The clever hybridization of quantum computing concepts and evolutionary algorithms (EAs) resulted in a new field called quantum-inspired evolutionary algorithms (QIEAs). Unlike traditional EAs, QIEAs employ quantum bits to adopt a probabilistic representation of the state of a feature in a given solution. This unprecedented feature enables them to achieve better diversity and perform global search, effectively yielding a tradeoff between exploration and exploitation. We conducted a comprehensive survey across various publishers and gathered 56 papers. We thoroughly analyzed these publications, focusing on the novelty elements and types of heuristics employed by the extant quantum-inspired evolutionary algorithms (QIEAs) proposed to solve the feature subset selection (FSS) problem. Importantly, we provided a detailed analysis of the different types of objective functions and popular quantum gates, i.e., rotation gates, employed throughout the literature. Additionally, we suggested several open research problems to attract the attention of the researchers.
{"title":"Quantum-Inspired Evolutionary Algorithms for Feature Subset Selection: A Comprehensive Survey","authors":"Yelleti Vivek, Vadlamani Ravi, P. Radha Krishna","doi":"arxiv-2407.17946","DOIUrl":"https://doi.org/arxiv-2407.17946","url":null,"abstract":"The clever hybridization of quantum computing concepts and evolutionary\u0000algorithms (EAs) resulted in a new field called quantum-inspired evolutionary\u0000algorithms (QIEAs). Unlike traditional EAs, QIEAs employ quantum bits to adopt\u0000a probabilistic representation of the state of a feature in a given solution.\u0000This unprecedented feature enables them to achieve better diversity and perform\u0000global search, effectively yielding a tradeoff between exploration and\u0000exploitation. We conducted a comprehensive survey across various publishers and\u0000gathered 56 papers. We thoroughly analyzed these publications, focusing on the\u0000novelty elements and types of heuristics employed by the extant\u0000quantum-inspired evolutionary algorithms (QIEAs) proposed to solve the feature\u0000subset selection (FSS) problem. Importantly, we provided a detailed analysis of\u0000the different types of objective functions and popular quantum gates, i.e.,\u0000rotation gates, employed throughout the literature. Additionally, we suggested\u0000several open research problems to attract the attention of the researchers.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"132 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777696","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The NKCS model was introduced to explore coevolutionary systems, that is, systems in which multiple species are closely interconnected. The fitness landscapes of the species are coupled to a controllable amount, where the underlying properties of the individual landscapes are also controllable. No previous work has explored the use of hierarchical control within the model. This paper explores the effects of using a confederation, based on Bookchins communalism, and a single point of global control. Significant changes in behaviour from the traditional model are seen across the parameter space.
{"title":"An NKCS Model of Bookchins Communalism","authors":"Larry Bull","doi":"arxiv-2407.18218","DOIUrl":"https://doi.org/arxiv-2407.18218","url":null,"abstract":"The NKCS model was introduced to explore coevolutionary systems, that is,\u0000systems in which multiple species are closely interconnected. The fitness\u0000landscapes of the species are coupled to a controllable amount, where the\u0000underlying properties of the individual landscapes are also controllable. No\u0000previous work has explored the use of hierarchical control within the model.\u0000This paper explores the effects of using a confederation, based on Bookchins\u0000communalism, and a single point of global control. Significant changes in\u0000behaviour from the traditional model are seen across the parameter space.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"37 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Co-evolutionary algorithms (CoEAs), which pair candidate designs with test cases, are frequently used in adversarial optimisation, particularly for binary test-based problems where designs and tests yield binary outcomes. The effectiveness of designs is determined by their performance against tests, and the value of tests is based on their ability to identify failing designs, often leading to more sophisticated tests and improved designs. However, CoEAs can exhibit complex, sometimes pathological behaviours like disengagement. Through runtime analysis, we aim to rigorously analyse whether CoEAs can efficiently solve test-based adversarial optimisation problems in an expected polynomial runtime. This paper carries out the first rigorous runtime analysis of $(1,lambda)$ CoEA for binary test-based adversarial optimisation problems. In particular, we introduce a binary test-based benchmark problem called Diagonal problem and initiate the first runtime analysis of competitive CoEA on this problem. The mathematical analysis shows that the $(1,lambda)$-CoEA can efficiently find an $varepsilon$ approximation to the optimal solution of the Diagonal problem, i.e. in expected polynomial runtime assuming sufficiently low mutation rates and large offspring population size. On the other hand, the standard $(1,lambda)$-EA fails to find an $varepsilon$ approximation to the optimal solution of the Diagonal problem in polynomial runtime. This suggests the promising potential of coevolution for solving binary adversarial optimisation problems.
