Bounded rational decision-making with adaptive neural network priors
Beitrag zu einer Konferenz
Braun, Daniel A.
FacultiesFakultät für Ingenieurwissenschaften, Informatik und Psychologie
InstitutionsInstitut für Neuroinformatik
Artificial Neural Networks in Pattern Recognition ; 2018 (2018). - S. 213-225. - ISBN 978-3-319-99977-7, ISBN 978-3-319-99978-4. - ISSN 0302-9743
Link to original publicationhttps://dx.doi.org/10.1007/978-3-319-99978-4_17
8th IAPR TC3 Workshop on Artificial Neural Networks in Pattern Recognition, 2018-09-19 - 2018-09-21, Siena
LicenseCC BY 4.0 International
Bounded rationality investigates utility-optimizing decision-makers with limited information-processing power. In particular, information theoretic bounded rationality models formalize resource constraints abstractly in terms of relative Shannon information, namely the Kullback-Leibler Divergence between the agents' prior and posterior policy. Between prior and posterior lies an anytime deliberation process that can be instantiated by sample-based evaluations of the utility function through Markov Chain Monte Carlo (MCMC) optimization. The most simple model assumes a fixed prior and can relate abstract information-theoretic processing costs to the number of sample evaluations. However, more advanced models would also address the question of learning, that is how the prior is adapted over time such that generated prior proposals become more efficient. In this work we investigate generative neural networks as priors that are optimized concurrently with anytime sample-based decision-making processes such as MCMC. We evaluate this approach on toy examples.
BRISC / Bounded Rationality in Sensorimotor Coordination / EC / H2020 / 678082
Subject HeadingsEingeschränkte Rationalität [GND]
Decoders (Electronics) [LCSH]
Markov processes [LCSH]
Monte Carlo method [LCSH]