Inductive logic is concerned with assigning probabilities to sentences given probabilistic constraints. The Maximum Entropy Approach to inductive logic I here consider assigns probabilities to all sentences of a first order predicate logic. This assignment is built on an application of the Maximum Entropy Principle, which requires that probabilities for uncertain inference have maximal Shannon Entropy. This paper puts forward two different modified applications of this principle to first order predicate logic and shows that the original and the two modified applications agree in many cases. A third promising modification is studied and rejected.
A Triple Uniqueness of the Maximum Entropy Approach / J. Landes (LECTURE NOTES IN COMPUTER SCIENCE). - In: Symbolic and Quantitative Approaches to Reasoning with Uncertainty : / [a cura di] J. Vejnarová, N. Wilson. - Cham : Springer Science and Business Media Deutschland GmbH, 2021. - ISBN 978-3-030-86771-3. - pp. 644-656 (( Intervento presentato al 16. convegno European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty tenutosi a Prague : September 21–24 nel 2021 [10.1007/978-3-030-86772-0_46].
A Triple Uniqueness of the Maximum Entropy Approach
J. Landes
2021
Abstract
Inductive logic is concerned with assigning probabilities to sentences given probabilistic constraints. The Maximum Entropy Approach to inductive logic I here consider assigns probabilities to all sentences of a first order predicate logic. This assignment is built on an application of the Maximum Entropy Principle, which requires that probabilities for uncertain inference have maximal Shannon Entropy. This paper puts forward two different modified applications of this principle to first order predicate logic and shows that the original and the two modified applications agree in many cases. A third promising modification is studied and rejected.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.