Method and circuits to virtually increase the number of prototypes in artifical neural networks
First Claim
1. A device to virtually increase a number of prototypes of a neural network, comprising:
- a neural network formed by a plurality of neurons, each neuron having a prototype memory dedicated to store a prototype and an evaluator to generate a local result of an evaluation based on a comparison between an input pattern presented to the neuron and the prototype stored therein, wherein the neural network is adapted to generate, according to at least one specified parameter and the local results, a global result corresponding to the input pattern;
a memory coupled to the neural network and comprising a plurality of prototypes, a plurality of input patterns to be presented to the neural network, and an amount of storage space to store a predetermined number of global results, wherein each of a plurality of slices is associated with a plurality of the prototypes; and
a circuit coupled to the memory and configured to compare, according to predetermined criteria, a current global result from a current slice with a previously obtained global result from a previous slice, wherein both the current and previously obtained global result are determined from the same input pattern, and wherein the circuit can store a new global result in the memory based on the comparison.
1 Assignment
0 Petitions
Accused Products
Abstract
An improved Artificial Neural Network (ANN) is disclosed that comprises a conventional ANN, a database block, and a compare and update circuit. The conventional ANN is formed by a plurality of neurons, each neuron having a prototype memory dedicated to store a prototype and a distance evaluator to evaluate the distance between the input pattern presented to the ANN and the prototype stored therein. The database block has: all the prototypes arranged in slices, each slice being capable to store up to a maximum number of prototypes; the input patterns or queries to be presented to the ANN; and the distances resulting of the evaluation performed during the recognition/classification phase. The compare and update circuit compares the distance with the distance previously found for the same input pattern updates or not the distance previously stored.
-
Citations
12 Claims
-
1. A device to virtually increase a number of prototypes of a neural network, comprising:
-
a neural network formed by a plurality of neurons, each neuron having a prototype memory dedicated to store a prototype and an evaluator to generate a local result of an evaluation based on a comparison between an input pattern presented to the neuron and the prototype stored therein, wherein the neural network is adapted to generate, according to at least one specified parameter and the local results, a global result corresponding to the input pattern;
a memory coupled to the neural network and comprising a plurality of prototypes, a plurality of input patterns to be presented to the neural network, and an amount of storage space to store a predetermined number of global results, wherein each of a plurality of slices is associated with a plurality of the prototypes; and
a circuit coupled to the memory and configured to compare, according to predetermined criteria, a current global result from a current slice with a previously obtained global result from a previous slice, wherein both the current and previously obtained global result are determined from the same input pattern, and wherein the circuit can store a new global result in the memory based on the comparison. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. A method for increasing the number of prototypes to be processed in a neural network having a predetermined number of neurons, comprising:
-
a) partitioning a plurality of prototypes to be processed into a number of slices, wherein the number of prototypes is greater than the predetermined number of neurons of the neural network;
b) presenting one of a plurality of input patterns to the neural network for evaluation, the neural network producing a global result based on at least one specified parameter;
c) comparing, according to predetermined criteria, a current global result from a current slice with a previously obtained global result from a previous slice;
d) retaining, based on the predetermined criteria, either the current or previously obtained global result. e) repeating steps b), c) and d) until all the input patterns have been presented to the neural network; and
f) repeating steps b) to e) until all the slices have been loaded. - View Dependent Claims (12)
-
Specification