Title :
A fault-tolerant digital artificial neuron
Author_Institution :
Dept. of Electr. Eng., Akron Univ., OH, USA
Abstract :
A simple fault-tolerant digital artificial neuron is introduced. Two digital implementations based on two different adders are examined. Reliability, fault coverage, and hardware redundancy analyses are carried out to characterize the proposed fault-tolerant digital neural module. These analyses reveal that, for a 114% increase in hardware, a 92.07% fault detection coverage and a 21.15% fault recovery coverage are attained for a four-input neural module.<>
Keywords :
adders; fault tolerant computing; neural nets; redundancy; reliability; adders; digital artificial neuron; fault coverage; fault recovery; fault-tolerant; hardware redundancy; reliability; Aging; Fault location; Fault tolerance; Hardware; Joining processes; Mechanical factors; Neural networks; Neurons; Process design; Redundancy;
Journal_Title :
Design & Test of Computers, IEEE
