On the Fisher metric of conditional probability polytopes (Q296470): Difference between revisions

From MaRDI portal
Importer (talk | contribs)
Created a new Item
 
ReferenceBot (talk | contribs)
Changed an Item
(5 intermediate revisions by 4 users not shown)
Property / review text
 
Summary: We consider three different approaches to define natural Riemannian metrics on polytopes of stochastic matrices. First, we define a natural class of stochastic maps between these polytopes and give a metric characterization of Chentsov type in terms of invariance with respect to these maps. Second, we consider the Fisher metric defined on arbitrary polytopes through their embeddings as exponential families in the probability simplex. We show that these metrics can also be characterized by an invariance principle with respect to morphisms of exponential families. Third, we consider the Fisher metric resulting from embedding the polytope of stochastic matrices in a simplex of joint distributions by specifying a marginal distribution. All three approaches result in slight variations of products of Fisher metrics. This is consistent with the nature of polytopes of stochastic matrices, which are Cartesian products of probability simplices. The first approach yields a scaled product of Fisher metrics; the second, a product of Fisher metrics; and the third, a product of Fisher metrics scaled by the marginal distribution.
Property / review text: Summary: We consider three different approaches to define natural Riemannian metrics on polytopes of stochastic matrices. First, we define a natural class of stochastic maps between these polytopes and give a metric characterization of Chentsov type in terms of invariance with respect to these maps. Second, we consider the Fisher metric defined on arbitrary polytopes through their embeddings as exponential families in the probability simplex. We show that these metrics can also be characterized by an invariance principle with respect to morphisms of exponential families. Third, we consider the Fisher metric resulting from embedding the polytope of stochastic matrices in a simplex of joint distributions by specifying a marginal distribution. All three approaches result in slight variations of products of Fisher metrics. This is consistent with the nature of polytopes of stochastic matrices, which are Cartesian products of probability simplices. The first approach yields a scaled product of Fisher metrics; the second, a product of Fisher metrics; and the third, a product of Fisher metrics scaled by the marginal distribution. / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 62F99 / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 60D99 / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 62H11 / rank
 
Normal rank
Property / zbMATH DE Number
 
Property / zbMATH DE Number: 6593654 / rank
 
Normal rank
Property / zbMATH Keywords
 
Fisher information metric
Property / zbMATH Keywords: Fisher information metric / rank
 
Normal rank
Property / zbMATH Keywords
 
information geometry
Property / zbMATH Keywords: information geometry / rank
 
Normal rank
Property / zbMATH Keywords
 
convex support polytope
Property / zbMATH Keywords: convex support polytope / rank
 
Normal rank
Property / zbMATH Keywords
 
conditional model
Property / zbMATH Keywords: conditional model / rank
 
Normal rank
Property / zbMATH Keywords
 
Markov morphism
Property / zbMATH Keywords: Markov morphism / rank
 
Normal rank
Property / zbMATH Keywords
 
isometric embedding
Property / zbMATH Keywords: isometric embedding / rank
 
Normal rank
Property / zbMATH Keywords
 
natural gradient
Property / zbMATH Keywords: natural gradient / rank
 
Normal rank
Property / MaRDI profile type
 
Property / MaRDI profile type: MaRDI publication profile / rank
 
Normal rank
Property / OpenAlex ID
 
Property / OpenAlex ID: W3105865414 / rank
 
Normal rank
Property / arXiv ID
 
Property / arXiv ID: 1404.0198 / rank
 
Normal rank
Property / cites work
 
Property / cites work: A new mathematical framework for the study of linkage and selection / rank
 
Normal rank
Property / cites work
 
Property / cites work: Simulation-based optimization of Markov reward processes / rank
 
Normal rank
Property / cites work
 
Property / cites work: Axiomatic Geometry of Conditional Models / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q4167890 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3995193 / rank
 
Normal rank
Property / cites work
 
Property / cites work: On a notion of linear replicator equations / rank
 
Normal rank
links / mardi / namelinks / mardi / name
 

Revision as of 03:38, 12 July 2024

scientific article
Language Label Description Also known as
English
On the Fisher metric of conditional probability polytopes
scientific article

    Statements

    On the Fisher metric of conditional probability polytopes (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    15 June 2016
    0 references
    Summary: We consider three different approaches to define natural Riemannian metrics on polytopes of stochastic matrices. First, we define a natural class of stochastic maps between these polytopes and give a metric characterization of Chentsov type in terms of invariance with respect to these maps. Second, we consider the Fisher metric defined on arbitrary polytopes through their embeddings as exponential families in the probability simplex. We show that these metrics can also be characterized by an invariance principle with respect to morphisms of exponential families. Third, we consider the Fisher metric resulting from embedding the polytope of stochastic matrices in a simplex of joint distributions by specifying a marginal distribution. All three approaches result in slight variations of products of Fisher metrics. This is consistent with the nature of polytopes of stochastic matrices, which are Cartesian products of probability simplices. The first approach yields a scaled product of Fisher metrics; the second, a product of Fisher metrics; and the third, a product of Fisher metrics scaled by the marginal distribution.
    0 references
    Fisher information metric
    0 references
    information geometry
    0 references
    convex support polytope
    0 references
    conditional model
    0 references
    Markov morphism
    0 references
    isometric embedding
    0 references
    natural gradient
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references