Pub Date : 2015-07-12DOI: 10.1109/IJCNN.2015.7280512
Jingwei Yang, Huaping Liu, F. Sun, Meng Gao
Tactile sensors in the robotic fingertips are used to capture multiple object properties such as texture, roughness, spatial features, compliance or friction and therefore becomes a very important sense modality for intelligent robot. However, existing work neglects the intrinsic relation between different fingers which simultaneously contact the object. In this paper, a joint kernel sparse coding model is developed to tackle the multi-finger tactile sequence classification problem. In this model, the intrinsic relations between fingers are explicitly considered using the joint sparse coding which encourages different modal coding to share the same support. The experimental results show that the joint sparse coding achieves better performance than conventional sparse coding.
{"title":"Tactile sequence classification using joint kernel sparse coding","authors":"Jingwei Yang, Huaping Liu, F. Sun, Meng Gao","doi":"10.1109/IJCNN.2015.7280512","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280512","url":null,"abstract":"Tactile sensors in the robotic fingertips are used to capture multiple object properties such as texture, roughness, spatial features, compliance or friction and therefore becomes a very important sense modality for intelligent robot. However, existing work neglects the intrinsic relation between different fingers which simultaneously contact the object. In this paper, a joint kernel sparse coding model is developed to tackle the multi-finger tactile sequence classification problem. In this model, the intrinsic relations between fingers are explicitly considered using the joint sparse coding which encourages different modal coding to share the same support. The experimental results show that the joint sparse coding achieves better performance than conventional sparse coding.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"15 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75044003","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280764
E. Capecci, Josafath Israel Espinosa Ramos, N. Mammone, N. Kasabov, J. Duun-Henriksen, T. Kjaer, M. Campolo, F. L. Foresta, F. Morabito
Epilepsy is the most diffuse brain disorder that can affect people's lives even on its early stage. In this paper, we used for the first time the spiking neural networks (SNN) framework called NeuCube for the analysis of electroencephalography (EEG) data recorded from a person affected by Absence Epileptic (AE), using permutation entropy (PE) features. Our results demonstrated that the methodology constitutes a valuable tool for the analysis and understanding of functional changes in the brain in term of its spiking activity and connectivity. Future applications of the model aim at personalised modelling of epileptic data for the analysis and the event prediction.
{"title":"Modelling Absence Epilepsy seizure data in the NeuCube evolving spiking neural network architecture","authors":"E. Capecci, Josafath Israel Espinosa Ramos, N. Mammone, N. Kasabov, J. Duun-Henriksen, T. Kjaer, M. Campolo, F. L. Foresta, F. Morabito","doi":"10.1109/IJCNN.2015.7280764","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280764","url":null,"abstract":"Epilepsy is the most diffuse brain disorder that can affect people's lives even on its early stage. In this paper, we used for the first time the spiking neural networks (SNN) framework called NeuCube for the analysis of electroencephalography (EEG) data recorded from a person affected by Absence Epileptic (AE), using permutation entropy (PE) features. Our results demonstrated that the methodology constitutes a valuable tool for the analysis and understanding of functional changes in the brain in term of its spiking activity and connectivity. Future applications of the model aim at personalised modelling of epileptic data for the analysis and the event prediction.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"31 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75278856","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280582
Zehong Cao, L. Ko, K. Lai, Song-Bo Huang, Shuu-Jiun Wang, Chin-Teng Lin
Migraine is a chronic neurological disease characterized by recurrent moderate to severe headaches during a period like one month often in association with symptoms in human brain and autonomic nervous system. Normally, migraine symptoms can be categorized into four different stages: inter-ictal, pre-ictal, ictal, and post-ictal stages. Since migraine patients are difficulty knowing when they will suffer migraine attacks, therefore, early detection becomes an important issue, especially for low-frequency migraine patients who have less than 5 times attacks per month. The main goal of this study is to develop a migraine-stage classification system based on migraineurs' resting-state EEG power. We collect migraineurs' O1 and O2 EEG activities during closing eyes from occipital lobe to identify pre-ictal and non-pre-ictal stages. Self-Constructing Neural Fuzzy Inference Network (SONFIN) is adopted as the classifier in the migraine stages classification which can reach the better classification accuracy (66%) in comparison with other classifiers. The proposed system is helpful for migraineurs to obtain better treatment at the right time.σ
