The Experts below are selected from a list of 324 Experts worldwide ranked by ideXlab platform
Prasad Tadepalli - One of the best experts on this subject based on the ideXlab platform.
-
The Choice Function Framework for Online Policy Improvement
Proceedings of the AAAI Conference on Artificial Intelligence, 2020Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect leaf evaluation Function and transition model. Indeed, simple counterexamples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the empirical utility of the framework.
-
The Choice Function Framework for Online Policy Improvement
arXiv: Artificial Intelligence, 2019Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect evaluation Function and transition model. Indeed, simple counter examples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the framework's empirical utility.
Irina Georgescu - One of the best experts on this subject based on the ideXlab platform.
-
Arrow Index of a Fuzzy Choice Function
SSRN Electronic Journal, 2009Co-Authors: Irina GeorgescuAbstract:The Arrow index of a fuzzy Choice Function C is a measure of the degree to which C satisfies the Fuzzy Arrow Axiom, a fuzzy version of the classical Arrow Axiom. The main result of this paper shows that A(C) characterizes the degree to which C is full rational. We also obtain a method for computing A(C). The Arrow index allows to rank the fuzzy Choice Functions with respect to their rationality. Thus, if for solving a decision problem several fuzzy Choice Functions are proposed, by the Arrow index the most rational one will be chosen.
-
Acyclic rationality indicators of fuzzy Choice Functions
Fuzzy Sets and Systems, 2009Co-Authors: Irina GeorgescuAbstract:We introduce the degree of acyclicity of a fuzzy preference relation and two acyclic rationality indicators of a fuzzy Choice Function C. These two indicators measure the degree to which the Choice Function C is acyclic G-rational, respectively, acyclic M-rational. We study the fuzzy Condorcet property of a fuzzy Choice Function and the corresponding indicators. We express the indicator of G-rationality in terms of the Condorcet property and the indicator of @a-consistency.
-
Ranking fuzzy Choice Functions by their rationality indicators
Fuzzy Optimization and Decision Making, 2007Co-Authors: Irina GeorgescuAbstract:In this paper we study two rationality indicators and two normality indicators of a fuzzy Choice Function. They express the degree of rationality or normality of this fuzzy Choice Function. This way we can establish a hierarchy in a given family of fuzzy Choice Functions with respect to their degree of rationality.
-
Congruence indicators for fuzzy Choice Functions
Social Choice and Welfare, 2007Co-Authors: Irina GeorgescuAbstract:We introduce the congruence indicators WFCA(·) and SFCA(·) corresponding to fuzzy congruence axioms WFCA and SFCA. These indicators measure the degree to which a fuzzy Choice Function verifies the axioms WFCA and SFCA, respectively. The main result of the paper establishes for a given Choice Function the relationship between its congruence indicators and some rationality conditions. One obtains a fuzzy counterpart of the well-known Arrow–Sen theorem in crisp Choice Functions theory.
-
EUSFLAT Conf. (2) - Revealed Preference in a Fuzzy Framework.
2007Co-Authors: Irina GeorgescuAbstract:To a fuzzy Choice Function C we assign the indicators of revealed preference WAFRP (C), SAFRP (C) and the indicators of congruence WFCA(C), SFCA(C). These indicators measure the degree to which the fuzzy Choice Function C verifies the axioms of revealed preference WAFRP , SAFRP and of congruenceWFCA and SFCA, respectively.
Murugeswari Issakkimuthu - One of the best experts on this subject based on the ideXlab platform.
-
The Choice Function Framework for Online Policy Improvement
Proceedings of the AAAI Conference on Artificial Intelligence, 2020Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect leaf evaluation Function and transition model. Indeed, simple counterexamples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the empirical utility of the framework.
-
The Choice Function Framework for Online Policy Improvement
arXiv: Artificial Intelligence, 2019Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect evaluation Function and transition model. Indeed, simple counter examples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the framework's empirical utility.
Arunava Sen - One of the best experts on this subject based on the ideXlab platform.
-
Robust incentive compatibility of voting rules with positively correlated beliefs
Social Choice and Welfare, 2021Co-Authors: Dipjyoti Majumdar, Arunava SenAbstract:We investigate a voting model where each voter’s beliefs are positively correlated. We show that requiring a social Choice Function to be Ordinally Bayesian Incentive-Compatible (d’Aspremont and Peleg in Soc Choice Welf 5:261–280, 1988) with respect to all such beliefs is not equivalent to requiring it to be strategy-proof. However, if the social Choice Function is also required to be efficient, it must be strategy-proof and hence, dictatorial.
-
A Characterization of Single-Peaked Preferences via Random Social Choice Functions
Theoretical Economics, 2016Co-Authors: Shurojit Chatterji, Arunava Sen, Huaxia ZengAbstract:The paper proves the following result: every path-connected domain of preferences that admits a strategy-proof, unanimous, tops-only random social Choice Function satisfying a compromise property, is single-peaked. Conversely, every single-peaked domain admits a random social Choice Function satisfying these properties. Single-peakedness is dened with respect to arbitrary trees. The paper provides a justication of the salience of single-peaked preferences and evidence in favour of the Gul conjecture (Barber a (2010)).
-
weak monotonicity characterizes deterministic dominant strategy implementation
Econometrica, 2006Co-Authors: Sushil Bikhchandani, Shurojit Chatterji, Ron Lavi, Ahuva Mualem, Noam Nisan, Arunava SenAbstract:We characterize dominant-strategy incentive compatibility with multidimensional types. A deterministic social Choice Function is dominant-strategy incentive compatible if and only if it is weakly monotone (W-Mon). The W-Mon requirement is the following: If changing one agent's type (while keeping the types of other agents fixed) changes the outcome under the social Choice Function, then the resulting difference in utilities of the new and original outcomes evaluated at the new type of this agent must be no less than this difference in utilities evaluated at the original type of this agent.
Alan Fern - One of the best experts on this subject based on the ideXlab platform.
-
The Choice Function Framework for Online Policy Improvement
Proceedings of the AAAI Conference on Artificial Intelligence, 2020Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect leaf evaluation Function and transition model. Indeed, simple counterexamples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the empirical utility of the framework.
-
The Choice Function Framework for Online Policy Improvement
arXiv: Artificial Intelligence, 2019Co-Authors: Murugeswari Issakkimuthu, Alan Fern, Prasad TadepalliAbstract:There are notable examples of online search improving over hand-coded or learned policies (e.g. AlphaZero) for sequential decision making. It is not clear, however, whether or not policy improvement is guaranteed for many of these approaches, even when given a perfect evaluation Function and transition model. Indeed, simple counter examples show that seemingly reasonable online search procedures can hurt performance compared to the original policy. To address this issue, we introduce the Choice Function framework for analyzing online search procedures for policy improvement. A Choice Function specifies the actions to be considered at every node of a search tree, with all other actions being pruned. Our main contribution is to give sufficient conditions for stationary and non-stationary Choice Functions to guarantee that the value achieved by online search is no worse than the original policy. In addition, we describe a general parametric class of Choice Functions that satisfy those conditions and present an illustrative use case of the framework's empirical utility.