Pub Date : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660080
Simen Sverdrup-Thygeson, P. Haddow
An explosive growth of misleading and untrustworthy news articles has been observed over the last years. These news articles are often referred to as fake news and have been found to severely impact fair elections and democratic values. Computational Intelligence models may be applied to the classification of news articles, assuming that an efficient feature set is available as input to the model. However, the selection of appropriate feature sets is an open question for such high-dimensional tasks. A further challenge is the general applicability of feature selection strategies, where testing on a single dataset may convey misleading results. The work herein evaluates a wide-range of potential news article features resulting in twenty-five potential features. Feature selection, based on a combination of feature scoring, feature ranking and mutual information is then applied, evaluated on multiple datasets: Kaggle, Liar and FakeNewsNet. An Artificial Immune System model is applied in the feature ranking and as the classification model. The accuracy obtained is compared to state of the art fake news classification models, highlighting that the approach shows promise in terms of accuracy despite the small feature sets provided for classification.
{"title":"Feature Selection for Fake News Classification","authors":"Simen Sverdrup-Thygeson, P. Haddow","doi":"10.1109/SSCI50451.2021.9660080","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660080","url":null,"abstract":"An explosive growth of misleading and untrustworthy news articles has been observed over the last years. These news articles are often referred to as fake news and have been found to severely impact fair elections and democratic values. Computational Intelligence models may be applied to the classification of news articles, assuming that an efficient feature set is available as input to the model. However, the selection of appropriate feature sets is an open question for such high-dimensional tasks. A further challenge is the general applicability of feature selection strategies, where testing on a single dataset may convey misleading results. The work herein evaluates a wide-range of potential news article features resulting in twenty-five potential features. Feature selection, based on a combination of feature scoring, feature ranking and mutual information is then applied, evaluated on multiple datasets: Kaggle, Liar and FakeNewsNet. An Artificial Immune System model is applied in the feature ranking and as the classification model. The accuracy obtained is compared to state of the art fake news classification models, highlighting that the approach shows promise in terms of accuracy despite the small feature sets provided for classification.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129341214","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660028
Cláudio A. Monteiro, F. M. P. Neto
The use of properties that are intrinsic to quantum mechanics has made it possible to build quantum algorithms with greater efficiency than classical algorithms to solve problems whose classically efficient solution either does not exist or is not known. There are quantum neurons that can carry an exponential amount of information to a linear number of quantum information units (qubits) using the quantum property of superposition. In this paper, we compare the performance of three of these quantum neuron models applied to the diabetes classification problem. We also propose the use of different data preprocessing strategies. Quantum neurons were simulated using the IBM Qiskit tool. We compare the preprocessing approaches applied to two toy problems (1) simulating the XOR operator and (2) solving a generic nonlinear problem. The results of the experiments shows that a single quantum neuron is capable of achieving an accuracy rate of 100% in the XOR problem and an accuracy rate of 100% in a non-linear dataset, demonstrating that the quantum neurons with real weights are capable of modeling non-linearly separable problems. In the problem of diagnosing diabetes, quantum neurons achieved an accuracy rate of 76% and AUC-ROC of 88%, while its classic version, the perceptron, reached only 63% accuracy and the artificial neural network reached 80% AUC-ROC. These results indicate that a single quantum neuron performs better than its classical version and even the artificial neural network for AUC-ROC, demonstrating potential for use in healthcare applications in the near future.