{"title":"Overcoming Binary Adversarial Optimisation with Competitive Coevolution","authors":"Per Kristian Lehre, Shishen Lin","doi":"arxiv-2407.17875","DOIUrl":"https://doi.org/arxiv-2407.17875","url":null,"abstract":"Co-evolutionary algorithms (CoEAs), which pair candidate designs with test\u0000cases, are frequently used in adversarial optimisation, particularly for binary\u0000test-based problems where designs and tests yield binary outcomes. The\u0000effectiveness of designs is determined by their performance against tests, and\u0000the value of tests is based on their ability to identify failing designs, often\u0000leading to more sophisticated tests and improved designs. However, CoEAs can\u0000exhibit complex, sometimes pathological behaviours like disengagement. Through\u0000runtime analysis, we aim to rigorously analyse whether CoEAs can efficiently\u0000solve test-based adversarial optimisation problems in an expected polynomial\u0000runtime. This paper carries out the first rigorous runtime analysis of $(1,lambda)$\u0000CoEA for binary test-based adversarial optimisation problems. In particular, we\u0000introduce a binary test-based benchmark problem called Diagonal problem and\u0000initiate the first runtime analysis of competitive CoEA on this problem. The\u0000mathematical analysis shows that the $(1,lambda)$-CoEA can efficiently find an\u0000$varepsilon$ approximation to the optimal solution of the Diagonal problem,\u0000i.e. in expected polynomial runtime assuming sufficiently low mutation rates\u0000and large offspring population size. On the other hand, the standard\u0000$(1,lambda)$-EA fails to find an $varepsilon$ approximation to the optimal\u0000solution of the Diagonal problem in polynomial runtime. This suggests the\u0000promising potential of coevolution for solving binary adversarial optimisation\u0000problems.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"56 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777697","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ali Hummos, Felipe del Río, Brabeeba Mien Wang, Julio Hurtado, Cristian B. Calderon, Guangyu Robert Yang
Humans and many animals show remarkably adaptive behavior and can respond differently to the same input depending on their internal goals. The brain not only represents the intermediate abstractions needed to perform a computation but also actively maintains a representation of the computation itself (task abstraction). Such separation of the computation and its abstraction is associated with faster learning, flexible decision-making, and broad generalization capacity. We investigate if such benefits might extend to neural networks trained with task abstractions. For such benefits to emerge, one needs a task inference mechanism that possesses two crucial abilities: First, the ability to infer abstract task representations when no longer explicitly provided (task inference), and second, manipulate task representations to adapt to novel problems (task recomposition). To tackle this, we cast task inference as an optimization problem from a variational inference perspective and ground our approach in an expectation-maximization framework. We show that gradients backpropagated through a neural network to a task representation layer are an efficient heuristic to infer current task demands, a process we refer to as gradient-based inference (GBI). Further iterative optimization of the task representation layer allows for recomposing abstractions to adapt to novel situations. Using a toy example, a novel image classifier, and a language model, we demonstrate that GBI provides higher learning efficiency and generalization to novel tasks and limits forgetting. Moreover, we show that GBI has unique advantages such as preserving information for uncertainty estimation and detecting out-of-distribution samples.
