Pub Date : 2014-10-20DOI: 10.1109/UKCI.2014.6930152
Maktuba Mohid, J. Miller, Simon Harding, G. Tufte, O. R. Lykkebø, M. K. Massey, M. Petty
Evolution-in-materio (EIM) is a method that uses artificial evolution to exploit properties of materials to solve computational problems without requiring a detailed understanding of such properties. In this paper, we show that using a purpose-built hardware platform called Mecobo, it is possible to evolve voltages and signals applied to physical materials to solve computational problems. We demonstrate for the first time that this methodology can be applied to function optimization. We evaluate the approach on 23 function optimization benchmarks and in some cases results come very close to the global optimum or even surpass those provided by a well-known software-based evolutionary approach. This indicates that EIM has promise and further investigations would be fruitful.
{"title":"Evolution-in-materio: Solving function optimization problems using materials","authors":"Maktuba Mohid, J. Miller, Simon Harding, G. Tufte, O. R. Lykkebø, M. K. Massey, M. Petty","doi":"10.1109/UKCI.2014.6930152","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930152","url":null,"abstract":"Evolution-in-materio (EIM) is a method that uses artificial evolution to exploit properties of materials to solve computational problems without requiring a detailed understanding of such properties. In this paper, we show that using a purpose-built hardware platform called Mecobo, it is possible to evolve voltages and signals applied to physical materials to solve computational problems. We demonstrate for the first time that this methodology can be applied to function optimization. We evaluate the approach on 23 function optimization benchmarks and in some cases results come very close to the global optimum or even surpass those provided by a well-known software-based evolutionary approach. This indicates that EIM has promise and further investigations would be fruitful.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128457537","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930153
Longzhi Yang, D. Neagu
The recent development of information techniques, especially the state-of-the-art “big data” solutions, enables the extracting, gathering, and processing large amount of toxicity information from multiple sources. Facilitated by this technology advance, a framework named integrated testing strategies (ITS) has been proposed in the predictive toxicology domain, in an effort to intelligently jointly use multiple heterogeneous toxicity data records (through data fusion, grouping, interpolation/extrapolation etc.) for toxicity assessment. This will ultimately contribute to accelerating the development cycle of chemical products, reducing animal use, and decreasing development costs. Most of the current study in ITS is based on a group of consensus processes, termed weight of evidence (WoE), which quantitatively integrate all the relevant data instances towards the same endpoint into an integrated decision supported by data quality. Several WoE implementations for the particular case of toxicity data fusion have been presented in the literature, which are collectively studied in this paper. Noting that these uncertainty handling methodologies are usually not simply developed from conventional probability theory due to the unavailability of big datasets, this paper first investigates the mathematical foundations of these approaches. Then, the investigated data integration models are applied to a representative case in the predictive toxicology domain, with the experimental results compared and analysed.
{"title":"Integration strategies for toxicity data from an empirical perspective","authors":"Longzhi Yang, D. Neagu","doi":"10.1109/UKCI.2014.6930153","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930153","url":null,"abstract":"The recent development of information techniques, especially the state-of-the-art “big data” solutions, enables the extracting, gathering, and processing large amount of toxicity information from multiple sources. Facilitated by this technology advance, a framework named integrated testing strategies (ITS) has been proposed in the predictive toxicology domain, in an effort to intelligently jointly use multiple heterogeneous toxicity data records (through data fusion, grouping, interpolation/extrapolation etc.) for toxicity assessment. This will ultimately contribute to accelerating the development cycle of chemical products, reducing animal use, and decreasing development costs. Most of the current study in ITS is based on a group of consensus processes, termed weight of evidence (WoE), which quantitatively integrate all the relevant data instances towards the same endpoint into an integrated decision supported by data quality. Several WoE implementations for the particular case of toxicity data fusion have been presented in the literature, which are collectively studied in this paper. Noting that these uncertainty handling methodologies are usually not simply developed from conventional probability theory due to the unavailability of big datasets, this paper first investigates the mathematical foundations of these approaches. Then, the investigated data integration models are applied to a representative case in the predictive toxicology domain, with the experimental results compared and analysed.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124737528","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930155
Ali M. Abdulshahed, A. Longstaff, S. Fletcher
The fast and accurate modelling of thermal errors in machining is an important aspect for the implementation of thermal error compensation. This paper presents a novel modelling approach for thermal error compensation on CNC machine tools. The method combines the Adaptive Neuro Fuzzy Inference System (ANFIS) and Grey system theory to predict thermal errors in machining. Instead of following a traditional approach, which utilises original data patterns to construct the ANFIS model, this paper proposes to exploit Accumulation Generation Operation (AGO) to simplify the modelling procedures. AGO, a basis of the Grey system theory, is used to uncover a development tendency so that the features and laws of integration hidden in the chaotic raw data can be sufficiently revealed. AGO properties make it easier for the proposed model to design and predict. According to the simulation results, the proposed model demonstrates stronger prediction power than standard ANFIS model only with minimum number of training samples.
