Eyke Hullermeier

AAAI Conference 2020 Conference Paper

Reliable Multilabel Classification: Prediction with Partial Abstention

• Vu-Linh Nguyen
• Eyke Hullermeier

In contrast to conventional (single-label) classification, the setting of multilabel classification (MLC) allows an instance to belong to several classes simultaneously. Thus, instead of selecting a single class label, predictions take the form of a subset of all labels. In this paper, we study an extension of the setting of MLC, in which the learner is allowed to partially abstain from a prediction, that is, to deliver predictions on some but not necessarily all class labels. We propose a formalization of MLC with abstention in terms of a generalized loss minimization problem and present first results for the case of the Hamming loss, rank loss, and F-measure, both theoretical and experimental.

EWRL Workshop 2013 Workshop Paper

Preference-based Evolutionary Direct Policy Search

• Robert Busa-Fekete
• Balazs Szorenyi
• Paul Weng
• Weiwei Cheng
• Eyke Hullermeier

KR Conference 2002 Conference Paper

A fuzzy set approach to flexible case-based querying: methodology and experimentation

• Martine de Calmes
• Didier Dubois
• Eyke Hullermeier
• Henri Prade
• Florence Sedes

IJCAI Conference 1999 Conference Paper

Toward a Probabilistic Formalization of Case-Based Inference

• Eyke Hullermeier

We propose a formal framework for modelling case-based inference ( C B I ), which is a crucial part of the case-based reasoning ( C B R ) methodology. As a representation of the similarity structure of a system, the concept of a similarity profile is introduced. This concept makes it possible to formalize the C B R hypothesis that "similar problems have similar solutions" and to realize C B I in the form of constraint-based inference. In order to exploit the similarity structure more efficiently, a probabilistic generalization of the constraintbased view is developed. This formalization allows for realizing C B I in the context of probabilistic reasoning and statistical inference and, hence, makes a powerful methodological framework accessible to C B R. Within the generalized setting, a (formalized) C B R hypothesis corresponds to the assumption of a certain stochastic model, and a memory of cases can be seen as statistical data underlying the inference process. As a particular result we establish an approximate probabilistic reasoning scheme which generalizes the constraint-based approach.