N-gram based classification with associative processing unit
A system for N-gram classification in a field of interest via hyperdimensional computing includes an associative memory array and a controller. The associative memory array stores hyperdimensional vectors in rows of the array. The hyperdimensional vectors represent symbols in the field of interest and the array includes bit-line processors along portions of bit-lines of the array. The controller activates rows of the array to perform XNOR, permute, and add operations on the hyperdimensional vectors with the bit-line processors, to encode N-grams, having N symbols therein, to generate fingerprints of a portion of the field of interest from the N-grams, to store the fingerprints within the associative memory array, and to match an input sequence to one of the stored fingerprints.
1 . A system for N-gram classification in a field of interest via hyperdimensional computing, the system comprising:
an associative memory array, said array storing hyperdimensional vectors in vector rows of said array, the hyperdimensional vectors representing symbols in said field of interest, said array comprising a multiplicity of bit- lines connecting a column of cells, each of which effects at least one bit-line processor operating on activated cells in its column; and
a controller configured to activate at least two vector rows of said array to perform at least XNOR, and add operations on said hyperdimensional vectors in said activated vector rows with said bit-line processors, and to copy a temporary vector output of said bit-line processors into an interim row of said array, said controller to activate said interim row with respect to at least one other vector row,
said controller to perform said activate and copy operations to encode N- grams, having N symbols therein where N is greater than 2, to generate fingerprints of a portion of said field of interest from said N-grams, to store said fingerprints within said array, and to match an input sequence to one of said stored fingerprints.
2 . The system of claim 1 , wherein said field of interest is music.
3 . The system of claim 1 , said controller to store interim N-gram results of a first N-gram and to generate a later N-gram from said interim N-gram results of its previous N-gram.
4 . A method for N-gram classification in a field of interest via hyperdimensional computing, the method comprising:
storing hyperdimensional vectors in vector rows of an associative memory array, the hyperdimensional vectors representing symbols in said field of interest, the array comprising a multiplicity of bit-lines connecting a column of cells, each of which effects at least one bit-line processor operating on activated cells in its column;
activating at least two vector rows of said array to perform at least XNOR, and add operations on said hyperdimensional vectors in said activated vector rows with said bit-line processors,
copying a temporary vector output of said bit-line processors into an interim row of said array, said copying to activate said interim row with respect to at least one other vector row, and
performing said activating and copying to encode N-grams, having N symbols therein where N is greater than 2, to generate fingerprints of a portion of said field of interest from said N-grams, to store said fingerprints within said associative memory array, and to match an input sequence to one of said stored fingerprints.
5 . The method of claim 4 , wherein said field of interest is music.
6 . The method of claim 4 , and comprising storing interim N-gram results of a first N-gram and generating a later N-gram from said interim N-gram results of its previous N-gram.