{"title":"A novel approach for ANFIS modelling based on Grey system theory for thermal error compensation","authors":"Ali M. Abdulshahed, A. Longstaff, S. Fletcher","doi":"10.1109/UKCI.2014.6930155","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930155","url":null,"abstract":"The fast and accurate modelling of thermal errors in machining is an important aspect for the implementation of thermal error compensation. This paper presents a novel modelling approach for thermal error compensation on CNC machine tools. The method combines the Adaptive Neuro Fuzzy Inference System (ANFIS) and Grey system theory to predict thermal errors in machining. Instead of following a traditional approach, which utilises original data patterns to construct the ANFIS model, this paper proposes to exploit Accumulation Generation Operation (AGO) to simplify the modelling procedures. AGO, a basis of the Grey system theory, is used to uncover a development tendency so that the features and laws of integration hidden in the chaotic raw data can be sufficiently revealed. AGO properties make it easier for the proposed model to design and predict. According to the simulation results, the proposed model demonstrates stronger prediction power than standard ANFIS model only with minimum number of training samples.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129551795","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930193
Hilal Abbood, W. Al-Nuaimy, Ali Al-Ataby, Sameh A. Salem, Hamzah S. AlZu'bi
Fatigue is a mental process that grows gradually and affects human reaction time and the consciousness. It is one of the causes of road fatal accidents around the globe. Although it is now generally accepted that fatigue plays an important role in road safety, it is still largely left to individual drivers to manage. The recent research in this area focuses on fatigue detection and the existing systems alert the drivers in severe fatigued stage. These systems use either physiological signs of the fatigue or the behavioural reaction to generate alerts. This research investigates the feasibility of using a group of fatigue symptoms (such as pupil response, gaze patterns, steering reaction and EEG) to build a robust fatigue detection algorithm that can be used in a real-life system for the early prediction and avoidance of fatigue development. Intensive testing and validation stages are required to ensure the reliability and the suitability of the system that should be able to detect fatigue levels at different degrees of tiredness. Moreover, the proposed system predicts subsequent stages of fatigue and generates an approximate behavioural model for each individual driver to enable more personalised and effective intervention.