{"title":"Classification of migraine stages based on resting-state EEG power","authors":"Zehong Cao, L. Ko, K. Lai, Song-Bo Huang, Shuu-Jiun Wang, Chin-Teng Lin","doi":"10.1109/IJCNN.2015.7280582","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280582","url":null,"abstract":"Migraine is a chronic neurological disease characterized by recurrent moderate to severe headaches during a period like one month often in association with symptoms in human brain and autonomic nervous system. Normally, migraine symptoms can be categorized into four different stages: inter-ictal, pre-ictal, ictal, and post-ictal stages. Since migraine patients are difficulty knowing when they will suffer migraine attacks, therefore, early detection becomes an important issue, especially for low-frequency migraine patients who have less than 5 times attacks per month. The main goal of this study is to develop a migraine-stage classification system based on migraineurs' resting-state EEG power. We collect migraineurs' O1 and O2 EEG activities during closing eyes from occipital lobe to identify pre-ictal and non-pre-ictal stages. Self-Constructing Neural Fuzzy Inference Network (SONFIN) is adopted as the classifier in the migraine stages classification which can reach the better classification accuracy (66%) in comparison with other classifiers. The proposed system is helpful for migraineurs to obtain better treatment at the right time.σ","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"38 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74703307","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280736
Alexander Schulz, B. Hammer
Discriminative dimensionality reduction refers to the goal of visualizing given high-dimensional data in the plane such that the structure relevant for a specified aspect is displayed. While this framework has been successfully applied to visualize data with auxiliary label information, its extension to real-valued information is lacking. In this contribution, we propose a general way to shape data distances based on auxiliary real-valued information with the Fisher metric which is derived from a Gaussian process model of the data. This can directly be integrated into high quality non-linear dimensionality reduction methods such as t-SNE, as we will demonstrate in artificial as well as real life benchmarks.
{"title":"Discriminative dimensionality reduction for regression problems using the Fisher metric","authors":"Alexander Schulz, B. Hammer","doi":"10.1109/IJCNN.2015.7280736","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280736","url":null,"abstract":"Discriminative dimensionality reduction refers to the goal of visualizing given high-dimensional data in the plane such that the structure relevant for a specified aspect is displayed. While this framework has been successfully applied to visualize data with auxiliary label information, its extension to real-valued information is lacking. In this contribution, we propose a general way to shape data distances based on auxiliary real-valued information with the Fisher metric which is derived from a Gaussian process model of the data. This can directly be integrated into high quality non-linear dimensionality reduction methods such as t-SNE, as we will demonstrate in artificial as well as real life benchmarks.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"76 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73231328","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280369
A. Arista-Jalife, E. Gómez-Ramírez
The Polynomial Cellular Neural Network (PCNN) is a fully parallel, scalable, non-linear processor that uses polynomial terms to solve non-linear problems in a lattice fashion. The parallel nature of such processor allows every neuron (or cell) to gather information from the nearby neurons and independently process the retrieved values by employing non-linear functions and synaptic weights. Nonetheless, one of the main challenges of the PCNN is the determination of the synaptic weights in order to achieve the desired behavior. In this paper, a new training method is presented, based on two fundamental concepts: the root location training method and the polynomial surfaces. The proposed training method is able to straightforwardly determine the requested synaptic weights for any outer-totallistic cellular automata behavior. In order to deliver a proof of the potential of such proposition, several image processing tasks are performed with a single layered PCNN.