{"title":"Gradient-based inference of abstract task representations for generalization in neural networks","authors":"Ali Hummos, Felipe del Río, Brabeeba Mien Wang, Julio Hurtado, Cristian B. Calderon, Guangyu Robert Yang","doi":"arxiv-2407.17356","DOIUrl":"https://doi.org/arxiv-2407.17356","url":null,"abstract":"Humans and many animals show remarkably adaptive behavior and can respond\u0000differently to the same input depending on their internal goals. The brain not\u0000only represents the intermediate abstractions needed to perform a computation\u0000but also actively maintains a representation of the computation itself (task\u0000abstraction). Such separation of the computation and its abstraction is\u0000associated with faster learning, flexible decision-making, and broad\u0000generalization capacity. We investigate if such benefits might extend to neural\u0000networks trained with task abstractions. For such benefits to emerge, one needs\u0000a task inference mechanism that possesses two crucial abilities: First, the\u0000ability to infer abstract task representations when no longer explicitly\u0000provided (task inference), and second, manipulate task representations to adapt\u0000to novel problems (task recomposition). To tackle this, we cast task inference\u0000as an optimization problem from a variational inference perspective and ground\u0000our approach in an expectation-maximization framework. We show that gradients\u0000backpropagated through a neural network to a task representation layer are an\u0000efficient heuristic to infer current task demands, a process we refer to as\u0000gradient-based inference (GBI). Further iterative optimization of the task\u0000representation layer allows for recomposing abstractions to adapt to novel\u0000situations. Using a toy example, a novel image classifier, and a language\u0000model, we demonstrate that GBI provides higher learning efficiency and\u0000generalization to novel tasks and limits forgetting. Moreover, we show that GBI\u0000has unique advantages such as preserving information for uncertainty estimation\u0000and detecting out-of-distribution samples.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"161 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hossein Nematzadeh, Joseph Mani, Zahra Nematzadeh, Ebrahim Akbari, Radziah Mohamad
Feature selection poses a challenge in small-sample high-dimensional datasets, where the number of features exceeds the number of observations, as seen in microarray, gene expression, and medical datasets. There isn't a universally optimal feature selection method applicable to any data distribution, and as a result, the literature consistently endeavors to address this issue. One recent approach in feature selection is termed frequency-based feature selection. However, existing methods in this domain tend to overlook feature values, focusing solely on the distribution in the response variable. In response, this paper introduces the Distance-based Mutual Congestion (DMC) as a filter method that considers both the feature values and the distribution of observations in the response variable. DMC sorts the features of datasets, and the top 5% are retained and clustered by KMeans to mitigate multicollinearity. This is achieved by randomly selecting one feature from each cluster. The selected features form the feature space, and the search space for the Genetic Algorithm with Adaptive Rates (GAwAR) will be approximated using this feature space. GAwAR approximates the combination of the top 10 features that maximizes prediction accuracy within a wrapper scheme. To prevent premature convergence, GAwAR adaptively updates the crossover and mutation rates. The hybrid DMC-GAwAR is applicable to binary classification datasets, and experimental results demonstrate its superiority over some recent works. The implementation and corresponding data are available at https://github.com/hnematzadeh/DMC-GAwAR
{"title":"Distance-based mutual congestion feature selection with genetic algorithm for high-dimensional medical datasets","authors":"Hossein Nematzadeh, Joseph Mani, Zahra Nematzadeh, Ebrahim Akbari, Radziah Mohamad","doi":"arxiv-2407.15611","DOIUrl":"https://doi.org/arxiv-2407.15611","url":null,"abstract":"Feature selection poses a challenge in small-sample high-dimensional\u0000datasets, where the number of features exceeds the number of observations, as\u0000seen in microarray, gene expression, and medical datasets. There isn't a\u0000universally optimal feature selection method applicable to any data\u0000distribution, and as a result, the literature consistently endeavors to address\u0000this issue. One recent approach in feature selection is termed frequency-based\u0000feature selection. However, existing methods in this domain tend to overlook\u0000feature values, focusing solely on the distribution in the response variable.\u0000In response, this paper introduces the Distance-based Mutual Congestion (DMC)\u0000as a filter method that considers both the feature values and the distribution\u0000of observations in the response variable. DMC sorts the features of datasets,\u0000and the top 5% are retained and clustered by KMeans to mitigate\u0000multicollinearity. This is achieved by randomly selecting one feature from each\u0000cluster. The selected features form the feature space, and the search space for\u0000the Genetic Algorithm with Adaptive Rates (GAwAR) will be approximated using\u0000this feature space. GAwAR approximates the combination of the top 10 features\u0000that maximizes prediction accuracy within a wrapper scheme. To prevent\u0000premature convergence, GAwAR adaptively updates the crossover and mutation\u0000rates. The hybrid DMC-GAwAR is applicable to binary classification datasets,\u0000and experimental results demonstrate its superiority over some recent works.\u0000The implementation and corresponding data are available at\u0000https://github.com/hnematzadeh/DMC-GAwAR","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"66 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777741","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The current landscape of massive production industries is undergoing significant transformations driven by emerging customer trends and new smart manufacturing technologies. One such change is the imperative to implement mass customization, wherein products are tailored to individual customer specifications while still ensuring cost efficiency through large-scale production processes. These shifts can profoundly impact various facets of the industry. This study focuses on the necessary adaptations in shop-floor production planning. Specifically, it proposes the use of efficient evolutionary algorithms to tackle the flowshop with missing operations, considering different optimization objectives: makespan, weighted total tardiness, and total completion time. An extensive computational experimentation is conducted across a range of realistic instances, encompassing varying numbers of jobs, operations, and probabilities of missing operations. The findings demonstrate the competitiveness of the proposed approach and enable the identification of the most suitable evolutionary algorithms for addressing this problem. Additionally, the impact of the probability of missing operations on optimization objectives is discussed.
