Neuroevolution is a machine learning method for evolving neural networks parameters and topology with a high degree of flexibility that makes them applicable to a wide range of architectures. Neuroevolution has been p...
详细信息
Neuroevolution is a machine learning method for evolving neural networks parameters and topology with a high degree of flexibility that makes them applicable to a wide range of architectures. Neuroevolution has been popular in reinforcement learning and has also shown to be promising for deep learning. The major feature of Bayesian optimisation is in reducing computational load by approximating the actual model with an acquisition function (surrogate model) that is computationally cheaper. A major limitation of neuroevolution is the high computational time required for convergence since learning (evo-lution) typically does not utilize gradient information. Bayesian optimisation, which is also known as surrogate-assisted optimisation, has been popular for expensive engineering optimisation problems and hyper-parameter tuning in machine learning. It has potential for training deep learning models via neuroevolution given large datasets and complex models. Recent advances in parallel and distributed computing have enabled efficient implementation of neuroevolution for complex and computationally expensive neural models. In this paper, we present a Bayesian optimisation framework for deep neu-roevolution using a distributed architecture to provide computational efficiency in training. Our results demonstrate promising results for simple to deep neural network models such as convolutional neural networks which motivates further applications. (c) 2021 Elsevier B.V. All rights reserved.
This work presents the EvoSpace model for the development of pool-based evolutionaryalgorithms (Pool-EA). Conceptually, the EvoSpace model is built around a central repository or population store, incorporating some ...
详细信息
This work presents the EvoSpace model for the development of pool-based evolutionaryalgorithms (Pool-EA). Conceptually, the EvoSpace model is built around a central repository or population store, incorporating some of the principles of the tuple-space model and adding additional features to tackle some of the issues associated with Pool-EAs;such as, work redundancy, starvation of the population pool, unreliability of connected clients or workers, and a large parameter space. The model is intended as a platform to develop search algorithms that take an opportunistic approach to computing, allowing the exploitation of freely available services over the Internet or volunteer computing resources within a local network. A comprehensive analysis of the model at both the conceptual and implementation levels is provided, evaluating performance based on efficiency, optima found and speedup, while providing a comparison with a standard EA and an island-based model. The issues of lost connections and system parametrization are studied and validated experimentally with encouraging results, that suggest how EvoSpace can be used to develop and implement different Pool-EAs for search and optimization.
Parallelization is becoming a more important issue for solving difficult optimization problems. Island models combine phases of independent evolution with migration where genetic information is spread out to neighbore...
详细信息
Parallelization is becoming a more important issue for solving difficult optimization problems. Island models combine phases of independent evolution with migration where genetic information is spread out to neighbored islands. This can lead to an increased diversity within the population. Despite many successful applications, the reasons behind the success of island models are not well understood. We perform a first rigorous runtime analysis for island models and construct a function where phases of independent evolution as well as communication among the islands are essential. A simple island model with migration finds a global optimum in polynomial time, while panmictic populations as well as island models without migration need exponential time, with very high probability. Our results lead to new insights into the usefulness of migration, how information is propagated in island models, and how to set parameters such as the migration interval. This is a novel contribution to the theoretical foundation of parallel EAs. Further, we provide empirical results that complement the theoretical results, investigate the robustness with respect to the choice of the migration interval and compare various migration topologies using statistical tests.
Estimation of distribution algorithms (EDAs) are one of the most promising paradigms in today's evolutionary computation. In this field, there has been an incipient activity in the so-called parallel estimation of...
详细信息
Estimation of distribution algorithms (EDAs) are one of the most promising paradigms in today's evolutionary computation. In this field, there has been an incipient activity in the so-called parallel estimation of distribution algorithms (pEDAs). One of these approaches is the distributed estimation of distribution algorithms (dEDAs). This paper introduces a new initialization mechanism for each of the populations of the islands based on the Voronoi cells. To analyze the results, a series of different experiments using the benchmark suite for the special session on Real-parameter Optimization of the IEEE CEC 2005 conference has been carried out. The results obtained suggest that the Voronoi initialization method considerably improves the performance obtained from a traditional uniform initialization.
Large-scale optimization with high dimensionality and high computational cost becomes ubiquitous nowadays. To tackle such challenging problems efficiently, devising distributedevolutionary computation algorithms is i...
详细信息
Large-scale optimization with high dimensionality and high computational cost becomes ubiquitous nowadays. To tackle such challenging problems efficiently, devising distributedevolutionary computation algorithms is imperative. To this end, this paper proposes a distributed swarm optimizer based on a special master-slave model. Specifically, in this distributed optimizer, the master is mainly responsible for communication with slaves, while each slave iterates a swarm to traverse the solution space. An asynchronous and adaptive communication strategy based on the request-response mechanism is especially devised to let the slaves communicate with the master efficiently. Particularly, the communication between the master and each slave is adaptively triggered during the iteration. To aid the slaves to search the space efficiently, an elite-guided learning strategy is especially designed via utilizing elite particles in the current swarm and historically best solutions found by different slaves to guide the update of particles. Together, this distributed optimizer asynchronously iterates multiple swarms to collaboratively seek the optimum in parallel. Extensive experiments on a widely used large-scale benchmark set substantiate that the distributed optimizer could: 1) achieve competitive effectiveness in terms of solution quality as compared to the state-of-the-art large-scale methods;2) accelerate the execution of the algorithm in comparison with the sequential one and obtain almost linear speedup as the number of cores increases;and 3) preserve a good scalability to solve higher dimensional problems.