{"title":"Diabetes Prediction Using Quantum Neurons with Preprocessing Based on Hypercomplex Numbers","authors":"Cláudio A. Monteiro, F. M. P. Neto","doi":"10.1109/SSCI50451.2021.9660028","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660028","url":null,"abstract":"The use of properties that are intrinsic to quantum mechanics has made it possible to build quantum algorithms with greater efficiency than classical algorithms to solve problems whose classically efficient solution either does not exist or is not known. There are quantum neurons that can carry an exponential amount of information to a linear number of quantum information units (qubits) using the quantum property of superposition. In this paper, we compare the performance of three of these quantum neuron models applied to the diabetes classification problem. We also propose the use of different data preprocessing strategies. Quantum neurons were simulated using the IBM Qiskit tool. We compare the preprocessing approaches applied to two toy problems (1) simulating the XOR operator and (2) solving a generic nonlinear problem. The results of the experiments shows that a single quantum neuron is capable of achieving an accuracy rate of 100% in the XOR problem and an accuracy rate of 100% in a non-linear dataset, demonstrating that the quantum neurons with real weights are capable of modeling non-linearly separable problems. In the problem of diagnosing diabetes, quantum neurons achieved an accuracy rate of 76% and AUC-ROC of 88%, while its classic version, the perceptron, reached only 63% accuracy and the artificial neural network reached 80% AUC-ROC. These results indicate that a single quantum neuron performs better than its classical version and even the artificial neural network for AUC-ROC, demonstrating potential for use in healthcare applications in the near future.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129475554","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660117
F. Lezama, J. Soares, B. Canizes, Z. Vale
Evolutionary algorithms (EAs) have emerged as an efficient alternative to deal with real-world applications with high complexity. However, due to the stochastic nature of the results obtained using EAs, the design of benchmarks and competitions where such approaches can be evaluated and compared is attracting attention in the field. In the energy domain, the “2021 CEC-GECCO-PESGM Competition on Evolutionary Computation in the Energy Domain: Smart Grid Applications” provides a platform to test and compare new EAs to solve complex problems in the field. However, the metric used to rank the algorithms is based solely on the mean fitness value (related to the objective function value only), which does not give statistical significance to the performance of the algorithms. Thus, this paper presents a statistical analysis using the Wilcoxon pair-wise comparison to study the performance of algorithms with statistical grounds. Results suggest that, for track 1 of the competition, only the winner approach (first place) is significantly different and superior to the other algorithms; in contrast, the second place is already statistically comparable to some other contestants. For track 2, all the winner approaches (first, second, and third) are statistically different from each other and the rest of the contestants. This type of analysis is important to have a deeper understanding of the stochastic performance of algorithms.
{"title":"A Statistical Analysis of Performance in the 2021 CEC-GECCO-PESGM Competition on Evolutionary Computation in the Energy Domain","authors":"F. Lezama, J. Soares, B. Canizes, Z. Vale","doi":"10.1109/SSCI50451.2021.9660117","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660117","url":null,"abstract":"Evolutionary algorithms (EAs) have emerged as an efficient alternative to deal with real-world applications with high complexity. However, due to the stochastic nature of the results obtained using EAs, the design of benchmarks and competitions where such approaches can be evaluated and compared is attracting attention in the field. In the energy domain, the “2021 CEC-GECCO-PESGM Competition on Evolutionary Computation in the Energy Domain: Smart Grid Applications” provides a platform to test and compare new EAs to solve complex problems in the field. However, the metric used to rank the algorithms is based solely on the mean fitness value (related to the objective function value only), which does not give statistical significance to the performance of the algorithms. Thus, this paper presents a statistical analysis using the Wilcoxon pair-wise comparison to study the performance of algorithms with statistical grounds. Results suggest that, for track 1 of the competition, only the winner approach (first place) is significantly different and superior to the other algorithms; in contrast, the second place is already statistically comparable to some other contestants. For track 2, all the winner approaches (first, second, and third) are statistically different from each other and the rest of the contestants. This type of analysis is important to have a deeper understanding of the stochastic performance of algorithms.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124856917","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659540
Zhaoyang Liu, Tianchun Xiang, Tianhao Wang, C. Mu
In this paper, a consensus multi-agent deep reinforcement learning algorithm is introduced for distributed cooperative secondary voltage control of microgrids. To reduce dependence on the system model and enhance communication efficiency, we propose a fully decentralized multi-agent advantage actor critic (A2C) algorithm with local communication networks, which considers each distributed energy resource (DER) as an agent. Both local state and the messages received from neighbors are employed by each agent to learn a control strategy. Moreover, the maximum entropy reinforcement learning framework is applied to improve exploration of agents. The proposed algorithm is verified in two different scale microgrid setups, which are microgrid-6 and microgrid-20. Experiment results show the effectiveness and superiority of our proposed algorithm.
