Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870305
Felipe Rooke da Silva, A. Vieira, H. Bernardino, Victor Aquiles Alencar, Lucas Ribeiro Pessamilio, H. Barbosa
Automated Machine Learn (AutoML) process is target of large studies, both from academia and industry. AutoML reduces the demand for data scientists and makes specialists in specific fields able to use Machine Learn (ML) in their domains. An application of ML algorithms is over time-series forecasting, and about these, few works involve the application of AutoML. In this work, an AutoML approach that aggregates time-series forecasting models is proposed. Furthermore, a special focus is given to the optimization stage, which uses genetic algorithm to boost searching for hyper-parameters. In the end, results are compared with a recent time-series forecasting benchmark and we verify that the AutoML model proposed in this work surpasses the benchmark.
{"title":"Automated Machine Learning for Time Series Prediction","authors":"Felipe Rooke da Silva, A. Vieira, H. Bernardino, Victor Aquiles Alencar, Lucas Ribeiro Pessamilio, H. Barbosa","doi":"10.1109/CEC55065.2022.9870305","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870305","url":null,"abstract":"Automated Machine Learn (AutoML) process is target of large studies, both from academia and industry. AutoML reduces the demand for data scientists and makes specialists in specific fields able to use Machine Learn (ML) in their domains. An application of ML algorithms is over time-series forecasting, and about these, few works involve the application of AutoML. In this work, an AutoML approach that aggregates time-series forecasting models is proposed. Furthermore, a special focus is given to the optimization stage, which uses genetic algorithm to boost searching for hyper-parameters. In the end, results are compared with a recent time-series forecasting benchmark and we verify that the AutoML model proposed in this work surpasses the benchmark.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114958670","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870437
João Batista Rodrigues Neto, G. Ramos
The continuous use of satellite networks in the Low Earth Orbit (LEO) has accumulated a large amount of space debris. Given the actual state of the orbit, these debris are a threat to the active systems and to the feasibility of future operations in LEO. Now, Active Debris Removal (ADR) missions must be conducted to mitigate the debris through forced deorbitation. The best documented approaches for the ADR mission planning made use of metaheuristics, modeling the ADR as a complex variant of the TSP. However, these approaches usually fail to deal some of the ADR problem dynamics, such as large instances, mission constraints or the debris motion. In this paper we propose heuristic of continuous improvement on a genetic-based solution. Our work advances the state of the art by dealing with large real world instances, modeling all the constraints and considering the problem time dependence (motion). Experiments were conducted to evidence the improvements over the literature. With the ability of generating time-dependent results for scenarios with thousands of debris in a feasible time, our approach yielded missions 96.33 % more effective at the cleaning job than the present ones on the literature.
{"title":"An Interpolated Approach for Active Debris Removal","authors":"João Batista Rodrigues Neto, G. Ramos","doi":"10.1109/CEC55065.2022.9870437","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870437","url":null,"abstract":"The continuous use of satellite networks in the Low Earth Orbit (LEO) has accumulated a large amount of space debris. Given the actual state of the orbit, these debris are a threat to the active systems and to the feasibility of future operations in LEO. Now, Active Debris Removal (ADR) missions must be conducted to mitigate the debris through forced deorbitation. The best documented approaches for the ADR mission planning made use of metaheuristics, modeling the ADR as a complex variant of the TSP. However, these approaches usually fail to deal some of the ADR problem dynamics, such as large instances, mission constraints or the debris motion. In this paper we propose heuristic of continuous improvement on a genetic-based solution. Our work advances the state of the art by dealing with large real world instances, modeling all the constraints and considering the problem time dependence (motion). Experiments were conducted to evidence the improvements over the literature. With the ability of generating time-dependent results for scenarios with thousands of debris in a feasible time, our approach yielded missions 96.33 % more effective at the cleaning job than the present ones on the literature.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131225658","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870270
Ozgur Salman, Michael Kampouridis, D. Jarchi
The subject of financial forecasting has been re-searched for decades, and the driver behind its measured data has been fuelled by the selection of physical time series, which summarize data using fixed time intervals. For instance, time-series for daily stock data would be profiled at 252 points in one year. However, this episodic style neglects the important events, or price changes that occur between two intervals. Thus, we use Directional Changes (DC) as an event-based series, which is an alternative way to record price movements. In DC, unlike time-series methods, time intervals are constituted by price changes. The unique feature that decides the price change to be considered as a significant is called a threshold θ. The objective of our paper is to create DC-based trading strategies, and then optimize them using a Genetic Algorithm (GA). To construct such strategies, we use DC-based indicators and scaling laws that have been empirically identified under DC summaries. We first propose four novel DC-based trading strategies and then combine them with existing DC-based strategies and finally optimize them via the GA. We conduct trading experiments over 44 stocks. Results show that the GA-optimized strategies are able to generate new and profitable trading strategies, significantly outperforming the individual DC-based strategies, as well as a buy and sell benchmark.