{"title":"Enhancing Mass Customization Manufacturing: Multiobjective Metaheuristic Algorithms for flow shop Production in Smart Industry","authors":"Diego Rossit, Daniel Rossit, Sergio Nesmachnow","doi":"arxiv-2407.15802","DOIUrl":"https://doi.org/arxiv-2407.15802","url":null,"abstract":"The current landscape of massive production industries is undergoing\u0000significant transformations driven by emerging customer trends and new smart\u0000manufacturing technologies. One such change is the imperative to implement mass\u0000customization, wherein products are tailored to individual customer\u0000specifications while still ensuring cost efficiency through large-scale\u0000production processes. These shifts can profoundly impact various facets of the\u0000industry. This study focuses on the necessary adaptations in shop-floor\u0000production planning. Specifically, it proposes the use of efficient\u0000evolutionary algorithms to tackle the flowshop with missing operations,\u0000considering different optimization objectives: makespan, weighted total\u0000tardiness, and total completion time. An extensive computational\u0000experimentation is conducted across a range of realistic instances,\u0000encompassing varying numbers of jobs, operations, and probabilities of missing\u0000operations. The findings demonstrate the competitiveness of the proposed\u0000approach and enable the identification of the most suitable evolutionary\u0000algorithms for addressing this problem. Additionally, the impact of the\u0000probability of missing operations on optimization objectives is discussed.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Neural architecture search (NAS) enables re-searchers to automatically explore vast search spaces and find efficient neural networks. But NAS suffers from a key bottleneck, i.e., numerous architectures need to be evaluated during the search process, which requires a lot of computing resources and time. In order to improve the efficiency of NAS, a series of methods have been proposed to reduce the evaluation time of neural architectures. However, they are not efficient enough and still only focus on the accuracy of architectures. In addition to the classification accuracy, more efficient and smaller network architectures are required in real-world applications. To address the above problems, we propose the SMEM-NAS, a pairwise com-parison relation-assisted multi-objective evolutionary algorithm based on a multi-population mechanism. In the SMEM-NAS, a surrogate model is constructed based on pairwise compari-son relations to predict the accuracy ranking of architectures, rather than the absolute accuracy. Moreover, two populations cooperate with each other in the search process, i.e., a main population guides the evolution, while a vice population expands the diversity. Our method aims to provide high-performance models that take into account multiple optimization objectives. We conduct a series of experiments on the CIFAR-10, CIFAR-100 and ImageNet datasets to verify its effectiveness. With only a single GPU searching for 0.17 days, competitive architectures can be found by SMEM-NAS which achieves 78.91% accuracy with the MAdds of 570M on the ImageNet. This work makes a significant advance in the important field of NAS.