{"title":"Cooperative Optimization Strategy for Distributed Energy Resource System using Multi-Agent Reinforcement Learning","authors":"Zhaoyang Liu, Tianchun Xiang, Tianhao Wang, C. Mu","doi":"10.1109/SSCI50451.2021.9659540","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659540","url":null,"abstract":"In this paper, a consensus multi-agent deep reinforcement learning algorithm is introduced for distributed cooperative secondary voltage control of microgrids. To reduce dependence on the system model and enhance communication efficiency, we propose a fully decentralized multi-agent advantage actor critic (A2C) algorithm with local communication networks, which considers each distributed energy resource (DER) as an agent. Both local state and the messages received from neighbors are employed by each agent to learn a control strategy. Moreover, the maximum entropy reinforcement learning framework is applied to improve exploration of agents. The proposed algorithm is verified in two different scale microgrid setups, which are microgrid-6 and microgrid-20. Experiment results show the effectiveness and superiority of our proposed algorithm.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125267520","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659934
Thilina Perera, L. Wijerathna, Deshya Wijesundera, T. Srikanthan
On-demand public transit systems require real-time computation of routes to ensure a user-friendly responsive service while also minimizing the vehicle miles traveled (VMT) of the fleet for increasing the profits of an operator. To ensure responsiveness, heuristic algorithms that rapidly generate near-optimal solutions are preferred over time-consuming exact computations. In order to further ensure the scalability of heuristic algorithms, especially to solve large problems, parallel computing techniques need to distribute the workload evenly across several partitions, while keeping passengers on similar routes with less detour in a single partition to reduce the VMT. However, existing works ignore these factors when partitioning the workload. This work proposes a road-network aware tree partitioning algorithm that not only considers the shortest path based routes but also the workloads to create balanced partitions in real-time. Experimental results on a real road-network show that the proposed algorithm outperforms a well-known unsupervised learning algorithm in terms of quality of results and runtime.
{"title":"Road-network aware Dynamic Workload Balancing Technique for Real-time Route Generation in On-Demand Public Transit","authors":"Thilina Perera, L. Wijerathna, Deshya Wijesundera, T. Srikanthan","doi":"10.1109/SSCI50451.2021.9659934","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659934","url":null,"abstract":"On-demand public transit systems require real-time computation of routes to ensure a user-friendly responsive service while also minimizing the vehicle miles traveled (VMT) of the fleet for increasing the profits of an operator. To ensure responsiveness, heuristic algorithms that rapidly generate near-optimal solutions are preferred over time-consuming exact computations. In order to further ensure the scalability of heuristic algorithms, especially to solve large problems, parallel computing techniques need to distribute the workload evenly across several partitions, while keeping passengers on similar routes with less detour in a single partition to reduce the VMT. However, existing works ignore these factors when partitioning the workload. This work proposes a road-network aware tree partitioning algorithm that not only considers the shortest path based routes but also the workloads to create balanced partitions in real-time. Experimental results on a real road-network show that the proposed algorithm outperforms a well-known unsupervised learning algorithm in terms of quality of results and runtime.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126227968","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660126
Gang Li, Ogechi Onuoha, Mark Mcgill, S. Brewster, C. Chen, F. Pollick
How the performance of autonomic physiological, and human vestibular network (HVN)-based brain functional connectivity (BFC) features differ in a virtual reality (VR) sickness classification task is underexplored. Therefore, this paper presents an artificial intelligence (AI)-aided comparative study of the two. Results from different AI models all show that autonomic physiological features represented by the combined heart rate, fingertip temperature and forehead temperature are superior to HVN-based BFC features represented by the phase-locking values of inter-electrode coherence (IEC) of electroencephalogram (EEG) in the same VR sickness condition (that is, as a result of experiencing tunnel travel-induced illusory self-motion (vection) about moving in-depth in this study). Regarding EEG features per se (IEC-BFC vs traditional power spectrum), we did not find much difference across AI models.
