Machine learning is a robust process by which a computer can discover characteristics of underlying data that enable it to create a model for making future predictions or classifications from new data. designing machi...
详细信息
Machine learning is a robust process by which a computer can discover characteristics of underlying data that enable it to create a model for making future predictions or classifications from new data. designing machine learning pipelines, unfortunately, is often as much an art as it is a science, requiring pairing of feature construction, feature selection, and learning methods, all with their own sets of parameters. No general machine learning pipeline solution exists; each dataset has unique characteristics that make a particular set of methods and parameters better suited to solving the problem than others. To respond to the challenge of machine learning pipeline design, the field of automated machine learning (autoML) has recently emerged. AutoML seeks to automate the often arduous work of a data scientist, so they can focus on the underlying meanings of the data and spend less time on the tedium of pipeline design and tuning. This dissertation adapts and applies genetic programming to the newly emergent field of automated machine learning. Genetic programming enables the artificial evolution of an algorithm through a nearly infinite search space that otherwise requires a randomized search. This dissertation shows that through the process of genetic programming, it is possible to produce machine learning pipelines, and the evolved pipelines can outperform those created by human researchers.
We live in interesting times - as individuals, as members of various communities and organisations, and as inhabitants of planet Earth, we face many challenges, ranging from climate change to resource limitations, fro...
详细信息
ISBN:
(纸本)9781450342063
We live in interesting times - as individuals, as members of various communities and organisations, and as inhabitants of planet Earth, we face many challenges, ranging from climate change to resource limitations, from market risks and uncertainties to complex diseases. To some extent, these challenges arise from the complexity of the systems we are dealing with and of the problems that arise from understanding, modelling and controlling these systems. As computing scientists and IT professionals, we have much to contribute: solving complex problems by means of computer systems, software and algorithms is an important part of what our field is about. In this talk, I will focus on one particular type of complexity that has been of central interest to the evolutionary computation community, to artificial intelligence and far beyond, namely computational complexity, and in particular, NP-hardness. I will investigate the question to which extent NP-hard problems are as formidable as is often thought, and present an overview of several directions of research that aim to characterise and improve the behaviour of cutting-edge algorithms for solving NP-hard problems in a pragmatic, yet principled way. For prominent problems ranging from propositional satisfiability (SAT) to TSP and from AI planning to mixed integer programming (MIP), I will demonstrate how automated analysis and design techniques can be used to model and enhance the performance characteristics of cutting-edge solvers, sharing some surprising insights along the way.
A meta-evolutionary framework called Differential Evolution Ensemble designer (DEED) has been proposed in this paper to automate the design of DE ensemble algorithms. Given the design components of DE ensembles and a ...
详细信息
A meta-evolutionary framework called Differential Evolution Ensemble designer (DEED) has been proposed in this paper to automate the design of DE ensemble algorithms. Given the design components of DE ensembles and a set of optimization problems, DEED evolves effective and robust DE ensemble designs. The design components of DE ensemble algorithms include population management, constituent algorithms in the ensemble, information mixing amongst the sub-populations in the ensemble and the numerical parameters associated with various aspects of the ensemble. DEED employs Dynamic Structured Grammatical Evolution (DSGE) as the meta-evolutionary algorithm. A Backus-Naur form (BNF) grammar has been developed in this paper to represent the design space of DE ensembles and used by DSGE to evolve DE ensemble designs. DEED has been employed to evolve DE ensemble designs for solving 30-dimensional CEC'17 benchmark functions. The evolved designs (both the best design as well as all the final evolved designs) have been validated on CEC'14 and CEC'17 functions at 10, 30 and 50 dimensions and on real-world numerical optimization problems in CEC'11 benchmark suite. The DEED evolved designs have also been tested against the state-of-the-art algorithm configurator -irace. The performance of DEED evolved ensemble designs have been observed to be very competitive against that of manually designed and tuned state-of-the-art DE ensemble algorithms in the literature. DEED has also been demonstrated to evolve both co-operative and competitive style DE ensembles. The simulation experiments demonstrate the effectiveness as well as robustness of the evolved ensemble designs and the reliability of DEED framework in consistently evolving effective DE ensemble designs.
