Bayesian Models of Perceptual Organization

Abstract

One of the central ideas in the study of perception is that the proximal stimulus-the pattern of energy that impinges on sensory receptors, such as the visual image-is not sufficient to specify the actual state of the world outside (the distal stimulus). That is, while the image of your grandmother on your retina might look like your grandmother, it also looks like an infinity of other arrangements of matter, each having a different combination of three-dimensional structures, surface properties, colour properties, etc., so that they happen to look just like your grandmother from a particular viewpoint. Naturally, the brain generally does not perceive these far-fetched alternatives, but rapidly converges on a single solution, which is what we consciously perceive. A shape on the retina might be a large object that is far away, or a smaller one that is closer, or anything in between. A mid-grey region on the retina might be a bright white object in dim light, a dark object in bright light, or anything in between. An elliptical shape on the retina might be an elliptical object face-on, a circular object slanted back in depth, or anything in between. Every proximal stimulus is consistent with an infinite family of possible scenes, only one of which is perceived. The central problem for the perceptual system is to quickly and reliably decide among all these alternatives, and the central problem for visual science is to figure out what rules, principles, or mechanisms the brain uses to do so. This process was called unconscious inference by Helmholtz, perhaps the first scientist to appreciate the problem, and is sometimes called inverse optics to convey the idea that the brain must in a sense invert the process of optical projection-to take the image and recover the world that gave rise to it. The modern history of visual science contains a wealth of proposals for how exactly this process works, far too numerous to review here. Some are very broad, like the Gestalt idea of Prägnanz (infer the simplest or most reasonable scene consistent with the image). Many others are narrowly addressed to specific aspects of the problem, like the inference of shape or surface colour. But historically, the vast majority of these proposals suffer from one or both of the following two problems. First, many (like Prägnanz and many other older suggestions) are too vague to be realized as computational mechanisms. They rest on central ideas, like the Gestalt term 'goodness of form' , that are subjectively defined and cannot be implemented algorithmically without a host of additional assumptions. Second, many proposed rules are arbitrary or unmotivated, meaning that is unclear exactly why the brain would choose them rather than an infinity of other equally effective ones. Of course, it cannot be taken for granted that mental processes are principled in this sense, and some have argued for a view of the brain as a 'bag of tricks' (Ramachandran 1985). Nevertheless, to many theorists, a mental function as central and evolutionarily ancient as...

The simplicity principle in perception and cognition

“…The close connection between Occam's razor and Bayes’ rule can be appreciated most directly simply by observing that the hypothesis with the highest posterior is, ipso facto, also the hypothesis with the minimum DL in Shannon's sense. In a Bayesian framework, the posterior belief in hypothesis H after considering data D , notated p ( H | D ), is proportional to the product of its prior p ( H ) and its likelihood p ( D | H ), (see Refs for tutorial introductions). The hypothesis that maximizes this quantity, sometimes called the maximum a posteriori or MAP, is the hypothesis that is the most probable, in that it maximizes the trade‐off between prior plausibility and fit to the observed data.…”

Bayesian hierarchical grouping: Perceptual grouping as mixture estimation.

“…The goal of these models, broadly speaking, is to quantify the degree of belief that ought to be assigned to each potential interpretation of image data. In these models each possible interpretation c j ∈ C = { c 1 … c J } of an image D is associated with posterior probability p ( C | D ), which according to Bayes’ rule is proportional to the product of a prior probability p ( C ) and likelihood p ( D | C ) (for introductions see Feldman, 2014; Kersten et al, 2004; Mamassian & Landy, 2002). Similary we propose a framework in which perceptual grouping can be viewed as a rational Bayesian procedure by regarding it as a kind of mixture estimation (see also Feldman et al, 2014; Froyen et al, under review).…”

Does Perceptual Consciousness Overflow Cognitive Access? The Challenge from Probabilistic, Hierarchical Processes

“…Such views are often developed in a Bayesian framework, but we needn't assume that perceptual systems are optimal. For an introduction to Bayesian modeling of perception, see Feldman (forthcoming). The Bayesian tide in cognitive science is not without its critics.…”