Stable but sensitive: multi-stable perception arbitrates to the exploitation-exploration dilemma
-
1
Otto-von-Guericke Universität, Germany
-
2
Centre de Recerca Matemàtica, Spain
-
3
Freie Universität Berlin, Germany
-
4
Universitat Politècnica de Catalunya, Spain
-
5
Institució Catalana de Recerca i Estudis Avançats, Spain
The dynamics of visual inference can be studied in the phenomenon of multi-stable visual perception: when a multi-stable visual display is viewed continuously, its phenomenal appearance reverses spontaneously at irregular intervals. Presumably, a multi-stable display stimulates recurrent neural networks with several distinct steady-states of neural activity (attractor states), which are continually destabilized by neural adaptation and by neural noise.
We set out to characterize the dynamical balance of stabilizing and destabilizing factors in the multi-stable perception. Despite high diversity of reversal statistics – differing by one or two orders of magnitude between observers and displays – we found a hidden consistency: almost all observers (21 of 24, viewing 3 displays) operated in a narrow dynamical regime near the brink of an oscillatory instability, or, equivalently in a marginally bistable regime.
This regime is functionally unique in that it combines relative stability of perceptual outcome and high sensitivity to input modulations. This combination can be understood as follows. Both dominance and response times are short at the bifurcation, but grow longer as the system enters more deeply into the bistable regime. A compromise is reached at some distance to the bifurcation. When the input changes from being balanced (I1=I2) to being biased (I1 < I2), the bifurcation border moves towards the bistable region. Accordingly, a system previously situated near the border may find now itself at the border and able to respond with a rapid reversal. In short, being near the bifurcation affords both stability when the input is constant and sensitivity when the input changes.
Our results may generalize to visual inference, where a goal conflict between stability and stability seems unavoidable. In terms of attractor dynamics, a stable appearance of a visual scene, resulting in reciprocal excitation between visual and memory activity, would be expected to stabilize a particular pattern of activity. The downside would be reduced sensitivity to incremental changes in the visual input, for attractor dynamics would tend to counteract any change. If the system is to remain sensitive, associative stabilization by memory traces must not go too far. A combination of neural noise and adaptation would seem to offer an appropriate strategy for balancing stability and sensitivity, ensuring that alternative interpretations are exhaustively explored.
In conclusion, we surmise that a marginally stable dynamics arbitrates the exploitation-exploration dilemma of a perceptual system that acquires its store of prior experience through reinforcement learning.
Keywords:
attractor dynamics,
Criticality,
multi-stable perception,
Visual Illusions
Conference:
BC11 : Computational Neuroscience & Neurotechnology Bernstein Conference & Neurex Annual Meeting 2011, Freiburg, Germany, 4 Oct - 6 Oct, 2011.
Presentation Type:
Poster
Topic:
neurons, networks and dynamical systems (please use "neurons, networks and dynamical systems" as keywords)
Citation:
Pastukhov
A,
García-Rodríguez
P,
Haenicke
J,
Guillamon
A,
Deco
G and
Braun
J
(2011). Stable but sensitive: multi-stable perception arbitrates to the exploitation-exploration dilemma.
Front. Comput. Neurosci.
Conference Abstract:
BC11 : Computational Neuroscience & Neurotechnology Bernstein Conference & Neurex Annual Meeting 2011.
doi: 10.3389/conf.fncom.2011.53.00145
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
23 Aug 2011;
Published Online:
04 Oct 2011.
*
Correspondence:
Dr. Alexander Pastukhov, Otto-von-Guericke Universität, Magdeburg, 39120, Germany, pastukhov.alexander@gmail.com