Due to the large volume of requests and the need to speed up the provision of services, production companies are migrating from a single service center to distributed centers. To support this migration, it is necessar...
详细信息
Due to the large volume of requests and the need to speed up the provision of services, production companies are migrating from a single service center to distributed centers. To support this migration, it is necessary to make intelligence decisions that benefit from automatic design of search algorithms. Considering these, this paper addresses the distributed hybrid flow shop scheduling problem with multiprocessor tasks (DHFSP-MT) as an extension of the hybrid flow shop scheduling problem with multiprocessor tasks (HFSP-MT) to minimize the maximum completion time among distributed factories. To provide effective decision support, we apply a novel framework called conditional markov chain search (CMCS) to automate the generation of heuristics, which is presented for the first time in the distributed shop scheduling problem to the best of our knowledge. We express the HFSP-MT as a markov decision process (MDP) and solve it through a hybrid Q-learning-local search algorithm. By using the characteristics of the problem under study, we introduce two new concepts, weight and impact, which are used to develop an initial construction algorithm and two local search methods. To balance jobs between factories at runtime, we propose a load balancing method, which transfers selected jobs from certain source factories to destination factories. We compare the proposed CMCS with two state-of-the-art metaheuristic algorithms from the literature using publicly available benchmark instances. The computational results show that the proposed CMCS provides better performance than that of the existing algorithms on solving the considered DHFSP-MT.(c) 2023 Elsevier B.V. All rights reserved.
The automation of meta-heuristic algorithm configuration holds the utmost significance in evolutionary computation. A hierarchical reinforcement learning-aware hyper-heuristic algorithm with fitness landscape analysis...
详细信息
The automation of meta-heuristic algorithm configuration holds the utmost significance in evolutionary computation. A hierarchical reinforcement learning-aware hyper-heuristic algorithm with fitness landscape analysis (HRLHH) is proposed to flexibly configure the suitable algorithms under various optimization scenarios. Two kinds of fitness landscape analysis techniques improved based on specific problem characteristics construct the state spaces for hierarchical reinforcement learning. Among them, an adaptive classification based on dynamic ruggedness of information entropy is designed to discern the complexity of problems, which serves as the basis for decision-making actions in upper-layer space. Additionally, an online dispersion metric based on knowledge is further presented to distinguish the precise landscape features in lower-layer space. In light of the characteristics of the state spaces, the hierarchical action spaces composed of meta-heuristics with disparate exploration and exploitation are designed, and various action selection strategies are introduced. Taking into account the real-time environment and algorithm evolution behavior, dynamic reward mechanisms based on evolutionary success rate and population convergence rate are utilized to enhance search efficiency. The experimental results on the IEEE Congress on Evolutionary Computation (CEC) 2017, CEC 2014, and large-scale CEC 2013 test suites demonstrate that the proposed HRLHH exhibits superiority in terms of accuracy, stability, and convergence speed, and possesses strong generalization.
A Meta-evolutionary Selection of Constituents in Ensemble DE (MeSCEDE) framework is proposed in this paper to automate the design of high-level multi-population ensemble Differential Evolution (DE) algorithms. The aut...
详细信息
A Meta-evolutionary Selection of Constituents in Ensemble DE (MeSCEDE) framework is proposed in this paper to automate the design of high-level multi-population ensemble Differential Evolution (DE) algorithms. The automateddesign of high-level multi-population ensemble DE algorithms involves both automated selection of constituent DE algorithms for the ensemble and automated configuration of ensemble related parameters. Grammatical Evolution has been used as the meta-evolutionary algorithm in MeSCEDE to search the space of design choices so as to evolve effective ensemble design(s) for given problem(s). The simulation experiments carried out in this paper involve applying MeSCEDE to evolve ensemble DE designs for solving 30-dimensional CEC'17 benchmark functions. The effectiveness of the evolved designs are validated on 30 and 50-dimensional CEC'14 functions as well as on 22 real-world problem instances from CEC'11 benchmark suite. The MeSCEDE evolved designs have exhibited a competitive performance against state-of-the-art ensemble DE algorithms. In addition, the potential of MeSCEDE has been demonstrated against irace, a state-of-the-art algorithm configurator. All simulation experiments reiterate the potential of MeSCEDE towards evolving effective and robust ensemble DE designs.
暂无评论