathematical physics Deep Learning data science.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

On dually flat Randers metrics

The notion of dually flat Finsler metrics arise from information geometry. In this paper, we will study a special class of Finsler metrics called Randers metrics to be dually flat. A simple characterization is provided and some non-trivial explicit examples are constructed. In particular, We will show that the dual flatness of a Randers metric always arises from that of some Riemannian metric b...

متن کامل

Edge selection based on the geometry of dually flat spaces for Gaussian graphical models

We propose a method for selecting edges in Gaussian graphical models. Our algorithm takes after our previous work, an extension of Least Angle Regression (LARS), and it is based on the information geometry of dually flat spaces. Non-diagonal elements of the inverse of the covariance matrix, the concentration matrix, play an important role in edge selection. Our iterative method estimates these ...

متن کامل

On dually flat general $(\alpha,\beta)$-metrics

Based on the previous research, in this paper we study the dual flatness of a special class of Finsler metrics called general (α, β)-metrics, which is defined by a Riemannian metric α and a 1-form β. By using a new kind of deformation technique, we construct many non-trivial explicit dually flat general (α, β)-metrics.

متن کامل

On dually flat $(\alpha,\beta)$-metrics

The dual flatness for Riemannian metrics in information geometry has been extended to Finsler metrics. The aim of this paper is to study the dual flatness of the so-called (α, β)-metrics in Finsler geometry. By doing some special deformations, we will show that the dual flatness of an (α, β)-metric always arises from that of some Riemannian metric in dimensional n ≥ 3.

متن کامل

Monte Carlo Information Geometry: The dually flat case

Exponential families and mixture families are parametric probability models that can be geometrically studied as smooth statistical manifolds with respect to any statistical divergence like the KullbackLeibler (KL) divergence or the Hellinger divergence. When equipping a statistical manifold with the KL divergence, the induced manifold structure is dually flat, and the KL divergence between dis...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Information geometry

سال: 2021

ISSN: ['2511-2481', '2511-249X']

DOI: https://doi.org/10.1007/s41884-021-00044-8