Recently, multi-core processors, which have more than one Central Processing Unit (CPU), are introduced widely into personal computers. Authors have been proposed a concurrent program of Differential Evolution (DE). The concurrent program of DE, which is called Parallelized DE (PDE), can generate and evaluate multiple individuals in parallel on a multi-core processor. In this paper, two implementation arts of PDE are presented and compared through the numerical experiment and the statistical test. © 2011 IEEE.

In this paper, a new Differential Evolution (DE) that has multiple populations, or islands, is proposed. The proposed DE is called Structured Differential Evolution (StDE). In order to generate a new individual from the current population, various characteristic strategies have been proposed for DE. However, the performances of these strategies depend on the kind of the optimization problem. The proposed StDE uses different strategies in respective islands. Therefore, it can be expected that the proposed StDE is effective for a wide range of optimization problems. Although various networks topologies among islands are reported for island-based evolutionary algorithms, the most popular ones, namely the ring network and the torus network, are employed by StDE. Furthermore, in order to enhance the performance of proposed StDE, various migration policies are examined in two kinds of networks though a variety of benchmark problems.

本論文では種々のネットワークに対する構造差分進化計算(Structued Differential Evolution, StDE)を提案する。
逐次進化計算(Sequential Differential Evolution, SqDE)は近年提案された進化計算(Evolutionary algorithm, EA)であり、SqDEは最適化問題を効率良く解く事ができる。
本論文で提案するStDEはSqDEを並列化したものである。
ベンチマーク問題に対する最適化問題において、ネットワークを用いたStDEはSqDEよりも解を高速に求めることができる。(英文)

Many of the conventional Differential Evolutions (DEs) have employed the discrete generation model that uses two populations, namely, old one and new one. Recently, a new DE based on the continuous generation model is proposed. In the continuous generation model, only one population is used. The new DE is sometimes called Sequential DE (SDE). Besides better convergence, SDE has some advantages. For instance, it becomes easy to introduce various survival selections into SDE. Therefore, four survival selections depending on the distance between two individuals are presented for SDE. Furthermore, in order to compare the effects of the distance dependent survival selections on SDE, not only the numerical experiment but also the statistical test is conducted on various benchmark problems.

進化計算の一種であるDEをMapReduceの概念に基づき並行プリグラムに拡張したCDEを考案し，マルチコア・プロセッサにおいて速度向上率を評価した．（英文）

Recently, general-purpose multi-core processors have been introduced widely into personal computers. In order to utilize the additional cores to execute costly application programs such as Evolutionary Algorithms (EAs), concurrent implementations of them are demanded. Even though EAs including various Differential Evolutions (DEs) are naturally prone to parallelism, Sequential DE (SDE) is especially suited for concurrent programming. Therefore, a concurrent implementation of SDE, which is based on the map and reduce framework, is proposed. Through the numerical experiment, the speedup of SDE due to the use of multiple cores is demonstrated. Furthermore, it is shown that the concurrent programming of SDE is efficient, simple, portable and scalable.

In this paper, a Structured Differential Evolution (StDE) that has multiple populations, or islands, is proposed. Since various characteristic strategies have been contrived for DE, the proposed StDE uses different strategies in respective islands. This technique is called mixed strategies. Therefore, it can be expected that the proposed StDE is effective for a wide range of optimization problems. Although various networks among islands are reported for island-based evolutionary algorithms, the most common one, namely the ring network is employed by StDE. However, in order to enhance the performance of proposed StDE, various migration policies are examined in the ring network though a variety of benchmark problems. © 2010 IEEE.

