Abstract
In order to utilise the computing power offered by modern multi-core computer systems, APBHC, a new parallel search algorithm is proposed in this paper. This algorithm uses a number of parallel, asynchronous threads, each performing hill climbing independently of each other whilst sharing information about the best solution found so far amongst all threads. This information is used to adapt the maximum step size during the search. One advantage of this approach is that this new algorithm has no control parameters, which would require tuning. The other advantage is that it can make use of all processing cores available in a computer system. The new method was applied to the problem of Spice Model Generation. It was shown that it out-performs Genetic Algorithms (GA), which were applied to this problem in the past, without the need of time consuming parameter tuning.