{"title":"Comparing Autonomic Physiological and Electroencephalography Features for VR Sickness Detection Using Predictive Models","authors":"Gang Li, Ogechi Onuoha, Mark Mcgill, S. Brewster, C. Chen, F. Pollick","doi":"10.1109/SSCI50451.2021.9660126","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660126","url":null,"abstract":"How the performance of autonomic physiological, and human vestibular network (HVN)-based brain functional connectivity (BFC) features differ in a virtual reality (VR) sickness classification task is underexplored. Therefore, this paper presents an artificial intelligence (AI)-aided comparative study of the two. Results from different AI models all show that autonomic physiological features represented by the combined heart rate, fingertip temperature and forehead temperature are superior to HVN-based BFC features represented by the phase-locking values of inter-electrode coherence (IEC) of electroencephalogram (EEG) in the same VR sickness condition (that is, as a result of experiencing tunnel travel-induced illusory self-motion (vection) about moving in-depth in this study). Regarding EEG features per se (IEC-BFC vs traditional power spectrum), we did not find much difference across AI models.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126766508","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 research of vehicle routing problem (VRP) is significant for people traveling and logistics distribution. Recently, in order to alleviate global warming, the VRP based on electric vehicles has attracted much attention from researchers. In this paper, a bi-level routing problem model based on electric vehicles is presented, which can simulate the actual logistics distribution process. The classic backpropagation neural network is used to predict the road conditions for applying the method in real life. We also propose a local search algorithm based on a dynamic constrained multiobjective optimization framework. In this algorithm, 26 local search operators are designed and selected adaptively to optimize initial solutions. We also make a comparison between our algorithm and 3 modified algorithms. Experimental results indicate that our algorithm can attain an excellent solution that can satisfy the constraints of the VRP with real-time traffic conditions and be more competitive than the other 3 modified algorithms.
{"title":"An Adaptive Evolutionary Algorithm for Bi- Level Multi-objective VRPs with Real-Time Traffic Conditions","authors":"Baojian Chen, Changhe Li, Sanyou Zeng, Shengxiang Yang, Michalis Mavrovouniotis","doi":"10.1109/SSCI50451.2021.9659933","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659933","url":null,"abstract":"The research of vehicle routing problem (VRP) is significant for people traveling and logistics distribution. Recently, in order to alleviate global warming, the VRP based on electric vehicles has attracted much attention from researchers. In this paper, a bi-level routing problem model based on electric vehicles is presented, which can simulate the actual logistics distribution process. The classic backpropagation neural network is used to predict the road conditions for applying the method in real life. We also propose a local search algorithm based on a dynamic constrained multiobjective optimization framework. In this algorithm, 26 local search operators are designed and selected adaptively to optimize initial solutions. We also make a comparison between our algorithm and 3 modified algorithms. Experimental results indicate that our algorithm can attain an excellent solution that can satisfy the constraints of the VRP with real-time traffic conditions and be more competitive than the other 3 modified algorithms.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"110 8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125882804","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659544
O. Odeyomi, G. Záruba
This paper discusses how clients in a federated learning system can collaborate with privacy guarantee in a fully decentralized setting without a central server. Most existing work includes a central server that aggregates the local updates from the clients and coordinates the training. Thus, the setting in this existing work is prone to communication and computational bottlenecks, especially when large number of clients are involved. Also, most existing federated learning algorithms do not cater for situations where the data distribution is time-varying such as in real-time traffic monitoring. To address these problems, this paper proposes a differentially-private online mirror descent algorithm. To provide additional privacy to the loss gradients of the clients, local differential privacy is introduced. Simulation results are based on a proposed differentially-private exponential gradient algorithm, which is a variant of differentially-private online mirror descent algorithm with entropic regularizer. The simulation shows that all the clients can converge to the global optimal vector over time. The regret bound of the proposed differentially-private exponential gradient algorithm is compared with the regret bounds of some state-of-the-art online federated learning algorithms found in the literature.