A structured implementation of Differential Evolution (DE), which can be executed in parallel by using various processor networks, is presented in this paper. Even though Evolutionary Algorithms (EAs) including DE have a parallel and distributed nature intrinsically, Sequential DE (SqDE) is especially suited for the structured implementation of DE. Therefore, the proposed Structured DE (StDE) is based on SqDE. Through the numerical experiment conducted on a variety of benchmark problems, the performances of StDE realized on some different network topologies are compared with the conventional SqDE that uses no processor network. As a result, it is shown that the number of generations spent by StDE to find optimal solutions is smaller than the number of them spent by the above SqDE in many benchmark problems. Therefore, the optimal solutions of almost of the benchmark problems are found more efficiently by using the proposed StDE realized on the processor network.

In this paper, we propose parallel algorithms to solve the selection problem on the Bulk-Synchronous Parallel (BSP) model and the BSP* model. The BSP and BSP* models are recently proposed parallel computation models. They can represent the communication cost, a vital element in the latest parallel computations, in terms of the parameters of the synchronization period L, the reciprocal g of the communication network bandwidth, and the packet size B. In this paper, we propose a parallel algorithm having an internal computation time O[ (n/p) + d log p log log n + L (log p log log n)/(log d)] and a communication time O[g(n/p) + (gd + L) (log p log log n)/(log d)], on the BSP model, and a parallel algorithm having an internal computation time O[ (n/p) + d log p log log n + L(log p log log n)/(log d)] and a communication time O[g(n/(pB)) + (n/p)1/7(log p)6/7 + (gd + L) (log p log log n)/(log d)] on the BSP* model for any integer d (1 ≤ d ≤ log n) to solve the problem of selecting n data with p processors.

In this paper, we present two parallel algorithms for computing the all nearest neighbors of an n x n binary image on the Bulk-Synchronous Parallel(BSP) model. The first algorithm is for weighted distance, and the second algorithm is for L-p distance. Both two algorithms run in O(n(2)/p + L) computation time and O(g n/root p + L) communication time using p (1 less than or equal to p less than or equal to n) processors and in O(n(2)/p + (d + L) log p/n/log(d+1)) computation time and in O(g n/root p + (gd + L) log p/n/log(d+1)) communication time using p (n < p less than or equal to n(2)) processors, for any integer d (1 less than or equal to d < p/n), where L denotes synchronization periodicity and g denotes a reciprocal of communication bandwidth.

プロセッサネットワーク上での混合戦略を用いた差分進化計算を提案し、いくつかのテスト関数に対して数値実験を行った．

本論文では種々のネットワークに対する構造差分進化計算(Structued Differential Evolution, StDE)を提案する。 逐次進化計算(Sequential Differential Evolution, SqDE)は近年提案された進化計算(Evolutionary algorithm, EA)であり、SqDEは最適化問題を効率良く解く事ができる。 本論文で提案するStDEはSqDEを並列化したものである。 ベンチマーク問題に対する最適化問題において、ネットワークを用いたStDEはSqDEよりも解を高速に求めることができる。(英文)

進化計算アルゴリズムの一種であるDifferential Evolutionを並行プログラムとしてマルチコア・プロセッサにより実装する技法を考案した。

本論文では種々のネットワークに対する構造差分進化計算(Structued Differential Evolution, StDE)を提案する。 逐次進化計算(Sequential Differential Evolution, SqDE)は近年提案された進化計算(Evolutionary algorithm, EA)であり、SqDEは最適化問題を効率良く解く事ができる。 本論文で提案するStDEはSqDEを並列化したものである。 ベンチマーク問題に対する最適化問題において、ネットワークを用いたStDEはSqDEよりも解を高速に求めることができる。(英文)

本稿ではp プロセッサを用いたHBSP モデル上で、サイズn × n の2 次元データに対し、各プロセッサの速度に応じてデータを割り当てるデータ分割法を示し、また、その分割法を用いた行列積を解く並列アルゴリズムを示す。

本稿ではpプロセッサを用いたHBSPモデル上で、サイズn×nの2次元データに対し、各プロセッサの速度に応じてデータを割り当てるデータ分割法を示し、また、その分割法を用いた行列積を解く並列アルゴリズムを示す。