EvAg: a scalable peer-to-peer evolutionary algorithm

Abstract

This paper studies the scalability of an Evolutionary Algorithm (EA) whose population is structured by means of a gossiping protocol and where the evolutionary operators act exclusively within the local neighborhoods. This makes the algorithm inherently suited for parallel execution in a peer-to-peer fashion which, in turn, offers great advantages when dealing with computationally expensive problems because distributed execution implies massive scalability. In this paper we show another advantage of this algorithm: We experimentally demonstrate that it scales up better than traditional alternatives even when executed in a sequential fashion. In particular, we analyze the behavior of several EAs on wellknown deceptive trap functions with varying sizes and levels of deceptiveness. The results show that the new EA requires smaller optimal population sizes and fewer fitness evaluations to reach solutions. The relative advantage of the new EA is more outstanding as problem hardness and size increase. In some cases the new algorithm reduces the computational efforts of the traditional EAs by several orders of magnitude.

Validating a Peer-to-Peer Evolutionary Algorithm

“…The experimental setup in this paper is based on the simulations performed in [8] on the scalability of the Evolvable Agent model when tackling trap functions [1]. In order to validate the model, we will try to reproduce such results in a parallel infrastructure.…”

“…To that aim, we consider the results published in [8] on the scalability of the Evolvable Agent model (i.e. a P2P EA model) in a simulated based environment.…”

The Seamless Peer and Cloud Evolution Framework

“…To handle the decisions of how many virtual machines are desired 17 , a virtual machine manager (VMM) has been implemented. The VMM watches the state of the sched- uler (contents of the queue, number of connected users, and current throughput), and decides the number of desired instances based on a specified criteria.…”

“…The best models obtained in each node are eventually fused offline to provide the user with an optimal meta-model for the regression task at hand (the system is, therefore, particularly suitable for solving large regression problems). Finally, an abstract model for scalable P2P evolutionary computation, named EvAg (Evolvable Agent), was proposed in [17], and a thorough analysis of the scalability of the system with respect to the problem size was provided. This model was further investigated in [16].…”

Oversized Populations and Cooperative Selection: Dealing with Massive Resources in Parallel Infrastructures

“…To that aim, we assume platforms in which the number of resources can be always considered sufficient (see e.g. [5]), i.e. large enough to allow a parallelized population to eventually converge to problem optima.…”