{"title":"Prediction of driver fatigue: Approaches and open challenges","authors":"Hilal Abbood, W. Al-Nuaimy, Ali Al-Ataby, Sameh A. Salem, Hamzah S. AlZu'bi","doi":"10.1109/UKCI.2014.6930193","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930193","url":null,"abstract":"Fatigue is a mental process that grows gradually and affects human reaction time and the consciousness. It is one of the causes of road fatal accidents around the globe. Although it is now generally accepted that fatigue plays an important role in road safety, it is still largely left to individual drivers to manage. The recent research in this area focuses on fatigue detection and the existing systems alert the drivers in severe fatigued stage. These systems use either physiological signs of the fatigue or the behavioural reaction to generate alerts. This research investigates the feasibility of using a group of fatigue symptoms (such as pupil response, gaze patterns, steering reaction and EEG) to build a robust fatigue detection algorithm that can be used in a real-life system for the early prediction and avoidance of fatigue development. Intensive testing and validation stages are required to ensure the reliability and the suitability of the system that should be able to detect fatigue levels at different degrees of tiredness. Moreover, the proposed system predicts subsequent stages of fatigue and generates an approximate behavioural model for each individual driver to enable more personalised and effective intervention.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132452583","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930171
Shouyong Jiang, Shengxiang Yang
Many real-world optimization problems appear to not only have multiple objectives that conflict each other but also change over time. They are dynamic multi-objective optimization problems (DMOPs) and the corresponding field is called dynamic multi-objective optimization (DMO), which has gained growing attention in recent years. However, one main issue in the field of DMO is that there is no standard test suite to determine whether an algorithm is capable of solving them. This paper presents a new benchmark generator for DMOPs that can generate several complicated characteristics, including mixed Pareto-optimal front (convexity-concavity), strong dependencies between variables, and a mixed type of change, which are rarely tested in the literature. Experiments are conducted to compare the performance of five state-of-the-art DMO algorithms on several typical test functions derived from the proposed generator, which gives a better understanding of the strengths and weaknesses of these tested algorithms for DMOPs.
{"title":"A benchmark generator for dynamic multi-objective optimization problems","authors":"Shouyong Jiang, Shengxiang Yang","doi":"10.1109/UKCI.2014.6930171","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930171","url":null,"abstract":"Many real-world optimization problems appear to not only have multiple objectives that conflict each other but also change over time. They are dynamic multi-objective optimization problems (DMOPs) and the corresponding field is called dynamic multi-objective optimization (DMO), which has gained growing attention in recent years. However, one main issue in the field of DMO is that there is no standard test suite to determine whether an algorithm is capable of solving them. This paper presents a new benchmark generator for DMOPs that can generate several complicated characteristics, including mixed Pareto-optimal front (convexity-concavity), strong dependencies between variables, and a mixed type of change, which are rarely tested in the literature. Experiments are conducted to compare the performance of five state-of-the-art DMO algorithms on several typical test functions derived from the proposed generator, which gives a better understanding of the strengths and weaknesses of these tested algorithms for DMOPs.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130929197","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930156
Gang Yao, F. Chao, Hualin Zeng, Minghui Shi, Min Jiang, Changle Zhou
Classifier ensembles improve the performance of single classifier system. However, a classifier ensemble with too many classifiers may occupy a large number of computational time. This paper proposes a new ensemble subset evaluation method that integrates classifier diversity measures into a classifier ensemble reduction framework. The approach is implemented by using three conventional diversity algorithms and one new developed diversity measure method to calculate the diversity's merits within the classifier ensemble reduction framework. The subset evaluation method is demonstrated by the experimental data: the method not only can meet the requirements of high accuracy rate and fewer size, but also its running time is greatly shortened. When the accuracy requirements are not very strict, but the the running time requirements is more stringent, the proposed method is a good choice.
