The Inversion of Sensory Processing by Feedback Pathways: A Model of Visual Cognitive Functions

Abstract

The mammalian visual system has a hierarchic structure with extensive reciprocal connections. A model is proposed in which the feedback pathways serve to modify afferent sensory stimuli in ways that enhance and complete sensory input patterns, suppress irrelevant features, and generate quasi-sensory patterns when afferent stimulation is weak or absent. Such inversion of sensory coding and feature extraction can be achieved by optimization processes in which scalar responses derived from high-level neural analyzers are used as cost functions to modify the filter properties of more peripheral sensory relays. An optimization algorithm, Alopex, which is used in the model, is readily implemented with known neural circuitry. The functioning of the system is investigated by computer simulations.

Free-energy and the brain

“…Conversely, "the back-projection from the higher area to the lower area provides a forward model of the optics" (Kawato et al 1993). See also Harth et al (1987). This perspective highlights the importance of backward connections and the role of empirical priors during Bayesian inversion of generative models.…”

Learning and inference in the brain

“…The architectures that emerge from these schemes suggest that "feedforward connections from the lower visual cortical area to the higher visual cortical area provides an approximated inverse model of the imaging process (optics), while the backprojection connection from the higher area to the lower area provides a forward model of the optics" (Kawato et al, 1993). See also Harth, Unnikrishnan, and Pandya (1987). The connection between this perspective on forward influences and the error minimisation can be seen by finessing the gradient decent in the E-step using a Newton -Raphson scheme.…”

Derivation and Analysis of Basic Computational Operations of Thalamocortical Circuits

“…Evidence suggests that the repeating thalamocortical, cortico-cortical, and corticothalamic projection patterns hold not only for primary sensory areas including VPM/VPL, LGd, and MGv to layer IV, and Pom, LP/Pul, and MGm to layer I of somatosensory, visual and auditory cortices, respectively (Killackey and Ebner, 1972;Ryugo and Killackey, 1974;Ribak and Peters, 1975;Herkenham, 1980;Kelly and Wong, 1981;Swadlow, 1983;Rieck and Carey, 1985;Herkenham, 1986;Jensen and Killackey, 1987;Winer and Larue, 1987;Scheel, 1988;Conley and Diamond, 1990;Rouiller and Welker, 1991;Bourassa and Deschenes, 1995;Huang and Winer, 2000), but also for a wide array of thalamic nuclei, intralaminar and nonintralaminar alike (Jones and Hendry, 1989;Rausell et al, 1992;Molinari et al, 1994;Molinari et al, 1995;Kuroda et al, 1998;Mitchell and Cauller, 2001;Rauschecker et al, 1997;Jones, 1998;Reep and Corwin, 1999;Linke and Schwegler, 2000;Jones, 2001)). For those cortical areas not receiving topographic projections from thalamus, the extensive topography-preserving cortico-cortical projections from superficial layers to recipient middle layers with reciprocal projections from the target's deep layers back to the source's superficial layers, may subserve a related function (Harth et al, 1987;Mumford, 1992;Olson and Musil, 1992;Miller 1996;Barbas and Rempel-Clower, 1997;…”