Comparison of quantum statistical models: Equivalent conditions for sufficiency (Q411358): Difference between revisions

From MaRDI portal
Importer (talk | contribs)
Created a new Item
 
Importer (talk | contribs)
Changed an Item
Property / review text
 
A physical system can be in different states \(\theta\in \Theta\). The only information about the actual state comes from measurements. Due to the inevitable measurement uncertainty, when the actual state is \(\theta\), we observe different measurement results \(\delta\in \Delta\) with known probabilities \(p_\theta(\delta)\geq 0\) (for which \(\sum_{\delta\in\Delta} p_\theta(\delta)=1\)). A \textit{statistical model} \(\mathcal E\) is defined as a triple consisting of the sets \(\Theta\) and \(\Delta\) and the probabilities \(p_\theta(\delta)\); both sets \(\Theta\) and \(\Delta\) are usually assumed to be finite. Let \(X\) be a (finite) set of possible alternatives. Then, a \textit{decision strategy} is a function that assigns to each measurement result \(\delta\) an appropriate \textit{mixed strategy}, i.e., a probability distribution \(t_\delta\) on the set \(X\). Once we know a payoff function \(\ell:\Theta\times X\to \mathbb{R}\), then for each decision strategy and for each state \(\theta\), we can compute the expected payoff \(v_\theta\); payoffs \(v_\theta\) corresponding to different states forms a \textit{payoff vector} \(\vec v\). A statistical model \(\mathcal E\) is called \textit{more informative} than \({\mathcal E}'\) if for every set of alternatives and for every payoff function, if a payoff vector \(\vec v\) is attainable in \({\mathcal E}'\), then \(\vec v\) is also attainable in \(\mathcal E\). The known Blackwell-Sherman-Stein (BSS) Theorem shows that \(\mathcal E\) is more informative the \({\mathcal E}'\) if and only if there is a transition matrix \(M:\Theta\times\Theta'\to [0,1]\) so that each probability distribution \(p'_\theta\) from \({\mathcal E}'\) is a convex combination of distributions from \(\mathcal E\): \({\vec p}^{\,\prime}_\theta=M\vec p_\theta\). The author generalizes this equivalence result to \textit{quantum} statistical models, in which for every \(\theta\), instead of a probability distribution \(p_\theta\), we have a density matrix \(\rho_\theta\). This generalization includes, as particular cases, both the classical BSS theorem and its known quantum analogues. To prove this general theorem, the author introduces a generalization of a transition matrix \(M\) under which -- in contrast to non-quantum statistical models -- the quantum distributions \(\rho'_\theta\) are not necessarily obtainable from \(\rho_\theta\) by a physically implementable randomization process. Whether physically implementable randomizations are sufficient is still an open question.
Property / review text: A physical system can be in different states \(\theta\in \Theta\). The only information about the actual state comes from measurements. Due to the inevitable measurement uncertainty, when the actual state is \(\theta\), we observe different measurement results \(\delta\in \Delta\) with known probabilities \(p_\theta(\delta)\geq 0\) (for which \(\sum_{\delta\in\Delta} p_\theta(\delta)=1\)). A \textit{statistical model} \(\mathcal E\) is defined as a triple consisting of the sets \(\Theta\) and \(\Delta\) and the probabilities \(p_\theta(\delta)\); both sets \(\Theta\) and \(\Delta\) are usually assumed to be finite. Let \(X\) be a (finite) set of possible alternatives. Then, a \textit{decision strategy} is a function that assigns to each measurement result \(\delta\) an appropriate \textit{mixed strategy}, i.e., a probability distribution \(t_\delta\) on the set \(X\). Once we know a payoff function \(\ell:\Theta\times X\to \mathbb{R}\), then for each decision strategy and for each state \(\theta\), we can compute the expected payoff \(v_\theta\); payoffs \(v_\theta\) corresponding to different states forms a \textit{payoff vector} \(\vec v\). A statistical model \(\mathcal E\) is called \textit{more informative} than \({\mathcal E}'\) if for every set of alternatives and for every payoff function, if a payoff vector \(\vec v\) is attainable in \({\mathcal E}'\), then \(\vec v\) is also attainable in \(\mathcal E\). The known Blackwell-Sherman-Stein (BSS) Theorem shows that \(\mathcal E\) is more informative the \({\mathcal E}'\) if and only if there is a transition matrix \(M:\Theta\times\Theta'\to [0,1]\) so that each probability distribution \(p'_\theta\) from \({\mathcal E}'\) is a convex combination of distributions from \(\mathcal E\): \({\vec p}^{\,\prime}_\theta=M\vec p_\theta\). The author generalizes this equivalence result to \textit{quantum} statistical models, in which for every \(\theta\), instead of a probability distribution \(p_\theta\), we have a density matrix \(\rho_\theta\). This generalization includes, as particular cases, both the classical BSS theorem and its known quantum analogues. To prove this general theorem, the author introduces a generalization of a transition matrix \(M\) under which -- in contrast to non-quantum statistical models -- the quantum distributions \(\rho'_\theta\) are not necessarily obtainable from \(\rho_\theta\) by a physically implementable randomization process. Whether physically implementable randomizations are sufficient is still an open question. / rank
 