{"title":"Integrate classifier diversity evaluation to feature selection based classifier ensemble reduction","authors":"Gang Yao, F. Chao, Hualin Zeng, Minghui Shi, Min Jiang, Changle Zhou","doi":"10.1109/UKCI.2014.6930156","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930156","url":null,"abstract":"Classifier ensembles improve the performance of single classifier system. However, a classifier ensemble with too many classifiers may occupy a large number of computational time. This paper proposes a new ensemble subset evaluation method that integrates classifier diversity measures into a classifier ensemble reduction framework. The approach is implemented by using three conventional diversity algorithms and one new developed diversity measure method to calculate the diversity's merits within the classifier ensemble reduction framework. The subset evaluation method is demonstrated by the experimental data: the method not only can meet the requirements of high accuracy rate and fewer size, but also its running time is greatly shortened. When the accuracy requirements are not very strict, but the the running time requirements is more stringent, the proposed method is a good choice.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130115699","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930172
M. R. Kianifar, F. Campean, A. Wood
This paper presents the development and implementation of a customised Permutation Genetic Algorithm (PermGA) for a sequential Design of Experiment (DoE) framework based on space filling Optimal Latin Hypercube (OLH) designs. The work is motivated by multivariate engineering problems such as engine mapping experiments, which require efficient DoE strategies to minimise expensive testing. The DoE strategy is based on a flexible Model Building - Model Validation (MB-MV) sequence based on space filling OLH DoEs, which preserves the space filling and projection properties of the DoEs through the iterations. A PermGA algorithm was developed to generate MB OLHs, subsequently adapted for generation of infill MV test points as OLH DoEs, preserving good space filling and projection properties for the merged MB + MV test plan. The algorithm was further modified to address issues with non-orthogonal design spaces. A case study addressing the steady state engine mapping of a Gasoline Direct Injection was used to illustrate and validate the practical application of MB-MV sequence based on the developed PermGA algorithm.
{"title":"PermGA algorithm for a sequential optimal space filling DoE framework","authors":"M. R. Kianifar, F. Campean, A. Wood","doi":"10.1109/UKCI.2014.6930172","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930172","url":null,"abstract":"This paper presents the development and implementation of a customised Permutation Genetic Algorithm (PermGA) for a sequential Design of Experiment (DoE) framework based on space filling Optimal Latin Hypercube (OLH) designs. The work is motivated by multivariate engineering problems such as engine mapping experiments, which require efficient DoE strategies to minimise expensive testing. The DoE strategy is based on a flexible Model Building - Model Validation (MB-MV) sequence based on space filling OLH DoEs, which preserves the space filling and projection properties of the DoEs through the iterations. A PermGA algorithm was developed to generate MB OLHs, subsequently adapted for generation of infill MV test points as OLH DoEs, preserving good space filling and projection properties for the merged MB + MV test plan. The algorithm was further modified to address issues with non-orthogonal design spaces. A case study addressing the steady state engine mapping of a Gasoline Direct Injection was used to illustrate and validate the practical application of MB-MV sequence based on the developed PermGA algorithm.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114489632","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930151
J. Stamford, Arjab Singh Khuman, Jenny Carter, S. Ahmadi
One of the key aspects of games development is making Non-Playable Characters (NPC) behave more realistically in the environment. One of the main challenges is creating an NPC that is aware of its surroundings and acts accordingly. The A* Algorithm is widely used in the games development community to allow AI based characters to move around the environment however unlike real characters they are often given information about the environment without having to explore. This paper combines the use of the A* Algorithm with the occupancy grid technique to allow Non-Playable Characters to build their own representation of the environment and plan paths based on this information. The paper demonstrates the application of the approach and shows a range of testing and its limitations.
{"title":"Pathfinding in partially explored games environments: The application of the A∗ Algorithm with occupancy grids in Unity3D","authors":"J. Stamford, Arjab Singh Khuman, Jenny Carter, S. Ahmadi","doi":"10.1109/UKCI.2014.6930151","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930151","url":null,"abstract":"One of the key aspects of games development is making Non-Playable Characters (NPC) behave more realistically in the environment. One of the main challenges is creating an NPC that is aware of its surroundings and acts accordingly. The A* Algorithm is widely used in the games development community to allow AI based characters to move around the environment however unlike real characters they are often given information about the environment without having to explore. This paper combines the use of the A* Algorithm with the occupancy grid technique to allow Non-Playable Characters to build their own representation of the environment and plan paths based on this information. The paper demonstrates the application of the approach and shows a range of testing and its limitations.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129467464","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930173
Ahmad Alhindi, Qingfu Zhang, E. Tsang
This paper proposes an idea of using heuristic local search procedures specific for single-objective optimisation in multiobjectie evolutionary algorithms (MOEAs). In this paper, a multiobjective evolutionary algorithm based on decomposition (MOEA/D) hybridised with a multi-start single-objective metaheuristic called greedy randomised adaptive search procedure (GRASP). In our method a multiobjetive optimisation problem (MOP) is decomposed into a number of single-objecive subproblems and optimised in parallel by using neighbourhood information. The proposed GRASP alternates between subproblems to help them escape local Pareto optimal solutions. Experimental results have demonstrated that MOEA/D with GRASP outperforms the classical MOEA/D algorithm on the multiobjective 0-1 knapsack problem that is commonly used in the literature. It has also demonstrated that the use of greedy genetic crossover can significantly improve the algorithm performance.
