Analogy-Based Preference Learning with Kernels Machine Learning

Building on a specific formalization of analogical relationships of the form "A relates to B as C relates to D", we establish a connection between two important subfields of artificial intelligence, namely analogical reasoning and kernel-based machine learning. More specifically, we show that so-called analogical proportions are closely connected to kernel functions on pairs of objects. Based on this result, we introduce the analogy kernel, which can be seen as a measure of how strongly four objects are in analogical relationship. As an application, we consider the problem of object ranking in the realm of preference learning, for which we develop a new method based on support vector machines trained with the analogy kernel. Our first experimental results for data sets from different domains (sports, education, tourism, etc.) are promising and suggest that our approach is competitive to state-of-the-art algorithms in terms of predictive accuracy.

An Analogy Ontology for Integrating Analogical Processing and First-principles Reasoning

AAAI Conferences

This paper describes an analogy ontology, a formal representation of some key ideas in analogical processing, that supports the integration of analogical processing with first-principles reasoners. The ontology is based on Gentner's structure-mapping theory, a psychological account of analogy and similarity. The semantics of the ontology are enforced via procedural attachment, using cognitive simulations of structure-mapping to provide analogical processing services. Introduction There is mounting psychological evidence that human cognition centrally involves similarity computations over structured representations, in tasks ranging from high-level visual perception to problem solving, learning, and conceptual change [21]. Understanding how to integrate analogical processing into AI systems seems crucial to creating more humanlike reasoning systems [12].

Companion Cognitive Systems: A step towards human-level AI

AAAI Conferences

We are developing Companion Cognitive Systems, a new kind of software that can be effectively treated as a collaborator. Aside from their potential utility, we believe this effort is important because it focuses on three key problems that must be solved to achieve human-level AI: Robust reasoning and learning, performance and longevity, and interactivity. We describe the ideas we are using to develop the first architecture for Companions: Analogical processing, grounded in cognitive science for reasoning and learning, a distributed agent architecture hosted on a cluster to achieve performance and longevity, and sketching and concept maps to provide interactivity.


AAAI Conferences

Qualitative physics addresses the problem of modeling physical systems and reasoning about *This research has been sponsored by the Belgian Government with the contract "Incentive Program For Fundamental Research In Artificial Intelligence; Project: Self-organization in subsymbolic computation" and "Geconcerteerde Actie: Artifici le intelligentie, Parallelle Architecturen en Interfaces". Part of this research has also been funded by the Esprit Program with the contract P440: "Message Passing" Furthermore the basic strategy which consists in qualifying the equations of physics in order to obtain a useful set of inference rules is not applicable to liquids. We elaborate on these criticisms in the next section. The rest of this paper is divided into two parts. In section three we propose a hybrid architecture for the representation of the behavior of liquids. This architecture is composed of 216 a traditional symbolic reasoning module and an analogical simulation module coupled through different interpretation and visualization routines. We also briefly comment on the kind of analogical simulation needed for the purpose of predicting qualitatively the behavior of liquids in a large variety of possible situations. In the fourth section we discuss the complementary aspects of analogical and symbolic representation and argue that both are needed to build powerful and accurate models of liquid behavior.