Evolution and Optimization for Evolvable Hardware with Gene Expression Programming

Evolutionary design of electronic circuits is an important aspect in the research of Evolvable HardWare (EHW).The main contribution of this paper includes: (1) Based on fusing the techniques in data mining, gene expression programming and traditional EHW, this paper proposes a two-phased EHW framework including evolution phase based on ETGP (Express Tree Genetic Programming) algorithm and optimization phase based on MFDC (Mining Frequency Digital Circuit) algorithm. (2) By extensive experiments show that the average number of generation needed for the sixth order polynomial regression is 442, and the number of generations for multiplier is 2292. The figure is 6 times faster than CGP (Certain Genetic Programming) and NEHF (Novel Evolvable Hardware Framework). As a result, it gives the optimization circuit of multiplier by MFDC that is more efficient than the traditional circuit.