Self-organizing sequential memory pattern machine and reinforcement learning method
First Claim
1. A self-organizing computing machine for mapping from a plurality of patterns contained within at least one predetermined set of provided inputs to at least one invariant perception distinguishable by a name or a label among a plurality of categories, wherein the self-organizing computing machine comprises:
- at least one network of at least three nodes interconnected by variable connections into at least two hierarchical node levels including at least a lower node level and a higher node level;
at least one feature extractor arranged to receive the at least one predetermined set of provided inputs, to process the at least one predetermined set of provided inputs to determine at least one hierarchical set of at least two correlants commensurate with at least two hierarchical correlant levels including at least a lower correlant level and a higher correlant level, and to communicate the determined hierarchical sets of at least two correlants to the at least two distinct nodes of the at least two distinct hierarchical node levels commensurate with the at least two correlants of the at least two distinct correlant levels such that the correlants of the lower correlant level communicate to the corresponding nodes of the lower node level and that the correlants of the higher correlant level communicates to the corresponding nodes of the higher node level; and
at least one output unit arranged to interface the at least one invariant perception distinguishable by a name, or a label, among the plurality of categories;
wherein, the at least one node at each hierarchical node level incorporate at least one reinforcement learning sub-network combined with at least one ensemble learning sub-network;
wherein, the at least one reinforcement learning sub-network has been arranged to receive the commensurate correlants of the hierarchical sets of at least two correlants, to determine a plurality of output values and to output the output values from the determined plurality of output values to the nodes of the higher node level and the nodes of the lower node level; and
wherein, the at least one ensemble learning sub-network has been arranged to receive and to combine at least one output value from the at least one node of the higher node level and to receive and to combine at least one output value from the at least one node of the lower node level.
8 Assignments
0 Petitions
Accused Products
Abstract
A self-organizing computing machine utilizes a method for mapping from a plurality of patterns contained within provided inputs to an invariant perception, distinguishable by a name or a label. The self-organizing computing machine includes a network of at least three nodes arranged in at least two hierarchical levels, at least one feature extractor, and at least one output unit arranged to interface the invariant perception. The nodes may include a reinforcement learning sub-network combined with an ensemble learning sub-network. The reinforcement learning sub-network may be arranged to receive at least two correlants, to determine a plurality of output values and to output the output values to the nodes of the higher level and the nodes of the lower level. Also, the ensemble learning sub-network may be arranged to receive and to combine output values from nodes of the higher level and nodes of the lower level.
-
Citations
29 Claims
-
1. A self-organizing computing machine for mapping from a plurality of patterns contained within at least one predetermined set of provided inputs to at least one invariant perception distinguishable by a name or a label among a plurality of categories, wherein the self-organizing computing machine comprises:
-
at least one network of at least three nodes interconnected by variable connections into at least two hierarchical node levels including at least a lower node level and a higher node level; at least one feature extractor arranged to receive the at least one predetermined set of provided inputs, to process the at least one predetermined set of provided inputs to determine at least one hierarchical set of at least two correlants commensurate with at least two hierarchical correlant levels including at least a lower correlant level and a higher correlant level, and to communicate the determined hierarchical sets of at least two correlants to the at least two distinct nodes of the at least two distinct hierarchical node levels commensurate with the at least two correlants of the at least two distinct correlant levels such that the correlants of the lower correlant level communicate to the corresponding nodes of the lower node level and that the correlants of the higher correlant level communicates to the corresponding nodes of the higher node level; and at least one output unit arranged to interface the at least one invariant perception distinguishable by a name, or a label, among the plurality of categories; wherein, the at least one node at each hierarchical node level incorporate at least one reinforcement learning sub-network combined with at least one ensemble learning sub-network; wherein, the at least one reinforcement learning sub-network has been arranged to receive the commensurate correlants of the hierarchical sets of at least two correlants, to determine a plurality of output values and to output the output values from the determined plurality of output values to the nodes of the higher node level and the nodes of the lower node level; and wherein, the at least one ensemble learning sub-network has been arranged to receive and to combine at least one output value from the at least one node of the higher node level and to receive and to combine at least one output value from the at least one node of the lower node level. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
-
14. A self-organizing computing process for mapping from a plurality of patterns contained within at least one predetermined set of provided inputs to at least one invariant perception distinguishable, by a name or a label, among a plurality of categories, the self-organizing computing process comprises:
-
a) providing at least one self-organizing computing machine incorporating at least one network of at least three nodes arranged in at least two hierarchical node levels including at least a lower node level and a higher node level;
at least one feature extractor for receiving the at least one predetermined set of provided inputs, processing the at least one predetermined set of provided inputs to determine a hierarchical set of at least two correlants commensurate with the at least two hierarchical correlant levels including at least a lower correlant level and a higher correlant level, and communicating the determined hierarchical sets of at least two correlants to the at least two distinct nodes of the at least two distinct hierarchical node levels commensurate with the at least two correlants of the at least two distinct correlant levels such that the correlants of the lower correlant level communicates to the corresponding nodes of the lower node level and that the correlants of the higher correlant level communicates to the corresponding nodes of the higher node level;
at least one output unit for interfacing the at least one output one invariant perception distinguishable, by a name or a label, among categories;
wherein, the at least one node at each hierarchical node level includes at least one reinforcement learning sub-network combined with at least one ensemble learning sub-network;
wherein, the at least one reinforcement learning sub-network have been arranged to receive the commensurate correlants of the hierarchical sets of at least two correlants, to determine a plurality of output values and to output the output values from the determined plurality of output values to the nodes of the higher node level nodes and the nodes of the lower node level; and
wherein, the at least one ensemble learning sub-network has been arranged to receive and to combine at least one output value from the at least one node of the higher node level and to receive and combine at least one output value from the at least one node of the lower node level;b) providing at least one predetermined initial set of inputs, to the at least one feature extractor and determining the hierarchical set of at least two correlants commensurate with the at least two hierarchical correlant levels; c) communicating the determined hierarchical sets of at least two correlants to the at least two distinct nodes of the at least two distinct hierarchical node levels commensurate with the at least two correlants such that the correlants of the lower correlant level communicate to the corresponding nodes of the lower node level and that the correlants of the higher correlant level communicate to the corresponding nodes of the higher node level; d) determining at least one output value from each of the at least two distinct nodes and providing the determined output values from each node to proximal nodes of the at least one network of the at least one self-organizing computing machine; e) after a predetermined time period providing at least another subsequent set of inputs, to the at least one feature extractor and determining the hierarchical set of at least two subsequent correlants commensurate with the at least two hierarchical correlant levels; f) communicating the determined hierarchical sets of at least two subsequent correlants to the at least two distinct nodes of the at least two distinct hierarchical node levels commensurate with the at least two subsequent correlants such that the correlants of the lower correlant level communicates to the corresponding nodes of the lower node level and that the correlants of the higher correlant level communicates to the corresponding nodes of the higher node level; g) determining at least one subsequent output value from each of the at least two distinct nodes and providing the determined subsequent output values from each node to proximal nodes of the at least one network of the at least one self-organizing computing machine; h) determining, based on the at least one subsequent output value of the at least one updated invariant perception distinguishable, by a name or a label, among categories; i) repeating sequentially steps c)-h) for another predetermined time period, or for a duration of time necessary to achieve a predetermined convergence of the at least one subsequent output value of a preselected node of the at least one network; j) interfacing the at least one updated invariant perception distinguishable, by a name or a label, among categories. - View Dependent Claims (15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29)
-
Specification