Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.737898
Title: Ethological decision making with non-stationary inputs using MSPRT based mechanisms
Author: Nunes, Luana F.
ISNI:       0000 0004 7225 7745
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
One of the most widely implemented models for multi-alternative decision-making is the Multihypothesis Sequential Probability Ratio Test (MSPRT); this model has found application in biological decision-making despite limitations to discrete ('trial-based'), non-time-varying scenarios. Real world situations are continuous and entail stimulus non-stationarity, making them incompatible with the MSPRT; to address this issue, we introduce a new decision mechanism by augmenting the MSPRT with an integration window which allows selection and de-selection of options as their evidence changes dynamically. In this research we explored different scenarios which did not obey the strict derivation of this algorithm, but rather constituted an empirical study of the boundaries of the applications of strict theoretical Sequential Probability Ratio Tests. In order to bridge the abstract laboratory experiments where the inputs tend to be defined as 'signals' with the real decisions performed by animals, we took an existing model of shepherd-sheep behaviour and embedded our full MSPRT based model in place of the authors' instant decision mechanism; this allowed the shepherd to perform decisions based on evidence accumulated over time, thus removing the constraint of perfect information being made available to it. The resulting model encompasses the best characteristics of the MSPRT whilst still being representative of ethological decision making algorithms.
Supervisor: Gurney, Kevin ; Gross, Roderich ; Stafford, Tom Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.737898  DOI: Not available
Share: