**6. Conclusion**

In this chapter we did a summary outline of GAs and discussed some possible applications. We presented three extrinsic evolutionary designs of digital circuits at gate level using GAs.

Future research must be done in this area. Firstly it is important to find a better representation of the circuit in chromosomes, because complex functions need a great number of architecture bits, which directly influences the GA search space. EHW successfully succeeds only when fitness reaches 100% and in huge search spaces this condition may be not always possible. This is the main reason that for the time being the complexity of evolved circuits is so far small. In our opinion, conclusion drawn in the paper (Yao & Higuchi, 1999) is still available: "EHW research needs to address issues, such as scalability, online adaptation, generalization, circuit correctness, and potential risk of evolving hardware in a real physical environment. It is argued that a theoretical foundation of EHW should be established before rushing to large-scale EHW implementations".

Recently appeared the idea of hybridization of a GA with elements of quantum computation (Han & Kim, 2002; Han, 2003). We have proposed a new quantum inspired genetic algorithm (QIGA) considerably faster than other similar algorithms, based on the idea of introducing a new parameter, which we called the probability of collapse, and to initiate the collapse of the quantum chromosome in order to generate a conventional population of chromosomes from time to time, and not each generation, as usually is done. We believe that some improvements in this method may be found in a future research, by establishing of a new method of updating the quantum chromosome from the current generation to the next one. Finally, some hybridization techniques may be useful for new quantum inspired evolutionary algorithms. (Rubinstein, 2001) used Genetic Programming to evolve quantum circuits with various properties, and (Moore & Venayagamoorthy, 2005) has developed an algorithm inspired from quantum evolution and Particle Swarm to evolve conventional combinational logic circuits.

#### **7. References**

118 Bio-Inspired Computational Algorithms and Their Applications

Evolution of CGA, SCQGA, and QIGA

Parameter CGA SCQGA QIGA Global time 73.990 s 38.599 s 19.263 s Self time 2.447 s 1.417 s 1.390 s Evaluation time 59.561 s 31.536 s 11.750 s

CGA SCQGA QIGA

Calls of evaluation function 25200 19200 4836 Ratio between evaluation and global time 80.5 % 81.7 % 60.9 %

<sup>0</sup> <sup>50</sup> <sup>100</sup> <sup>150</sup> <sup>200</sup> <sup>250</sup> <sup>300</sup> <sup>70</sup>

Number of generations

Number of generations 300 300 300

Unfortunately, the number of successful runs in 300 generations is only in the order of 70% for CGA, and 60% for the rest two algorithms. It occurs due to the constraint that only 100% in fitness evaluation is accepted. In other applications, this constraint may be not critical.

In this chapter we did a summary outline of GAs and discussed some possible applications. We presented three extrinsic evolutionary designs of digital circuits at gate level using GAs. Future research must be done in this area. Firstly it is important to find a better representation of the circuit in chromosomes, because complex functions need a great number of architecture bits, which directly influences the GA search space. EHW successfully succeeds only when fitness reaches 100% and in huge search spaces this condition may be not always possible. This is the main reason that for the time being the

of 100%) 7 6 6

Fig. 10. The evolutions of CGA, SCQGA and QIGA

75

80

85

Fitness evaluation in %

90

95

100

Successful runnings in 10 attempts (with fitness

**6. Conclusion** 

Table 1. A comparison between CGA, SCQGA and IQGA


**Part 2** 

**New Applications of Genetic Algorithm** 

