Evolutionary Algorithms are powerful function optimizers, but suffer from premature convergence. Quantum-Inspired Evolutionary Algorithm (QEA) has been shown to be less prone to this on an important class of binary encoded problems. QEA uses Q-bits in place of ordinary bits, introducing a rational parameter into an otherwise binary search space. The essential feature of QEA is that the fitness of individuals in the population is defined stochastically by sampling from discrete points in the landscape. The probability of a particular point being sampled is based on the proximity of an individual to that point, where the individual represents a point in the solid hypercube spanned by the possible discrete solutions. This paper presents Probabilistically Interpolated Rational Hypercube Landscape Evolutionary Algorithm (PIRHLEA), which generalizes QEA by relaxing its two vestigial quantum mechanical attributes: quadratic and angular parameterization of probabilities and using single samples to determine fitness estimates of individuals. This is accomplished by replacing each Q-bit with a rational parameter between zero and one. Compared to QEA, PIRHLEA is simpler to code, more computationally efficient, and easier to visualize. PIRHLEA also permits multiple samples from points in the landscape to determine individuals' fitness.
- Hypercube Landscape,
- Landscape Sampling,
- Quantum Computing,
- Quantum-Inspired Evolutionary Algorithm
Available at: http://works.bepress.com/daniel-tauritz/58/