Normal rank
Property / reviewed by
 
Property / reviewed by: Vladik Ya. Kreinovich / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 81P16 / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 81P15 / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 91B06 / rank
 
Normal rank
Property / Mathematics Subject Classification ID
 
Property / Mathematics Subject Classification ID: 62C05 / rank
 
Normal rank
Property / zbMATH DE Number
 
Property / zbMATH DE Number: 6021961 / rank
 
Normal rank
Property / zbMATH Keywords
 
statistical models
Property / zbMATH Keywords: statistical models / rank
 
Normal rank
Property / zbMATH Keywords
 
quantum statistical models
Property / zbMATH Keywords: quantum statistical models / rank
 
Normal rank

Revision as of 18:37, 29 June 2023

scientific article
Language Label Description Also known as
English
Comparison of quantum statistical models: Equivalent conditions for sufficiency
scientific article

    Statements

    Comparison of quantum statistical models: Equivalent conditions for sufficiency (English)
    0 references
    0 references
    4 April 2012
    0 references
    A physical system can be in different states \(\theta\in \Theta\). The only information about the actual state comes from measurements. Due to the inevitable measurement uncertainty, when the actual state is \(\theta\), we observe different measurement results \(\delta\in \Delta\) with known probabilities \(p_\theta(\delta)\geq 0\) (for which \(\sum_{\delta\in\Delta} p_\theta(\delta)=1\)). A \textit{statistical model} \(\mathcal E\) is defined as a triple consisting of the sets \(\Theta\) and \(\Delta\) and the probabilities \(p_\theta(\delta)\); both sets \(\Theta\) and \(\Delta\) are usually assumed to be finite. Let \(X\) be a (finite) set of possible alternatives. Then, a \textit{decision strategy} is a function that assigns to each measurement result \(\delta\) an appropriate \textit{mixed strategy}, i.e., a probability distribution \(t_\delta\) on the set \(X\). Once we know a payoff function \(\ell:\Theta\times X\to \mathbb{R}\), then for each decision strategy and for each state \(\theta\), we can compute the expected payoff \(v_\theta\); payoffs \(v_\theta\) corresponding to different states forms a \textit{payoff vector} \(\vec v\). A statistical model \(\mathcal E\) is called \textit{more informative} than \({\mathcal E}'\) if for every set of alternatives and for every payoff function, if a payoff vector \(\vec v\) is attainable in \({\mathcal E}'\), then \(\vec v\) is also attainable in \(\mathcal E\). The known Blackwell-Sherman-Stein (BSS) Theorem shows that \(\mathcal E\) is more informative the \({\mathcal E}'\) if and only if there is a transition matrix \(M:\Theta\times\Theta'\to [0,1]\) so that each probability distribution \(p'_\theta\) from \({\mathcal E}'\) is a convex combination of distributions from \(\mathcal E\): \({\vec p}^{\,\prime}_\theta=M\vec p_\theta\). The author generalizes this equivalence result to \textit{quantum} statistical models, in which for every \(\theta\), instead of a probability distribution \(p_\theta\), we have a density matrix \(\rho_\theta\). This generalization includes, as particular cases, both the classical BSS theorem and its known quantum analogues. To prove this general theorem, the author introduces a generalization of a transition matrix \(M\) under which -- in contrast to non-quantum statistical models -- the quantum distributions \(\rho'_\theta\) are not necessarily obtainable from \(\rho_\theta\) by a physically implementable randomization process. Whether physically implementable randomizations are sufficient is still an open question.
    0 references
    statistical models
    0 references
    quantum statistical models
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references