{"title":"Trading Strategies Optimization by Genetic Algorithm under the Directional Changes Paradigm","authors":"Ozgur Salman, Michael Kampouridis, D. Jarchi","doi":"10.1109/CEC55065.2022.9870270","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870270","url":null,"abstract":"The subject of financial forecasting has been re-searched for decades, and the driver behind its measured data has been fuelled by the selection of physical time series, which summarize data using fixed time intervals. For instance, time-series for daily stock data would be profiled at 252 points in one year. However, this episodic style neglects the important events, or price changes that occur between two intervals. Thus, we use Directional Changes (DC) as an event-based series, which is an alternative way to record price movements. In DC, unlike time-series methods, time intervals are constituted by price changes. The unique feature that decides the price change to be considered as a significant is called a threshold θ. The objective of our paper is to create DC-based trading strategies, and then optimize them using a Genetic Algorithm (GA). To construct such strategies, we use DC-based indicators and scaling laws that have been empirically identified under DC summaries. We first propose four novel DC-based trading strategies and then combine them with existing DC-based strategies and finally optimize them via the GA. We conduct trading experiments over 44 stocks. Results show that the GA-optimized strategies are able to generate new and profitable trading strategies, significantly outperforming the individual DC-based strategies, as well as a buy and sell benchmark.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134101737","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870232
Stephen Y. Chen, Antonio Bolufé-Röhler, James Montgomery, Wenxuan Zhang, T. Hendtlass
It is well known that metaheuristics for numerical optimization tend to decrease in performance as dimensionality increases. These effects are commonly referred to as “The Curse of Dimensionality”. An obvious change to search spaces with increasing dimensionality is that their volume grows exponentially, and this has led to large amounts of research on improved exploration. A recent insight is that the shape of attraction basins can also change drastically with increasing dimensionality, and this has led to selection-based approaches to combat the Curse of Dimensionality. Average-Fitness Based Selection is introduced as a means to reduce the selection errors caused by Fitness-Based Selection. Experimental results show that the rate of selection errors grows much more slowly for Average-Fitness Based Selection with Increasing dimensionality.
{"title":"Using Average-Fitness Based Selection to Combat the Curse of Dimensionality","authors":"Stephen Y. Chen, Antonio Bolufé-Röhler, James Montgomery, Wenxuan Zhang, T. Hendtlass","doi":"10.1109/CEC55065.2022.9870232","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870232","url":null,"abstract":"It is well known that metaheuristics for numerical optimization tend to decrease in performance as dimensionality increases. These effects are commonly referred to as “The Curse of Dimensionality”. An obvious change to search spaces with increasing dimensionality is that their volume grows exponentially, and this has led to large amounts of research on improved exploration. A recent insight is that the shape of attraction basins can also change drastically with increasing dimensionality, and this has led to selection-based approaches to combat the Curse of Dimensionality. Average-Fitness Based Selection is introduced as a means to reduce the selection errors caused by Fitness-Based Selection. Experimental results show that the rate of selection errors grows much more slowly for Average-Fitness Based Selection with Increasing dimensionality.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131261534","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870331
Naruhiko Nimura, A. Oyama
A new topology optimization method using genetic programming is proposed. To simultaneously achieve the gen-eration of shapes with high degrees of freedom and efficient optimization, the quadtree used in image processing is employed to reduce the number of design variables. Because the quadtree used in image processing implicitly holds coordinate information, we propose a new crossover and mutation method that inherits this information. For validation of the proposed approach, shape optimization and topology optimization are demonstrated where target airfoils including multi-element airfoils are reproduced. As a result, it is confirmed that the proposed method works for shape and topology optimizations with high efficiency.
{"title":"Evolutionary Topology Optimization Using Quadtree Genetic Programming","authors":"Naruhiko Nimura, A. Oyama","doi":"10.1109/CEC55065.2022.9870331","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870331","url":null,"abstract":"A new topology optimization method using genetic programming is proposed. To simultaneously achieve the gen-eration of shapes with high degrees of freedom and efficient optimization, the quadtree used in image processing is employed to reduce the number of design variables. Because the quadtree used in image processing implicitly holds coordinate information, we propose a new crossover and mutation method that inherits this information. For validation of the proposed approach, shape optimization and topology optimization are demonstrated where target airfoils including multi-element airfoils are reproduced. As a result, it is confirmed that the proposed method works for shape and topology optimizations with high efficiency.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133515325","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870394
Rafael R. M. Ribeiro, Carlos Dias Maciel
Recently there is great interest in optimization, especially on meta-heuristic algorithms. Many works have proposed improvements for these algorithms for general and specific applications. In this paper the Adaptive Genetic Algorithm with Varying Population Size (AGAVaPS) is proposed, an improvement of Genetic Algorithm. On the AGAVaPS each solution has their own mutation rate and number of iterations that the solution will be in the population. The proposed optimizer is tested against six other well established optimizers on the CEC2017 single objective optimization benchmark functions considering coverage of the search space and quality of solution obtained. It is also tested for feature selection and Bayesian network structural learning. The evolution of the population size over the iterations is also analysed. The results obtained show that the AGAVaPS has a very competitive performance in both, coverage and quality of solution.
{"title":"AGAVaPS - Adaptive Genetic Algorithm with Varying Population Size","authors":"Rafael R. M. Ribeiro, Carlos Dias Maciel","doi":"10.1109/CEC55065.2022.9870394","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870394","url":null,"abstract":"Recently there is great interest in optimization, especially on meta-heuristic algorithms. Many works have proposed improvements for these algorithms for general and specific applications. In this paper the Adaptive Genetic Algorithm with Varying Population Size (AGAVaPS) is proposed, an improvement of Genetic Algorithm. On the AGAVaPS each solution has their own mutation rate and number of iterations that the solution will be in the population. The proposed optimizer is tested against six other well established optimizers on the CEC2017 single objective optimization benchmark functions considering coverage of the search space and quality of solution obtained. It is also tested for feature selection and Bayesian network structural learning. The evolution of the population size over the iterations is also analysed. The results obtained show that the AGAVaPS has a very competitive performance in both, coverage and quality of solution.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132040221","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870234
V. Gori, Giacomo Veneri, Valeria Ballarini
We apply a Recurrent Neural Network (RNN), Kullback-Leibler (KL) divergence and a continual learning approach to check the status of several hundreds of sensors during turbo-machinery prototype testing. Turbo-machinery prototypes can be instrumented with up to thousands of sensors. Therefore, checking the health of each sensor is a time consuming activity. Prototypes are also tested on several different and a-priori unknown operating conditions, so we cannot apply a purely supervised model to detect potential anomalies of sensors and, moreover, we have to take into account a covariate shift because measurements drift continuously day by day. We continuously train a RNN (daily) to build a virtual sensor from other sensors and we compare the predicted signal vs the real signal to raise (in case) an anomaly. Furthermore, KL is used to estimate the overlap between the input distributions available at training time and the ones seen at test time, and thus the confidence level of the prediction. Finally we implement an end-to-end system to automatically train and evaluate the models. The paper presents the system and reports the application to a test campaign of about five hundred sensors.
{"title":"Continual Learning for anomaly detection on turbomachinery prototypes - A real application","authors":"V. Gori, Giacomo Veneri, Valeria Ballarini","doi":"10.1109/CEC55065.2022.9870234","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870234","url":null,"abstract":"We apply a Recurrent Neural Network (RNN), Kullback-Leibler (KL) divergence and a continual learning approach to check the status of several hundreds of sensors during turbo-machinery prototype testing. Turbo-machinery prototypes can be instrumented with up to thousands of sensors. Therefore, checking the health of each sensor is a time consuming activity. Prototypes are also tested on several different and a-priori unknown operating conditions, so we cannot apply a purely supervised model to detect potential anomalies of sensors and, moreover, we have to take into account a covariate shift because measurements drift continuously day by day. We continuously train a RNN (daily) to build a virtual sensor from other sensors and we compare the predicted signal vs the real signal to raise (in case) an anomaly. Furthermore, KL is used to estimate the overlap between the input distributions available at training time and the ones seen at test time, and thus the confidence level of the prediction. Finally we implement an end-to-end system to automatically train and evaluate the models. The paper presents the system and reports the application to a test campaign of about five hundred sensors.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132822662","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870288
Caie Hu, Sanyou Zeng, Changhe Li
Gaussian processes (GPs) are a kind of non-parametric Bayesian approach. They are widely used as surrogate models in data-driven optimization to approximate the exact functions. However, the cubic computation complexity is involved in building GPs. This paper proposes hyperparameters adaptive sharing based on transfer learning for scalable GPs to address the limitation. In this method, the hyperparameters across source tasks are adaptively shared to the target task by the linear predictor. This method can reduce the computation cost of building GPs without losing capability based on experimental analyses. The method's effectiveness is demonstrated on a set of benchmark problems.
