Scalable flow virtual learning neurocomputer
First Claim
1. A computer system comprising a virtual neural synapse processor apparatus including:
- an N physical neuron structure, wherein N is an integer greater than or equal to one, partitioned into groups includingmeans for synapse processing including means for storing instructions and data, means for receiving instructions and data, means for executing instructions, and means for interfacing with external data storage for weights and neuron output values for V neurons, where V is greater than N and V is a number of neurons to be emulated (virtual neurons) on the N neuron structure,means for executing group instructions, and communicating adder trees, said N neuron structure further includingneuron activation means for controlling a start of a neuron activation function,means for interconnecting said groups into the N neuron structure, andmeans for communicating instructions, data, and outputs of the neuron activation means back to input means for synapse processing through the communicating adder trees.
1 Assignment
0 Petitions
Accused Products
Abstract
A scalable flow virtual learning neurocomputer system and appratus with a scalable hybrid control flow/data flow employing a group partitioning algorithm, and a scalable virtual learning architecture, synapse processor architecture mapping, inner square folding and array separation, with capability of back propagation for virtual learning. The group partitioning algorithm creates a common building block of synapse processors containing their own external memory. The processor groups are used to create a general purpose virtual learning machine which maintains complete connectivity with high performance. The synapse processor group allows a system to be scalable in virtual size and direct execution capabilities. Internal to the processor group, the synapse processors are designed as a hybrid control flow/data flow architecture with external memory access and reduced synchronization problems.
40 Citations
40 Claims
-
1. A computer system comprising a virtual neural synapse processor apparatus including:
-
an N physical neuron structure, wherein N is an integer greater than or equal to one, partitioned into groups including means for synapse processing including means for storing instructions and data, means for receiving instructions and data, means for executing instructions, and means for interfacing with external data storage for weights and neuron output values for V neurons, where V is greater than N and V is a number of neurons to be emulated (virtual neurons) on the N neuron structure, means for executing group instructions, and communicating adder trees, said N neuron structure further including neuron activation means for controlling a start of a neuron activation function, means for interconnecting said groups into the N neuron structure, and means for communicating instructions, data, and outputs of the neuron activation means back to input means for synapse processing through the communicating adder trees. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
-
14. A neural synapse processor apparatus comprising a neural network emulation structure including:
-
partitioned groups comprised of at least one synapse processing unit including means for storing instructions and data, means for receiving instructions and data, means for executing instructions, means for interfacing with external storage units to communicate weights and neuron output values, means for executing group instructions, and communicating adder trees; neuron activation units; means for interconnecting the groups into a N physical neuron structure supporting the emulation of V neurons (virtual neurons), where N and V are positive integers and V is greater than or equal to N; and means for communicating instructions, data, and outputs of the neuron activation units back to the synapse processing units through the communicating adder trees; the synapse processing units each associated with a connection weight in a neutral network, the neural network in the form of a N by N matrix folded along a diagonal, including diagonal cells and general cells, each of the diagonal cells including one of the synapse processing units and each of the general cells including two of the synapse processing units. - View Dependent Claims (15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28)
-
-
29. A neural synapse processor apparatus comprising:
-
N2 synapse processing units; N communicating adder trees; a programmable processor controller; and N neuron activation units for a N physical neuron structure partitioned into G groups of H synapse processing units per group, each said group interfacing with means for external storage for providing storage capabilities for V virtual neurons, where V is greater than N, with N external summation switching mechanisms, where G, H, V, and N are positive integers. - View Dependent Claims (30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40)
-
Specification