{"title":"Privacy-Preserving Online Mirror Descent for Federated Learning with Single-Sided Trust","authors":"O. Odeyomi, G. Záruba","doi":"10.1109/SSCI50451.2021.9659544","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659544","url":null,"abstract":"This paper discusses how clients in a federated learning system can collaborate with privacy guarantee in a fully decentralized setting without a central server. Most existing work includes a central server that aggregates the local updates from the clients and coordinates the training. Thus, the setting in this existing work is prone to communication and computational bottlenecks, especially when large number of clients are involved. Also, most existing federated learning algorithms do not cater for situations where the data distribution is time-varying such as in real-time traffic monitoring. To address these problems, this paper proposes a differentially-private online mirror descent algorithm. To provide additional privacy to the loss gradients of the clients, local differential privacy is introduced. Simulation results are based on a proposed differentially-private exponential gradient algorithm, which is a variant of differentially-private online mirror descent algorithm with entropic regularizer. The simulation shows that all the clients can converge to the global optimal vector over time. The regret bound of the proposed differentially-private exponential gradient algorithm is compared with the regret bounds of some state-of-the-art online federated learning algorithms found in the literature.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"7 7","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120891697","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660078
Jaeyoon Lee, Hyuntak Lim, Ki-Seok Chung
Deep neural networks reveal their usefulness through learning from large amounts of data. However, unless the data is correctly labeled, it may be very difficult to properly train a neural network. Labeling the large set of data is a time-consuming and labor-intensive task. To overcome the risk of mislabeling, several methods that are robust against the label noise have been proposed. In this paper, we propose an effective label correction method called Curriculum Label Correction (CLC). With reference to the loss distribution from self-supervised learning, CLC identifies and corrects noisy labels utilizing curriculum learning. Our experimental results verify that CLC shows outstanding performance especially in a harshly noisy condition, 91.06% test accuracy on CIFAR-10 at a noise rate of 0.8. Code is available at https://github.com/LJY-HY/CLC.
{"title":"CLC: Noisy Label Correction via Curriculum Learning","authors":"Jaeyoon Lee, Hyuntak Lim, Ki-Seok Chung","doi":"10.1109/SSCI50451.2021.9660078","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660078","url":null,"abstract":"Deep neural networks reveal their usefulness through learning from large amounts of data. However, unless the data is correctly labeled, it may be very difficult to properly train a neural network. Labeling the large set of data is a time-consuming and labor-intensive task. To overcome the risk of mislabeling, several methods that are robust against the label noise have been proposed. In this paper, we propose an effective label correction method called Curriculum Label Correction (CLC). With reference to the loss distribution from self-supervised learning, CLC identifies and corrects noisy labels utilizing curriculum learning. Our experimental results verify that CLC shows outstanding performance especially in a harshly noisy condition, 91.06% test accuracy on CIFAR-10 at a noise rate of 0.8. Code is available at https://github.com/LJY-HY/CLC.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"39 7","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121011295","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660029
Koen Van Der Pool, A. Eiben
Robots are arguably essential for space research in the future, but designing and producing robots for unknown environments represents a grand challenge. The field of Evolutionary Robotics offers a solution by applying the principles of natural evolution to robot design. In this paper, we consider a Moon-like environment and investigate the joint evolution of morphologies (bodies) and controllers (brains) when fitness is determined by the ability to locomote. In particular, we are interested in the evolved morphologies and compare the emerging 'life forms' in a Moonlike environment to those evolved under Earth-like conditions. To model the Moon we change two environmental properties of our baseline environment that represents the Earth: gravity is set to a low value and the flat terrain is replaced by the NASA model of the Moon landing site of the Apollo 14. The results show that changing only one of these does not lead to different evolved robot morphologies, but changing both does. Our evolved Moonwalkers are usually bigger, have fewer limbs and a less space filling shape than the robots evolved on Earth.
{"title":"Moonwalkers: Evolving Robots for Locomotion in a Moon-like Environment","authors":"Koen Van Der Pool, A. Eiben","doi":"10.1109/SSCI50451.2021.9660029","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660029","url":null,"abstract":"Robots are arguably essential for space research in the future, but designing and producing robots for unknown environments represents a grand challenge. The field of Evolutionary Robotics offers a solution by applying the principles of natural evolution to robot design. In this paper, we consider a Moon-like environment and investigate the joint evolution of morphologies (bodies) and controllers (brains) when fitness is determined by the ability to locomote. In particular, we are interested in the evolved morphologies and compare the emerging 'life forms' in a Moonlike environment to those evolved under Earth-like conditions. To model the Moon we change two environmental properties of our baseline environment that represents the Earth: gravity is set to a low value and the flat terrain is replaced by the NASA model of the Moon landing site of the Apollo 14. The results show that changing only one of these does not lead to different evolved robot morphologies, but changing both does. Our evolved Moonwalkers are usually bigger, have fewer limbs and a less space filling shape than the robots evolved on Earth.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"17 12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125765444","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：