Towards a mathematical theory of cortical micro-circuits.
Bottom Line:
Anatomical data provide a contrasting set of organizational constraints.We describe the pattern recognition capabilities of HTM networks and demonstrate the application of the derived circuits for modeling the subjective contour effect.We also discuss how the theory and the circuit can be extended to explain cortical features that are not explained by the current model and describe testable predictions that can be derived from the model.
View Article:
PubMed Central - PubMed
Affiliation: Numenta Inc., Redwood City, California, United States of America. dgeorge@numenta.com
ABSTRACT
Show MeSH
The theoretical setting of hierarchical Bayesian inference is gaining acceptance as a framework for understanding cortical computation. In this paper, we describe how Bayesian belief propagation in a spatio-temporal hierarchical model, called Hierarchical Temporal Memory (HTM), can lead to a mathematical model for cortical circuits. An HTM node is abstracted using a coincidence detector and a mixture of Markov chains. Bayesian belief propagation equations for such an HTM node define a set of functional constraints for a neuronal implementation. Anatomical data provide a contrasting set of organizational constraints. The combination of these two constraints suggests a theoretically derived interpretation for many anatomical and physiological features and predicts several others. We describe the pattern recognition capabilities of HTM networks and demonstrate the application of the derived circuits for modeling the subjective contour effect. We also discuss how the theory and the circuit can be extended to explain cortical features that are not explained by the current model and describe testable predictions that can be derived from the model. |
Related In:
Results -
Collection
License getmorefigures.php?uid=PMC2749218&req=5
Mentions: We then tested the network on a Kanizsa square test pattern. Figure 14 shows the response of the network to the Kanizsa square test pattern. The network classifies this pattern as a rectangle, even though this type of pattern was not seen during training. We examined the network for the presence of illusory contour responses. Illusory contour responses are characterized by top-down activations with no bottom-up activation. We used the capability of Numenta's software to inspect the node states of a network to probe for illusory contour responses. Figure 15 shows the feed-forward and feedback inputs to nodes at 4 different locations. The subjective contour effect can be seen in Figure 15(C). There are no actual contours in the receptive field of this node. Therefore, the feed-forward input of this node is zero. However, the feedback input is nonzero because the network expects the edges of a rectangle. This is the subjective contour effect. |
View Article: PubMed Central - PubMed
Affiliation: Numenta Inc., Redwood City, California, United States of America. dgeorge@numenta.com