Avoiding excess computation in asynchronous evolutionary algorithms

Asynchronous evolutionary algorithms are becoming increasingly popular as a means of making full use of many processors while solving computationally expensive search and optimization problems. These algorithms excel at keeping large clusters fully utilized, but may sometimes inefficiently sample an excess of fast-evaluating solutions at the expense of higher-quality, slow-evaluating ones. We introduce a steady-state parent selection strategy, SWEET (“Selection whilE EvaluaTing”), that sometimes selects individuals that are still being evaluated and allows them to reproduce early. This gives slow-evaluating individuals that have higher fitnesses an increased ability to multiply in the population. We find that SWEET appears effective in simulated take-over time analysis, but that its benefit is confined mostly to early in the run, and our preliminary study on an autonomous vehicle controller problem that involves tuning a spiking neural network proves inconclusive.

[1]  L. Darrell Whitley,et al.  The GENITOR Algorithm and Selection Pressure: Why Rank-Based Allocation of Reproductive Trials is Best , 1989, ICGA.

[2]  Trevor N. Mudge,et al.  A Parallel Genetic Algorithm for Multiobjective Microprocessor Design , 1995, ICGA.

[3]  Blaise Agüera y Arcas,et al.  Communication-Efficient Learning of Deep Networks from Decentralized Data , 2016, AISTATS.

[4]  Edwin Lughofer,et al.  On the Performance of Master-Slave Parallelization Methods for Multi-Objective Evolutionary Algorithms , 2013, ICAISC.

[5]  Catherine D. Schuman,et al.  A Small, Low Cost Event-Driven Architecture for Spiking Neural Networks on FPGAs , 2020, ICONS.

[6]  Andrew Philippides,et al.  Tool sequence optimization using synchronous and asynchronous parallel multi-objective evolutionary algorithms with heterogeneous evaluations , 2013, 2013 IEEE Congress on Evolutionary Computation.

[7]  Domenico Pascarella,et al.  Simulation-Based Evolutionary Optimization of Air Traffic Management , 2020, IEEE Access.

[8]  Brian D. Davison,et al.  Effect of global parallelism on the behavior of a steady state genetic algorithm for design optimization , 1999, Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat. No. 99TH8406).

[9]  In-So Kweon,et al.  An Efficient Asynchronous Method for Integrating Evolutionary and Gradient-based Policy Search , 2020, NeurIPS.

[10]  Kenneth A. De Jong,et al.  Evaluation-Time Bias in Quasi-Generational and Steady-State Asynchronous Evolutionary Algorithms , 2016, GECCO.

[11]  Sean Luke,et al.  Distributed, Automated Calibration of Agent-based Model Parameters and Agent Behaviors , 2020, AAMAS.

[12]  Huzefa Rangwala,et al.  Asynchronous Online Federated Learning for Edge Devices with Non-IID Data , 2019, 2020 IEEE International Conference on Big Data (Big Data).

[13]  Rahul Mangharam,et al.  F1TENTH: An Open-source Evaluation Environment for Continuous Control and Reinforcement Learning , 2019, NeurIPS.

[14]  Zeyi Tao,et al.  A survey of federated learning for edge computing: Research problems and solutions , 2021, High-Confidence Computing.

[15]  Jinwoo Kim,et al.  Hierarchical asynchronous genetic algorithms for parallel/distributed simulation-based optimization , 1995 .

[16]  Ivan Garibay,et al.  Evolutionary model discovery of causal factors behind the socio-agricultural behavior of the ancestral Pueblo , 2018 .

[17]  Enrique Alba,et al.  Parallel Metaheuristics: A New Class of Algorithms , 2005 .

[18]  Matjaz Depolli,et al.  Asynchronous Master-Slave Parallelization of Differential Evolution for Multi-Objective Optimization , 2013, Evolutionary Computation.

[19]  Kenneth DeJong.  Evolutionary computation: a unified approach , 2007, GECCO.

[20]  Mark Coletti,et al.  Troubleshooting deep-learner training data problems using an evolutionary algorithm on Summit , 2020, IBM J. Res. Dev..

[21]  Jeffrey K. Bassett,et al.  Library for evolutionary algorithms in Python (LEAP) , 2020, GECCO Companion.

[22]  Enrique Alba,et al.  A study of master-slave approaches to parallelize NSGA-II , 2008, 2008 IEEE International Symposium on Parallel and Distributed Processing.

[23]  Tomohiro Harada,et al.  Mathematical Model of Asynchronous Parallel Evolutionary Algorithm to Analyze Influence of Evaluation Time Bias , 2019, 2019 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE).

[24]  Marc Schoenauer,et al.  Asynchronous Evolutionary Multi-Objective Algorithms with heterogeneous evaluation costs , 2011, 2011 IEEE Congress of Evolutionary Computation (CEC).

[25]  Huzefa Rangwala,et al.  FedAT: A High-Performance and Communication-Efficient Federated Learning System with Asynchronous Tiers , 2020, SC21: International Conference for High Performance Computing, Networking, Storage and Analysis.

[26]  Kenneth A. De Jong,et al.  Understanding Simple Asynchronous Evolutionary Algorithms , 2015, FOGA.

[27]  FedAT , 2021, Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis.