# Israel needs national vision for AI or risks falling behind, tech authority says

Israel Innovation Authority urges government, academia, industry to join forces for advances in artificial intelligence as global race is underway… blah blah blah

“‘Artificial Intelligence’ is a term everyone throws around without understanding what it really means… true Artificial Intelligence is far from being developed (500 years?), what most people coin as ‘artificial intelligence’ is just advanced algorithms… it is not truly a ‘learning’ machine, a self aware and sentient being, and still nowhere what we perceive as ‘consciousness’

(and Quantum computers will NEVER become a reality)

…

… But I know of a guy!” 😀

(much of this has disappeared from the public domain for some strange reason)

Top Differences Between Artificial Intelligence, Machine Learning & Deep Learning

😀

Method and system for pattern recognition and processing

Dec 23, 1998

The present invention provides a method and system for pattern recognition and processing. Information representative of physical characteristics or representations of physical characteristics is transformed into a Fourier series in Fourier space within an input context of the physical characteristics that is encoded in time as delays corresponding to modulation of the Fourier series at corresponding frequencies. Associations are formed between Fourier series by filtering the Fourier series and by using a spectral similarity between the filtered Fourier series to determine the association based on Poissonian probability. The associated Fourier series are added to form strings of Fourier series. Each string is ordered by filtering it with multiple selected filters to form multiple time order formatted subset Fourier series, and by establishing the order through associations with one or more initially ordered strings to form an ordered string. Associations are formed between the ordered strings to form complex ordered strings that relate similar items of interest. The components of the invention are active based on probability using weighting factors based on activation rates.

Description

This application claims the benefit of U.S. provisional application Ser. No. 60/068,834, filed Dec. 24, 1997.

BACKGROUND OF THE INVENTION

Attempts have been made to create pattern recognition systems using programming and hardware. The state of the art includes neural nets. Neural nets typically comprise three layers—an input layer, a hidden layer, and an output layer. The hidden layer comprises a series of nodes which serve to perform a weighted sum of the input to form the output. Output for a given input is compared to the desired output, and a back projection of the errors is carried out on the hidden layer by changing the weighting factors at each node, and the process is reiterated until a tolerable result is obtained. The strategy of neural nets is analogous to the sum of least squares algorithms. These algorithms are adaptive to provide reasonable output to variations in input, but they can not create totally unanticipated useful output or discover associations between multiple inputs and outputs. Their usefulness to create novel conceptual content is limited; thus, advances in pattern recognition systems using neural nets is limited.

SUMMARY OF THE INVENTION

The present invention is directed to a method and system for pattern recognition and processing involving processing information in Fourier space.

The system of the present invention includes an Input Layer for receiving data representative of physical characteristics or representations of physical characteristics capable of transforming the data into a Fourier series in Fourier space. The data is received within an input context representative of the physical characteristics that is encoded in time as delays corresponding to modulation of the Fourier series at corresponding frequencies. The system includes a memory that maintains a set of initial ordered Fourier series. The system also includes an Association Layer that receives a plurality of the Fourier series in Fourier space including at least one ordered Fourier series from the memory and forms a string comprising a sum of the Fourier series and stores the string in memory. The system also includes a String Ordering Layer that receives the string from memory and orders the Fourier series contained in the string to form an ordered string and stores the ordered string in memory. The system also includes a Predominant Configuration Layer that receives multiple ordered strings from the memory, forms complex ordered strings comprising associations between the ordered strings, and stores the complex ordered strings to the memory. The components of the system are active based on probability using weighting factors based on activation rates.

