Hierarchical Bayesian inference in the visual cortex.

By Tai Sing Lee, David Mumford. Journal of The Optical Society of America A-optics Image Science and Vision, 2003

1,593 citations recorded by the source corpus.

Traditional views of visual processing suggest that early visual neurons in areas V1 and V2 are static spatiotemporal filters that extract local features from a visual scene. The extracted information is then channeled through a feedforward chain of modules in successively higher visual areas for further analysis. Recent electrophysiological recordings from early visual neurons in awake behaving monkeys reveal that there are many levels of complexity in the information processing of the early visual cortex, as seen in the long-latency responses of its neurons. These new findings suggest that activity in the early visual cortex is tightly coupled and highly interactive with the rest of the visual system. They lead us to propose a new theoretical setting based on the mathematical framework of hierarchical Bayesian inference for reasoning about the visual system. In this framework, the recurrent feedforward/feedback loops in the cortex serve to integrate top-down contextual priors and bottom-up observations so as to implement concurrent probabilistic inference along the visual hierarchy. We suggest that the algorithms of particle filtering and Bayesian-belief propagation might model these interactive cortical computations. We review some recent neurophysiological evidences that support the plausibility of these ideas.

Subfields: Cognitive Neuroscience, Signal Processing

View Paper: https://www.semanticscholar.org/paper/5042cb5efad30f443adef472b8748e1b7bb0452f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Signal Processing

Citations in the corpus

Theory of cortical function (2017, methodology): “The model has since been applied to explain physiological measurements of neural activity in a wide variety of neural systems (43-60), and behavioral/ perceptual analogs of those physiological phenomena (e.”
Deep networks for robust visual recognition (2010, background): “(Lee & Mumford, 2003) proposed that the human visual cortex performs hierarchical Bayesian inference where " beliefs " are propagated up and down the hierarchy.”
Multistability and Perceptual Inference (2012, background): “…Griffiths, & Tenenbaum, 2009) and animal (Daw & Courville, 2007) cognition, and several authors have proposed that the spiking activity of neurons might be interpreted as a set of samples (Hoyer & Hyvärinen, 2003; Lee & Mumford, 2003; Shi & Griffiths, 2009; Fiser, Berkes, Orbán, & Lengyel, 2010).”
Deep Hierarchies in the Primate Visual Cortex: What Can We Learn for Computer Vision? (2013, background): “It also means that a system should be able to produce several hypotheses that are concurrently considered and possibly not resolved [102].”
The Visual System's Internal Model of the World (2015, background): “…hierarchical architecture of the biological visual system and share many fundamental characteristics, they represent different perspectives how the internal model can be learned and used for inference, and in my opinion, each capturing or emphasizing certain elements of the reality of the brain.”
A generative vision model that trains with high data efficiency and breaks text-based CAPTCHAs (2017, background): “Although the RCN network is extremely loopy, we found that message-passing (38), with a schedule inspired by the timing of activations in the visual cortex (9, 20), resulted in fast and accurate inference.”
The Predictive Coding Account of Psychosis (2018, background): “This process can be formalized as Bayesian inference (5,8), whereby a probabilistic prediction (prior) is combined with observed sensory data (likelihood) to compute”
Cortical High-Density Counterstream Architectures (2013, background): “dictive coding, and emerging concepts of inference in cortical function (18, 69, 70).”
Vanilla PP for Philosophers: A Primer on Predictive Processing (2017, background): “…levels in the hierarchy are conditionally independent, estimates can be updated in parallel at the different levels (cf. Friston 2003, p. 1342), which ideally yields a globally consistent set of estimates (in practice, things are complicated, as Lee and Mumford 2003, p. 1437, point out).”
Feature Learning for Image Classification Via Multiobjective Genetic Programming (2014, methodology): “In this section, we describe the details of our GP implementation and the relevant experimental results we obtain by our approach.”
Learning Deep Architectures for AI (2007, background): “, going back from later stages of visual processing to earlier ones) and these may serve a role not only in learning (as in RBMs) but also in integrating contextual priors with visual evidence [112].”
Beyond Manual Tuning of Hyperparameters (2015, neutral): “, [15, 23, 41, 57]).”
Anatomy of hierarchy: Feedforward and feedback pathways in macaque visual cortex (2013, background): “prediction error signals and descending predictions (Friston, 2010; Lee and Mumford, 2003).”
Towards a Mathematical Theory of Cortical Micro-circuits (2009, background): “Lee and Mumford [1] suggested that the formation of illusory contours is primarily a top-down mechanism, as also suggested in our experiments, in combination with lateral mechanisms proposed in [66].”