This article proposes a distributed differential evolution which employs a novel self-adaptive scheme, namely scale factor inheritance. In the proposed algorithm, the population is distributed over several sub-populat...
详细信息
This article proposes a distributed differential evolution which employs a novel self-adaptive scheme, namely scale factor inheritance. In the proposed algorithm, the population is distributed over several sub-populations allocated according to a ring topology. Each sub-population is characterized by its own scale factor value. With a probabilistic criterion, that individual displaying the best performance is migrated to the neighbor population and replaces a pseudo-randomly selected individual of the target sub-population. The target sub-population inherits not only this individual but also the scale factor if it seems promising at the current stage of evolution. In addition, a perturbation mechanism enhances the exploration feature of the algorithm. The proposed algorithm has been run on a set of various test problems and then compared to two sequential differential evolution algorithms and three distributed differential evolution algorithms recently proposed in literature and representing state-of-the-art in the field. Numerical results show that the proposed approach seems very efficient for most of the analyzed problems, and outperforms all other algorithms considered in this study.
The dynamical properties of structures, such as natural frequencies, damping ratios and mode shapes, can be obtained by several identification methods. Some are based on the direct signal processing in a time domain;o...
详细信息
The dynamical properties of structures, such as natural frequencies, damping ratios and mode shapes, can be obtained by several identification methods. Some are based on the direct signal processing in a time domain;others transform response data to the frequency domain. The development of these techniques is useful in the production of more accurate structural models;they can be also used to test the level of damage in structures (or verify their strength to support new load actions) by using experimental data. There are situations where frequency domain algorithms and conventional system identification techniques fail, do not allow adequate solution of the identification problems or become trapped in a local optimum. It is in these cases where evolutionary optimization techniques are important tools for evaluating the dynamical properties of structural systems in practical applications. This article presents a methodology to determine the dynamic properties of structures knowing their response in terms of displacement, velocities or accelerations in the time domain when they are subjected to a free vibration excitation. In order to do that, a specialized evolutionary algorithm capable of adapting its parameters to the different types of registers obtained from the dynamic time response of a structure is implemented in a robust way, making this approach useful in practical situations. A distributed real genetic algorithm (DRGA) based on an island model of different subpopulations is used to adjust a simulated response signal of a building to the real response signal. Initially, computer-generated synthetic response signals are used but, in future, the approach will be validated with signals obtained from free vibration experimental tests and will be extended to other types of dynamical excitation signals. Finally, the method will be tested with data obtained from earthquake events.
DEAP is a novel evolutionary computation framework for rapid prototyping and testing of ideas. Its design departs from most other existing frameworks in that it seeks to make algorithms explicit and data structures tr...
详细信息
DEAP is a novel evolutionary computation framework for rapid prototyping and testing of ideas. Its design departs from most other existing frameworks in that it seeks to make algorithms explicit and data structures transparent, as opposed to the more common black-box frameworks. Freely available with extensive documentation at http://***, DEAP is an open source project under an LGPL license.
We propose in this paper to consider code-smells detection as a distributed optimization problem. The idea is that different methods are combined in parallel during the optimization process to find a consensus regardi...
详细信息
We propose in this paper to consider code-smells detection as a distributed optimization problem. The idea is that different methods are combined in parallel during the optimization process to find a consensus regarding the detection of code-smells. To this end, we used Parallel evolutionaryalgorithms (P-EA) where many evolutionaryalgorithms with different adaptations (fitness functions, solution representations, and change operators) are executed, in a parallel cooperative manner, to solve a common goal which is the detection of code-smells. An empirical evaluation to compare the implementation of our cooperative P-EA approach with random search, two single population-based approaches and two code-smells detection techniques that are not based on meta-heuristics search. The statistical analysis of the obtained results provides evidence to support the claim that cooperative P-EA is more efficient and effective than state of the art detection approaches based on a benchmark of nine large open source systems where more than 85 percent of precision and recall scores are obtained on a variety of eight different types of code-smells.
evolutionaryalgorithms provide gradient-free optimisation which is beneficial for models that have difficulty in obtaining gradients;for instance, geoscientific landscape evolution models. However, such models are at...
详细信息
evolutionaryalgorithms provide gradient-free optimisation which is beneficial for models that have difficulty in obtaining gradients;for instance, geoscientific landscape evolution models. However, such models are at times computationally expensive and even distributed swarm-based optimisation with parallel computing struggle. We can incorporate efficient strategies such as surrogate-assisted optimisation to address the challenges;however, implementing inter-process communication for surrogate-based model training is difficult. In this paper, we implement surrogate-based estimation of fitness evaluation in distributed swarm optimisation over a parallel computing architecture. We first test the framework on a set of benchmark optimisation problems and then apply to a geoscientifc model that features landscape evolution model. Our results demonstrate very promising results for benchmark functions and the Badlands landscape evolution model. We obtain a reduction in computationally time while retaining optimisation solution accuracy through the use of surrogates in a parallel computing environment. The major contribution of the paper is in the application of surrogate-based optimisation for geoscientific models which can in the future help in better understanding of paleoclimate and geomorphology.
暂无评论