Another aspect of the present invention is directed to ordering a string representing the information. This aspect of the invention utilizes a High Level Memory section of the memory that maintains an initial set of ordered Fourier series. This aspect of the invention includes obtaining a string from the memory and selecting at least two filters from a selected set of filters stored in the memory. This aspect also includes sampling the string with the filters such that each of the filters produce a sampled Fourier series. Each Fourier series comprises a subset of the string. This aspect also includes modulating each of the sampled Fourier series in Fourier space with the corresponding selected filter such that each of the filters produce an order formatted Fourier series. Furthermore, this aspect includes adding the order formatted Fourier series produced by each filter to form a summed Fourier series in Fourier space, obtaining an ordered Fourier series from the memory, determining a spectral similarity between the summed Fourier series and the ordered Fourier series, determining a probability expectation value based on the spectral similarity, and generating a probability operand having a value selected from a set of zero and one, based on the probability expectation value. These steps are repeated until the probability operand has a value of one. Once the probability operand has a value of one, this aspect includes storing the summed Fourier series to an intermediate memory section. Thereafter, this aspect includes removing the selected filters from the selected set of filters to form an updated set of filters, removing the subsets from the string to obtain an updated string, and selecting an updated filter from the updated set of filters. This aspect further includes sampling the updated string with the updated filter to produce a sampled Fourier series comprising a subset of the string, modulating the sampled Fourier series in Fourier space with the corresponding selected updated filter to produce an updated order formatted Fourier series, recalling the summed Fourier series from the intermediate memory section, adding the updated order formatted Fourier series to the summed Fourier series to form an updated summed Fourier series in Fourier space, and obtaining an updated ordered Fourier series from the memory. This aspect further includes determining a spectral similarity between the updated summed Fourier series and the updated ordered Fourier series, determining a probability expectation value based on the spectral similarity, and generating a probability operand having a value selected from a set of zero and one, based on the probability expectation value. These steps are repeated until the probability operand has a value of one or all of the updated filters have been selected from the updated set of filters. If all of the updated filters have been selected before the probability operand has a value of one, then clearing the intermediate memory section and repeating the steps starting with selecting at least two filters from a selected set of filters. Once the probability operand has a value of one, the updated summed Fourier series is stored to the intermediate memory section and steps beginning with removing the selected filters from the selected set of filters to form an updated set of filters are repeated until one of the following set of conditions is satisfied: the updated set of filters is empty or the remaining subsets of the string is nil. If the remaining subsets of the string is nil, then the Fourier series in the intermediate memory section is stored in the High Level Memory section of the memory.

Another aspect of the present invention is directed to forming complex ordered strings by forming associations between a plurality of ordered strings. This aspect of the invention includes recording ordered strings to the High Level Memory section, forming associations of the ordered strings to form complex ordered strings, and recording the complex ordered strings to the High Level Memory section. A further aspect of the invention is directed to forming a predominant configuration based on probability. This aspect of the invention includes generating an activation probability parameter, storing the activation probability parameter in the memory, generating an activation probability operand having a value selected from a set of zero and one, based on the activation probability parameter, activating any one or more components of the present invention such as matrices representing functions, data parameters, Fourier components, Fourier series, strings, ordered strings, components of the Input Layer, components of the Association Layer, components of the String Ordering Layer, and components of the Predominant Configuration Layer, the activation of each component being based on the corresponding activation probability parameter, and weighting each activation probability parameter based on an activation rate of each component.

Novel method and system for pattern recognition and processing using data encoded as Fourier series in Fourier space

Article in Engineering Applications of Artificial Intelligence 19(2):219-234 · March 2006

Abstract

A method and system for pattern recognition and processing is reported that has a data structure and theoretical basis that are unique. This novel approach anticipates the signal processing action of an ensemble of neurons as a unit and intends to simulate aspects of brain that give rise to capabilities such as intelligence, pattern recognition, and reasoning that have not been reproduced with past approaches such as neural networks that are based individual simulated ”neuronal units.” Information representative of physical characteristics or representations of physical characteristics is transformed into a Fourier series in Fourier space within an input context of the physical characteristics that is encoded in time as delays corresponding to modulation of the Fourier series at corresponding frequencies. Associations are formed between Fourier series by filtering the Fourier series and by using a spectral similarity between the filtered Fourier series to determine the association based on Poissonian probability. The associated Fourier series are added to form strings of Fourier series. Each string is ordered by filtering it with multiple selected filters to form multiple time order formatted subset Fourier series, and by establishing the order through associations with one or more initially ordered strings to form an ordered string. Associations are formed between the ordered strings to form complex ordered strings that relate similar items of interest. The components of the system based on the algorithm are active based on probability using weighting factors based on activation rates