Learning with Local Models

Rüping, Stefan

doi:10.1007/11504245_10

Cited by 34 publications

(62 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[28,26]), it has surprisingly not been addressed so much in the RL community. However, works dealing with feature discovery [11], variable selection [14,9,7] or dimensionality reduction in RL [4] can indeed be considered as first steps towards interpretable solutions.…”

Section: Related Workmentioning

confidence: 99%

Policy Search in a Space of Simple Closed-form Formulas: Towards Interpretability of Reinforcement Learning

Maes

Fonteneau

Wehenkel

et al. 2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper, we address the problem of computing interpretable solutions to reinforcement learning (RL) problems. To this end, we propose a search algorithm over a space of simple closed-form formulas that are used to rank actions. We formalize the search for a high-performance policy as a multi-armed bandit problem where each arm corresponds to a candidate policy canonically represented by its shortest formula-based representation. Experiments, conducted on standard benchmarks, show that this approach manages to determine both efficient and interpretable solutions.

show abstract

Section: Related Workmentioning

confidence: 99%

Policy Search in a Space of Simple Closed-form Formulas: Towards Interpretability of Reinforcement Learning

Maes

Fonteneau

Wehenkel

et al. 2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…More recently, a method combining local and global models was proposed by Rüping. [4] In his study a local model algorithm is presented that learns a global classifier plus local models. The goal is to reduce the complexity of the global model, ensure the prediction quality of the combined model and provide guarantees that the combined and global model will differ only up to a user-specified degree.…”

Section: Introductionmentioning

confidence: 99%

Using Local Models to Improve (Q)SAR Predictivity

Buchwald

Girschick

Seeland

et al. 2011

Molecular Informatics

View full text Add to dashboard Cite

We present a novel (Q)SAR approach that detects groups of structures for local (Q)SAR modeling. The algorithm combines clustering and classification or regression for making predictions on chemical structure data. A clustering procedure producing clusters with shared structural scaffolds is applied as a preprocessing step, before a (local) model is learned for each relevant cluster. Instead of using only one global model (classical approach), we use weighted local models for predictions of query compounds dependent on cluster memberships. The approach is evaluated and compared against standard statistical (Q)SAR algorithms on various datasets. The results show that in many cases the application of local models significantly improves the predictive power of the derived (Q)SAR models compared to the classical approach, to models that are induced by a fingerprint-based or a hierarchical clustering approach and to locally weighted learning.

show abstract

“…As a result, it can work in real time (see, e.g., Bagarinao et al, 2003;DeHaan and Guay, 2006;Fei et al, 2011). A desirable feature of the model is also its transparency and interpretability because they guarantee the possibility of a better understanding of the analysed phenomenon (see, e.g., Johansson et al, 2011;Gacto et al, 2011;Rüping, 2006).…”

Section: Introductionmentioning

confidence: 99%

A new approach to nonlinear modelling of dynamic systems based on fuzzy rules

Bartczuk

Przybył

Cpałka

2016

International Journal of Applied Mathematics and Computer Science

View full text Add to dashboard Cite

For many practical weakly nonlinear systems we have their approximated linear model. Its parameters are known or can be determined by one of typical identification procedures. The model obtained using these methods well describes the main features of the system's dynamics. However, usually it has a low accuracy, which can be a result of the omission of many secondary phenomena in its description. In this paper we propose a new approach to the modelling of weakly nonlinear dynamic systems. In this approach we assume that the model of the weakly nonlinear system is composed of two parts: a linear term and a separate nonlinear correction term. The elements of the correction term are described by fuzzy rules which are designed in such a way as to minimize the inaccuracy resulting from the use of an approximate linear model. This gives us very rich possibilities for exploring and interpreting the operation of the modelled system. An important advantage of the proposed approach is a set of new interpretability criteria of the knowledge represented by fuzzy rules. Taking them into account in the process of automatic model selection allows us to reach a compromise between the accuracy of modelling and the readability of fuzzy rules.

show abstract

Learning with Local Models

Cited by 34 publications

References 10 publications

Policy Search in a Space of Simple Closed-form Formulas: Towards Interpretability of Reinforcement Learning

Policy Search in a Space of Simple Closed-form Formulas: Towards Interpretability of Reinforcement Learning

Using Local Models to Improve (Q)SAR Predictivity

A new approach to nonlinear modelling of dynamic systems based on fuzzy rules

Contact Info

Product

Resources

About