{"title":"A Pairwise Comparison Relation-assisted Multi-objective Evolutionary Neural Architecture Search Method with Multi-population Mechanism","authors":"Yu Xue, Chenchen Zhu, MengChu Zhou, Mohamed Wahib, Moncef Gabbouj","doi":"arxiv-2407.15600","DOIUrl":"https://doi.org/arxiv-2407.15600","url":null,"abstract":"Neural architecture search (NAS) enables re-searchers to automatically\u0000explore vast search spaces and find efficient neural networks. But NAS suffers\u0000from a key bottleneck, i.e., numerous architectures need to be evaluated during\u0000the search process, which requires a lot of computing resources and time. In\u0000order to improve the efficiency of NAS, a series of methods have been proposed\u0000to reduce the evaluation time of neural architectures. However, they are not\u0000efficient enough and still only focus on the accuracy of architectures. In\u0000addition to the classification accuracy, more efficient and smaller network\u0000architectures are required in real-world applications. To address the above\u0000problems, we propose the SMEM-NAS, a pairwise com-parison relation-assisted\u0000multi-objective evolutionary algorithm based on a multi-population mechanism.\u0000In the SMEM-NAS, a surrogate model is constructed based on pairwise compari-son\u0000relations to predict the accuracy ranking of architectures, rather than the\u0000absolute accuracy. Moreover, two populations cooperate with each other in the\u0000search process, i.e., a main population guides the evolution, while a vice\u0000population expands the diversity. Our method aims to provide high-performance\u0000models that take into account multiple optimization objectives. We conduct a\u0000series of experiments on the CIFAR-10, CIFAR-100 and ImageNet datasets to\u0000verify its effectiveness. With only a single GPU searching for 0.17 days,\u0000competitive architectures can be found by SMEM-NAS which achieves 78.91%\u0000accuracy with the MAdds of 570M on the ImageNet. This work makes a significant\u0000advance in the important field of NAS.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"62 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777738","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Energy based models (EBMs) are appealing for their generality and simplicity in data likelihood modeling, but have conventionally been difficult to train due to the unstable and time-consuming implicit MCMC sampling during contrastive divergence training. In this paper, we present a novel energy-based generative framework, Variational Potential Flow (VAPO), that entirely dispenses with implicit MCMC sampling and does not rely on complementary latent models or cooperative training. The VAPO framework aims to learn a potential energy function whose gradient (flow) guides the prior samples, so that their density evolution closely follows an approximate data likelihood homotopy. An energy loss function is then formulated to minimize the Kullback-Leibler divergence between density evolution of the flow-driven prior and the data likelihood homotopy. Images can be generated after training the potential energy, by initializing the samples from Gaussian prior and solving the ODE governing the potential flow on a fixed time interval using generic ODE solvers. Experiment results show that the proposed VAPO framework is capable of generating realistic images on various image datasets. In particular, our proposed framework achieves competitive FID scores for unconditional image generation on the CIFAR-10 and CelebA datasets.
{"title":"Variational Potential Flow: A Novel Probabilistic Framework for Energy-Based Generative Modelling","authors":"Junn Yong Loo, Michelle Adeline, Arghya Pal, Vishnu Monn Baskaran, Chee-Ming Ting, Raphael C. -W. Phan","doi":"arxiv-2407.15238","DOIUrl":"https://doi.org/arxiv-2407.15238","url":null,"abstract":"Energy based models (EBMs) are appealing for their generality and simplicity\u0000in data likelihood modeling, but have conventionally been difficult to train\u0000due to the unstable and time-consuming implicit MCMC sampling during\u0000contrastive divergence training. In this paper, we present a novel energy-based\u0000generative framework, Variational Potential Flow (VAPO), that entirely\u0000dispenses with implicit MCMC sampling and does not rely on complementary latent\u0000models or cooperative training. The VAPO framework aims to learn a potential\u0000energy function whose gradient (flow) guides the prior samples, so that their\u0000density evolution closely follows an approximate data likelihood homotopy. An\u0000energy loss function is then formulated to minimize the Kullback-Leibler\u0000divergence between density evolution of the flow-driven prior and the data\u0000likelihood homotopy. Images can be generated after training the potential\u0000energy, by initializing the samples from Gaussian prior and solving the ODE\u0000governing the potential flow on a fixed time interval using generic ODE\u0000solvers. Experiment results show that the proposed VAPO framework is capable of\u0000generating realistic images on various image datasets. In particular, our\u0000proposed framework achieves competitive FID scores for unconditional image\u0000generation on the CIFAR-10 and CelebA datasets.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"65 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777747","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jose Guadalupe Hernandez, Anil Kumar Saini, Jason H. Moore
Lexicase selection is a successful parent selection method in genetic programming that has outperformed other methods across multiple benchmark suites. Unlike other selection methods that require explicit parameters to function, such as tournament size in tournament selection, lexicase selection does not. However, if evolutionary parameters like population size and number of generations affect the effectiveness of a selection method, then lexicase's performance may also be impacted by these `hidden' parameters. Here, we study how these hidden parameters affect lexicase's ability to exploit gradients and maintain specialists using diagnostic metrics. By varying the population size with a fixed evaluation budget, we show that smaller populations tend to have greater exploitation capabilities, whereas larger populations tend to maintain more specialists. We also consider the effect redundant test cases have on specialist maintenance, and find that high redundancy may hinder the ability to optimize and maintain specialists, even for larger populations. Ultimately, we highlight that population size, evaluation budget, and test cases must be carefully considered for the characteristics of the problem being solved.