{"title":"Hybridisation of decomposition and GRASP for combinatorial multiobjective optimisation","authors":"Ahmad Alhindi, Qingfu Zhang, E. Tsang","doi":"10.1109/UKCI.2014.6930173","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930173","url":null,"abstract":"This paper proposes an idea of using heuristic local search procedures specific for single-objective optimisation in multiobjectie evolutionary algorithms (MOEAs). In this paper, a multiobjective evolutionary algorithm based on decomposition (MOEA/D) hybridised with a multi-start single-objective metaheuristic called greedy randomised adaptive search procedure (GRASP). In our method a multiobjetive optimisation problem (MOP) is decomposed into a number of single-objecive subproblems and optimised in parallel by using neighbourhood information. The proposed GRASP alternates between subproblems to help them escape local Pareto optimal solutions. Experimental results have demonstrated that MOEA/D with GRASP outperforms the classical MOEA/D algorithm on the multiobjective 0-1 knapsack problem that is commonly used in the literature. It has also demonstrated that the use of greedy genetic crossover can significantly improve the algorithm performance.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"100 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115545901","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 : 2014-10-20DOI: 10.1109/UKCI.2014.6930163
Mohd-Hanif Yusoff, Yaochu Jin
In this paper, we investigate the influence of neural plasticity on the learning performance of echo state networks (ESNs) and supervised learning algorithms in training readout connections for two time series prediction problems including the sunspot time series and the Mackey Glass chaotic system. We implement two different plasticity rules that are expected to improve the prediction performance, namely, anti-Oja learning rule and the Bienenstock-Cooper-Munro (BCM) learning rule combined with both offline and online learning of the readout connections. Our experimental results have demonstrated that the neural plasticity can more significantly enhance the learning in offline learning than in online learning.
本文针对太阳黑子时间序列和Mackey Glass混沌系统这两个时间序列预测问题，研究了神经可塑性对回声状态网络(echo state network, ESNs)和监督学习算法的学习性能的影响。我们实现了两种不同的有望提高预测性能的可塑性规则，即anti-Oja学习规则和结合读出连接离线和在线学习的Bienenstock-Cooper-Munro (BCM)学习规则。我们的实验结果表明，与在线学习相比，神经可塑性对离线学习的促进作用更为显著。
{"title":"Modeling neural plasticity in echo state networks for time series prediction","authors":"Mohd-Hanif Yusoff, Yaochu Jin","doi":"10.1109/UKCI.2014.6930163","DOIUrl":"https://doi.org/10.1109/UKCI.2014.6930163","url":null,"abstract":"In this paper, we investigate the influence of neural plasticity on the learning performance of echo state networks (ESNs) and supervised learning algorithms in training readout connections for two time series prediction problems including the sunspot time series and the Mackey Glass chaotic system. We implement two different plasticity rules that are expected to improve the prediction performance, namely, anti-Oja learning rule and the Bienenstock-Cooper-Munro (BCM) learning rule combined with both offline and online learning of the readout connections. Our experimental results have demonstrated that the neural plasticity can more significantly enhance the learning in offline learning than in online learning.","PeriodicalId":315044,"journal":{"name":"2014 14th UK Workshop on Computational Intelligence (UKCI)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115203293","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：