{"title":"Hyperparameters Adaptive Sharing Based on Transfer Learning for Scalable GPs","authors":"Caie Hu, Sanyou Zeng, Changhe Li","doi":"10.1109/CEC55065.2022.9870288","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870288","url":null,"abstract":"Gaussian processes (GPs) are a kind of non-parametric Bayesian approach. They are widely used as surrogate models in data-driven optimization to approximate the exact functions. However, the cubic computation complexity is involved in building GPs. This paper proposes hyperparameters adaptive sharing based on transfer learning for scalable GPs to address the limitation. In this method, the hyperparameters across source tasks are adaptively shared to the target task by the linear predictor. This method can reduce the computation cost of building GPs without losing capability based on experimental analyses. The method's effectiveness is demonstrated on a set of benchmark problems.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133254352","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870271
Jordan Maslen, B. Ross
Mixed media in the real world involves the creation of works of art that creatively combine a variety of media on the canvas, for example, watercolour, acrylic paint, and photographs. We present an evolutionary art system that implements a digital version of mixed media. A genetic programming system uses a language that renders different digital effects on a canvas. Each rendered effect takes the form of an “art object”, and the tree defines a s et o fa rt o bjects that together comprise a final rendered image. Available effects include procedural images (textures), image filters, and bitmaps. A n art o bject is rendered onto the canvas via a pre-defined mask shape, which c an range from simple geometric shapes such as circles or squares, to com-plex paintbrush strokes and paint splatters. Fitness evaluation measures the pixel-by-pixel colour distance between a rendered canvas and an input target image, which acts as a compositional guide for rendered images. Various runs of the system have produced an interesting variety of stylized, mixed-effect results, often appearing as abstract “glitchy” interpretations of target images.
{"title":"Mixed Media in Evolutionary Art","authors":"Jordan Maslen, B. Ross","doi":"10.1109/CEC55065.2022.9870271","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870271","url":null,"abstract":"Mixed media in the real world involves the creation of works of art that creatively combine a variety of media on the canvas, for example, watercolour, acrylic paint, and photographs. We present an evolutionary art system that implements a digital version of mixed media. A genetic programming system uses a language that renders different digital effects on a canvas. Each rendered effect takes the form of an “art object”, and the tree defines a s et o fa rt o bjects that together comprise a final rendered image. Available effects include procedural images (textures), image filters, and bitmaps. A n art o bject is rendered onto the canvas via a pre-defined mask shape, which c an range from simple geometric shapes such as circles or squares, to com-plex paintbrush strokes and paint splatters. Fitness evaluation measures the pixel-by-pixel colour distance between a rendered canvas and an input target image, which acts as a compositional guide for rendered images. Various runs of the system have produced an interesting variety of stylized, mixed-effect results, often appearing as abstract “glitchy” interpretations of target images.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116294373","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}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870434
André Ramos Fernandes Da Silva, L. M. Pavelski, Luiz Alberto Queiroz Cordovil Júnior, Paulo Henrique De Oliveira Gomes, Layane Menezes Azevedo, Francisco Erivaldo Fernandes Junior
Neural Architecture Search (NAS) is a busy research field growing exponentially in recent years. State-of-the-art deep neural networks usually require a specialist to fine-tune the model to solve a specific problem. NAS research aims to design neural network architectures automatically, thus easing the need for machine learning specialists to spend a lot of effort on hand-crafted attempts. As artificial intelligence applications are becoming ubiquitous, there is also a growing interest in efficient applications that could be deployed to smartphones, smart wearable devices, and other edge devices. Gender recognition in unfiltered images — such as those we find in real-world situations like pictures taken with smartphones and video shots from surveillance cameras — is one of such challenging applications. In this work, we developed an evolutionary NAS algorithm that consistently finds efficient ResNet-based architectures, named RENNAS, which have a good trade-off between classification accuracy and architectural and computational complexities. We demonstrate our algorithm's performance on Adience dataset of unfiltered images for gender recognition.
{"title":"An evolutionary search algorithm for efficient ResNet-based architectures: a case study on gender recognition","authors":"André Ramos Fernandes Da Silva, L. M. Pavelski, Luiz Alberto Queiroz Cordovil Júnior, Paulo Henrique De Oliveira Gomes, Layane Menezes Azevedo, Francisco Erivaldo Fernandes Junior","doi":"10.1109/CEC55065.2022.9870434","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870434","url":null,"abstract":"Neural Architecture Search (NAS) is a busy research field growing exponentially in recent years. State-of-the-art deep neural networks usually require a specialist to fine-tune the model to solve a specific problem. NAS research aims to design neural network architectures automatically, thus easing the need for machine learning specialists to spend a lot of effort on hand-crafted attempts. As artificial intelligence applications are becoming ubiquitous, there is also a growing interest in efficient applications that could be deployed to smartphones, smart wearable devices, and other edge devices. Gender recognition in unfiltered images — such as those we find in real-world situations like pictures taken with smartphones and video shots from surveillance cameras — is one of such challenging applications. In this work, we developed an evolutionary NAS algorithm that consistently finds efficient ResNet-based architectures, named RENNAS, which have a good trade-off between classification accuracy and architectural and computational complexities. We demonstrate our algorithm's performance on Adience dataset of unfiltered images for gender recognition.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"22 2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133839438","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：