{"title":"Lexicase Selection Parameter Analysis: Varying Population Size and Test Case Redundancy with Diagnostic Metrics","authors":"Jose Guadalupe Hernandez, Anil Kumar Saini, Jason H. Moore","doi":"arxiv-2407.15056","DOIUrl":"https://doi.org/arxiv-2407.15056","url":null,"abstract":"Lexicase selection is a successful parent selection method in genetic\u0000programming that has outperformed other methods across multiple benchmark\u0000suites. Unlike other selection methods that require explicit parameters to\u0000function, such as tournament size in tournament selection, lexicase selection\u0000does not. However, if evolutionary parameters like population size and number\u0000of generations affect the effectiveness of a selection method, then lexicase's\u0000performance may also be impacted by these `hidden' parameters. Here, we study\u0000how these hidden parameters affect lexicase's ability to exploit gradients and\u0000maintain specialists using diagnostic metrics. By varying the population size\u0000with a fixed evaluation budget, we show that smaller populations tend to have\u0000greater exploitation capabilities, whereas larger populations tend to maintain\u0000more specialists. We also consider the effect redundant test cases have on\u0000specialist maintenance, and find that high redundancy may hinder the ability to\u0000optimize and maintain specialists, even for larger populations. Ultimately, we\u0000highlight that population size, evaluation budget, and test cases must be\u0000carefully considered for the characteristics of the problem being solved.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"36 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777739","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Manuel Eberhardinger, Florian Rupp, Johannes Maucher, Setareh Maghsudi
Despite tremendous progress, machine learning and deep learning still suffer from incomprehensible predictions. Incomprehensibility, however, is not an option for the use of (deep) reinforcement learning in the real world, as unpredictable actions can seriously harm the involved individuals. In this work, we propose a genetic programming framework to generate explanations for the decision-making process of already trained agents by imitating them with programs. Programs are interpretable and can be executed to generate explanations of why the agent chooses a particular action. Furthermore, we conduct an ablation study that investigates how extending the domain-specific language by using library learning alters the performance of the method. We compare our results with the previous state of the art for this problem and show that we are comparable in performance but require much less hardware resources and computation time.
{"title":"Unveiling the Decision-Making Process in Reinforcement Learning with Genetic Programming","authors":"Manuel Eberhardinger, Florian Rupp, Johannes Maucher, Setareh Maghsudi","doi":"arxiv-2407.14714","DOIUrl":"https://doi.org/arxiv-2407.14714","url":null,"abstract":"Despite tremendous progress, machine learning and deep learning still suffer\u0000from incomprehensible predictions. Incomprehensibility, however, is not an\u0000option for the use of (deep) reinforcement learning in the real world, as\u0000unpredictable actions can seriously harm the involved individuals. In this\u0000work, we propose a genetic programming framework to generate explanations for\u0000the decision-making process of already trained agents by imitating them with\u0000programs. Programs are interpretable and can be executed to generate\u0000explanations of why the agent chooses a particular action. Furthermore, we\u0000conduct an ablation study that investigates how extending the domain-specific\u0000language by using library learning alters the performance of the method. We\u0000compare our results with the previous state of the art for this problem and\u0000show that we are comparable in performance but require much less hardware\u0000resources and computation time.","PeriodicalId":501347,"journal":{"name":"arXiv - CS - Neural and Evolutionary Computing","volume":"56 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141777742","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：