Associative memory device and method based on wave propagation
First Claim
1. An associative memory device, comprising:
- an input device for accepting a set or a plurality of sets of input information;
one or a plurality of input buffers;
the input device being coupled to, or integrated into, the input buffer;
a means for transporting the input information to the input buffer;
the input buffer being adapted to transform the input information into corresponding sets of complex-valued wave fields as wave-modeled input data;
one or a plurality of recording structures, referred to as a cortex;
the input buffer being further adapted to propagate the wave-modeled input data to the cortex;
the cortex being adapted to associate the desired sets of wave-modeled input data through an invertible mathematical function or operation, thus creating one or a plurality of associations within the cortex;
the cortex being further adapted to store the associations in a distributed manner across a cortex surface with other previously stored associations;
the cortex being further adapted to form and store arbitrary linear combinations of associations;
an input device being operable to accept one or a plurality of retrieval prompts, referred to as a prompt;
the input buffer being operable to transform the prompt into a corresponding set of complex-valued wave fields, as a wave-modeled prompt;
the input buffer being operable to propagate the wave-modeled prompt to the cortex;
the cortex being operable to cause the wave-modeled prompt to mathematically operate upon the previously stored associations using a de-association operation that is the inverse of that originally used to form the associations;
the cortex being adapted to store the wave-modeled prompt with other previously stored associations and wave modeled prompts;
a means for the result of the mathematical de-association operation, representing a wave-modeled retrieval, to propagate from the cortex to one or a plurality of output buffers, said output buffer being referred to with the input buffer as the buffers;
the output buffer being operable to transform the wave-modeled retrieval into output data; and
a means for using control data to provide overall control of the associative memory device.
0 Assignments
0 Petitions
Accused Products
Abstract
An associative, or content-addressable, memory device (101) and method based on waves is described. In this invention, arbitrary inputs are written as patterns which are interpreted as values of complex waves, discretized or analog, on one or more buffers (102,104). Information is transported via wave propagation from the buffers (102,104) to a cortex (103) or to multiple cortices, where the patterns are (1) associated using a mathematical operation for storage purposes or (2) de-associated through the corresponding inverse operation for retrieval purposes. The present associative memory is shown to emulate important behavioral properties of the human brain, including higher-brain functions such as learning from experience, forming generalizations or abstractions, and autonomous operation.
-
Citations
78 Claims
-
1. An associative memory device, comprising:
-
an input device for accepting a set or a plurality of sets of input information;
one or a plurality of input buffers;
the input device being coupled to, or integrated into, the input buffer;
a means for transporting the input information to the input buffer;
the input buffer being adapted to transform the input information into corresponding sets of complex-valued wave fields as wave-modeled input data;
one or a plurality of recording structures, referred to as a cortex;
the input buffer being further adapted to propagate the wave-modeled input data to the cortex;
the cortex being adapted to associate the desired sets of wave-modeled input data through an invertible mathematical function or operation, thus creating one or a plurality of associations within the cortex;
the cortex being further adapted to store the associations in a distributed manner across a cortex surface with other previously stored associations;
the cortex being further adapted to form and store arbitrary linear combinations of associations;
an input device being operable to accept one or a plurality of retrieval prompts, referred to as a prompt;
the input buffer being operable to transform the prompt into a corresponding set of complex-valued wave fields, as a wave-modeled prompt;
the input buffer being operable to propagate the wave-modeled prompt to the cortex;
the cortex being operable to cause the wave-modeled prompt to mathematically operate upon the previously stored associations using a de-association operation that is the inverse of that originally used to form the associations;
the cortex being adapted to store the wave-modeled prompt with other previously stored associations and wave modeled prompts;
a means for the result of the mathematical de-association operation, representing a wave-modeled retrieval, to propagate from the cortex to one or a plurality of output buffers, said output buffer being referred to with the input buffer as the buffers;
the output buffer being operable to transform the wave-modeled retrieval into output data; and
a means for using control data to provide overall control of the associative memory device. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37)
-
-
38. An associative memory device, comprising:
-
one or a plurality of input devices;
one or a plurality of input buffers;
one or a plurality of recording structures, referred to as a cortex;
one or a plurality of output buffers, referred to with the input buffer as the buffers;
one or a plurality of output devices;
one or a plurality of overall system controllers;
one or a plurality of volatile memory units;
one or a plurality of permanent storage devices, and one or a plurality of processing units;
the input device operable to sense input information comprising physical or electronic stimuli of arbitrary nature and origin;
