The Dynamics of Perceptual Learning: An Incremental Reweighting Model.

Abstract

The mechanisms of perceptual learning are analyzed theoretically, probed in an orientation-discrimination experiment involving a novel nonstationary context manipulation, and instantiated in a detailed computational model. Two hypotheses are examined: modification of early cortical representations versus task-specific selective reweighting. Representation modification seems neither functionally necessary nor implied by the available psychophysical and physiological evidence. Computer simulations and mathematical analyses demonstrate the functional and empirical adequacy of selective reweighting as a perceptual learning mechanism. The stimulus images are processed by standard orientation- and frequency-tuned representational units, divisively normalized. Learning occurs only in the "read-out" connections to a decision unit; the stimulus representations never change. An incremental Hebbian rule tracks the task-dependent predictive value of each unit, thereby improving the signal-to-noise ratio of their weighted combination. Each abrupt change in the environmental statistics induces a switch cost in the learning curves as the system temporarily works with suboptimal weights.

A neural population model for visual pattern detection.

“…We introduced and evaluated the neural population model, designed to meet these goals. This theory builds on several earlier models and findings in the vision literature (e.g., Blakemore & Campbell, 1969;Campbell & Robson, 1968;Davis et al, 1983;Dosher & Lu, 1998;Foley, 1994;Geisler & Albrecht, 1997;Graham & Nachmias, 1971;Itti et al, 2000;Legge & Foley, 1980;Parker & Newsome, 1998;Pelli, 1985;Petrov et al, 2005). It is a mechanistic model in which visual information is encoded by a neurophysiologically inspired front-end resembling V1 and decoded by a deci- Figure 17.…”

“…The success of the various multichannel models that make use of linear-nonlinear cascades and population decoding to account for important aspects of detection, pattern masking (e.g., Itti et al, 2000), visual attention (e.g., Lee et al, 1999), and perceptual learning (e.g., Petrov et al, 2005) raises the question whether it is possible to formulate a single model that achieves all of this at once. Given that the proposed models differ in many of their Figure 18.…”

“…Models that bear similarity to ours have been applied before in the domains of pattern masking (Chirimuuta & Tolhurst, 2005;Geisler & Albrecht, 1997;Goris, Wichmann, & Henning, 2009;Itti, Koch, & Braun, 2000;Sanborn & Dayan, 2011), perceptual learning (Petrov et al, 2005), and visual attention (Lee, Itti, Koch, & Braun, 1999). In all these models, visual information is initially processed by a population of linear-nonlinear cascades tuned to the stimulus dimensions of interest.…”

“…This calls for a multichannel approach, whereby the encoding and decoding stage are separated and made explicit (for simulations and analysis of a multichannel variant of the PTM in the context of perceptual learning, see Dosher & Lu, 1998;Petrov, Dosher, & Lu, 2005).…”

The application of online transcranial random noise stimulation and perceptual learning in the improvement of visual functions in mild myopia

“…This change can be attributed to a specific and localised plasticity in V1 or alternatively to a more complex network involving changes in the inputs V1 receives from other higher order brain regions following practice (top-down modulation). Additionally, PL may maximize in decision mechanisms the read-out of sensory signals through internal reward (Kumano & Uka, 2013;Lu, Liu, & Dosher, 2010;Petrov, Dosher, & Lu, 2005).…”