Active Sensing as Bayes-Optimal Sequential Decision Making

Sheeraz Ahmad, Angela Yu
Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence, PMLR R11:15-24, 2013.

Abstract

Sensory inference under conditions of uncer- tainty is a major problem in both machine learning and computational neuroscience. An important but poorly understood aspect of sensory processing is the role of active sensing. Here, we present a Bayes-optimal inference and control framework for active sensing, C-DAC (Context-Dependent Active Controller). Unlike previously proposed al- gorithms that optimize abstract statistical objectives such as information maximization (Infomax) [Butko and Movellan, 2010] or one-step look-ahead accuracy [Najemnik and Geisler, 2005], our active sensing model di- rectly minimizes a combination of behavioral costs, such as temporal delay, response error, and sensor repositioning cost. We simulate these algorithms on a simple visual search task to illustrate scenarios in which context- sensitivity is particularly beneficial and op- timization with respect to generic statisti- cal objectives particularly inadequate. Mo- tivated by the geometric properties of the C- DAC policy, we present both parametric and non-parametric approximations, which retain context-sensitivity while significantly reduc- ing computational complexity. These ap- proximations enable us to investigate a more complex search problem involving peripheral vision, and we notice that the performance advantage of C-DAC over generic statistical policies is even more evident in this scenario.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR11-ahmad13a, title = {Active Sensing as {B}ayes-Optimal Sequential Decision Making}, author = {Ahmad, Sheeraz and Yu, Angela}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {15--24}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/ahmad13a/ahmad13a.pdf}, url = {https://proceedings.mlr.press/r11/ahmad13a.html}, abstract = {Sensory inference under conditions of uncer- tainty is a major problem in both machine learning and computational neuroscience. An important but poorly understood aspect of sensory processing is the role of active sensing. Here, we present a Bayes-optimal inference and control framework for active sensing, C-DAC (Context-Dependent Active Controller). Unlike previously proposed al- gorithms that optimize abstract statistical objectives such as information maximization (Infomax) [Butko and Movellan, 2010] or one-step look-ahead accuracy [Najemnik and Geisler, 2005], our active sensing model di- rectly minimizes a combination of behavioral costs, such as temporal delay, response error, and sensor repositioning cost. We simulate these algorithms on a simple visual search task to illustrate scenarios in which context- sensitivity is particularly beneficial and op- timization with respect to generic statisti- cal objectives particularly inadequate. Mo- tivated by the geometric properties of the C- DAC policy, we present both parametric and non-parametric approximations, which retain context-sensitivity while significantly reduc- ing computational complexity. These ap- proximations enable us to investigate a more complex search problem involving peripheral vision, and we notice that the performance advantage of C-DAC over generic statistical policies is even more evident in this scenario.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Active Sensing as Bayes-Optimal Sequential Decision Making %A Sheeraz Ahmad %A Angela Yu %B Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2013 %E Ann Nicholson %E Padhraic Smyth %F pmlr-vR11-ahmad13a %I PMLR %P 15--24 %U https://proceedings.mlr.press/r11/ahmad13a.html %V R11 %X Sensory inference under conditions of uncer- tainty is a major problem in both machine learning and computational neuroscience. An important but poorly understood aspect of sensory processing is the role of active sensing. Here, we present a Bayes-optimal inference and control framework for active sensing, C-DAC (Context-Dependent Active Controller). Unlike previously proposed al- gorithms that optimize abstract statistical objectives such as information maximization (Infomax) [Butko and Movellan, 2010] or one-step look-ahead accuracy [Najemnik and Geisler, 2005], our active sensing model di- rectly minimizes a combination of behavioral costs, such as temporal delay, response error, and sensor repositioning cost. We simulate these algorithms on a simple visual search task to illustrate scenarios in which context- sensitivity is particularly beneficial and op- timization with respect to generic statisti- cal objectives particularly inadequate. Mo- tivated by the geometric properties of the C- DAC policy, we present both parametric and non-parametric approximations, which retain context-sensitivity while significantly reduc- ing computational complexity. These ap- proximations enable us to investigate a more complex search problem involving peripheral vision, and we notice that the performance advantage of C-DAC over generic statistical policies is even more evident in this scenario. %Z Reissued by PMLR on 04 October 2026.
APA
Ahmad, S. & Yu, A.. (2013). Active Sensing as Bayes-Optimal Sequential Decision Making. Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R11:15-24 Available from https://proceedings.mlr.press/r11/ahmad13a.html. Reissued by PMLR on 04 October 2026.

Related Material