the input device being coupled to the input buffer and being further operable to transport the input information to the input buffer;
the input buffer operable to transform the input information into corresponding sets of complex-valued wave fields;
the input buffer being further operable to pre-process the input information and assign phase information;
the input buffer being further operable to propagate the sets of complex-valued wave fields representing the input information and referred to as wave-modeled input data to the cortex;
the cortex being operable to accept the wave-modeled input data;
the cortex being adapted to retain the wave-modeled input data from the input buffer in a distributed manner across a cortex surface;
the cortex being further operable to associate the desired sets of wave-modeled input data through an invertible mathematical function or operation, thus creating one or more associations within the cortex;
the cortex being further operable to create and store arbitrary linear combinations of associations;
the input device, when enabled, being further operable to sense additional sets of input information comprising physical or electronic stimuli or other input data to act as a prompt or prompts to make a retrieval or retrievals;
the input buffer being further operable to transform the prompting set of input information into corresponding sets of complex-valued wave fields, referred to as the wave-modeled prompt;
the input buffer being further operable to preprocess the prompting set of input information and assign phase information;
the input information to be retrieved from the cortex comprising any of the sets of physical or electronic stimuli stored in previously created associations, as determined by the operator;
the input buffer, when enabled, being further operable to propagate the wave-modeled prompt to the cortex;
the cortex being operable to accept a probe from the input buffer carried by a wave-modeled prompt using a de-association function or operation that is the inverse of that originally used to form the association;
the cortex being further operable to propagate the results of the de-association, representing a retrieval, to the output buffer;
the output buffer being adapted to receive, post-process, and export the retrieval to one or more output devices in human or machine-readable form. - View Dependent Claims (39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73)
-
-
74. A method of autonomous pattern recognition, comprising:
-
sensing physical or electronic stimuli with a sensor;
interfacing the output of the sensor to one or a plurality of input buffers;
converting the sensed physical or electronic stimuli at the input buffer into wave-modeled input data in the form of complex-valued wave fields;
contemporaneously pre-processing the input information and assigning phase values at the input buffer;
generating a unique internal identification pattern (IIP) based on a unique identification code assigned automatically or by a system operator to each specific set of input data;
storing the IIP in one or more IIP buffers;
linking each specific IIP to arbitrary sets of information such as data files to be recalled, sets of commands to be executed or communicated upon a positive identification, and threshold confidence levels required for execution of each particular command or sets of commands;
using wave propagation to transport the wave-modeled input data and IIPs from their respective buffers to one or a plurality of recording structures, referred to as a cortex;
associating the desired sets of wave-modeled input data and their respective IIPs through an invertible mathematical function or operation, thus creating one or more associations within the cortex;
retaining the associations in a distributed manner across the cortex;
creating and storing arbitrary linear combinations of associations;
inputting prompting data into the input buffer;
converting the prompting data at the input buffer into complex-valued wave fields as wave-modeled prompting data;
contemporaneously pre-processing the prompting data and assigning phase values;
using wave propagation to transport the wave-modeled prompting data to the cortex to prompt a response from the cortex;
mathematically operating upon the previously stored associations at the cortex using a de-association function that is the inverse of that originally used to form the associations;
propagating the result of the mathematical de-association operation, representing a retrieval, to one or more IIP buffers;
autonomously reading a retrieved pattern as accurately as possible;
deducing an identification code that could correspond to a previously created IIP;
generating a confidence estimate based on a relative certainty or uncertainty of the identification made;
recalling the associated files or other information related to the reported identification code and displaying the results in human or machine-readable form;
executing or communicating the appropriate commands based on the identification code, the reported confidence estimate, and the predefined confidence levels required for the relevant commands;
adaptively improving identification performance in response to feedback on the accuracy of past identifications, using the following equation for the case of false positive identifications; Cnew=Cold−
β
i<
Ri,Ij>where Cold is the original cortex value, Cnew is the revised cortex value, β
i is a positive weighting factor, Ri is the prompting pattern, and Ij is the incorrectly retrieved IIP and using the following equation for the case of false negative or overly weak positive identifications;Cnew=Cold+β
i<
Ri−
γ
iPi,Ii>where Cold is the original cortex value, Cnew is the revised cortex value, β
i is a positive weighting factor, Ri is the prompting pattern, Ii is the desired IIP, Pi is the original pattern associated with Ii, and γ
i is 0 for augmentation mode and 1 for replacement mode. - View Dependent Claims (75, 76, 77, 78)
-
Specification