{"title":"One-shot Training of Polynomial Cellular Neural Networks and applications in image processing","authors":"A. Arista-Jalife, E. Gómez-Ramírez","doi":"10.1109/IJCNN.2015.7280369","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280369","url":null,"abstract":"The Polynomial Cellular Neural Network (PCNN) is a fully parallel, scalable, non-linear processor that uses polynomial terms to solve non-linear problems in a lattice fashion. The parallel nature of such processor allows every neuron (or cell) to gather information from the nearby neurons and independently process the retrieved values by employing non-linear functions and synaptic weights. Nonetheless, one of the main challenges of the PCNN is the determination of the synaptic weights in order to achieve the desired behavior. In this paper, a new training method is presented, based on two fundamental concepts: the root location training method and the polynomial surfaces. The proposed training method is able to straightforwardly determine the requested synaptic weights for any outer-totallistic cellular automata behavior. In order to deliver a proof of the potential of such proposition, several image processing tasks are performed with a single layered PCNN.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"89 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72810675","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280368
M. Atencia, G. Joya
This paper proposes an adaptive control algorithm, which is designed by adding a parametric identification method to a non-linear controller. The identification module is built upon the Hopfield neural network, resulting in an unconventional network with time-varying weights and biases. The convergence of the estimations of the parameters of a dynamical system was proved in previous work, as long as the system inputs can be freely manipulated to provide persistent excitation. Henceforth the behaviour of the closed-loop system, when the inputs result from the controller equations, is here analyzed in order to assess both the tracking performance of the full adaptive controller and the identification ability of the neural estimator. The algorithm is applied to an idealized robotic system with two joints, whose positions and velocities are required to follow, as closely as possible, a prescribed reference trajectory. The simulation results show a satisfactory control performance, since the demanded trajectory is almost accurately followed. The estimated values also converge to the correct parameters, as long as the controller provides sufficiently rich signals to the system. The results are similar to a conventional least-squares adaptive controller, with a much lower computational cost.
{"title":"Hopfield networks: from optimization to adaptive control","authors":"M. Atencia, G. Joya","doi":"10.1109/IJCNN.2015.7280368","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280368","url":null,"abstract":"This paper proposes an adaptive control algorithm, which is designed by adding a parametric identification method to a non-linear controller. The identification module is built upon the Hopfield neural network, resulting in an unconventional network with time-varying weights and biases. The convergence of the estimations of the parameters of a dynamical system was proved in previous work, as long as the system inputs can be freely manipulated to provide persistent excitation. Henceforth the behaviour of the closed-loop system, when the inputs result from the controller equations, is here analyzed in order to assess both the tracking performance of the full adaptive controller and the identification ability of the neural estimator. The algorithm is applied to an idealized robotic system with two joints, whose positions and velocities are required to follow, as closely as possible, a prescribed reference trajectory. The simulation results show a satisfactory control performance, since the demanded trajectory is almost accurately followed. The estimated values also converge to the correct parameters, as long as the controller provides sufficiently rich signals to the system. The results are similar to a conventional least-squares adaptive controller, with a much lower computational cost.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"70 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74093713","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280623
Moritz Schneider, J. Adamy
Motivation and emotion are inseparable component factors of value systems in living beings, which enable them to act purposefully in a partially unknown and sometimes unforgiving environment. Value systems that drive innate reinforcement learning mechanisms have been identified as key factors in self-directed control and autonomous development towards higher intelligence and seem crucial in the development of a concept of “self” in sentient beings [1]. This contribution is concerned with the relationship between artificial learning control systems and innate value systems. In particular, we adapt the state-of-the-art model of motivational processes based on reduction of generalized drives towards higher flexibility, expressivity and representation capability. A framework for modelling self-adaptive value systems, which develop autonomously starting from an inherited (or designed) innate representation, within a learning control system architecture is formulated. We discuss the relationship of anticipated effects in this control architecture with psychological theory on motivations and contrast our framework with related approaches.
{"title":"Artificial motivations based on drive-reduction theory in self-referential model-building control systems","authors":"Moritz Schneider, J. Adamy","doi":"10.1109/IJCNN.2015.7280623","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280623","url":null,"abstract":"Motivation and emotion are inseparable component factors of value systems in living beings, which enable them to act purposefully in a partially unknown and sometimes unforgiving environment. Value systems that drive innate reinforcement learning mechanisms have been identified as key factors in self-directed control and autonomous development towards higher intelligence and seem crucial in the development of a concept of “self” in sentient beings [1]. This contribution is concerned with the relationship between artificial learning control systems and innate value systems. In particular, we adapt the state-of-the-art model of motivational processes based on reduction of generalized drives towards higher flexibility, expressivity and representation capability. A framework for modelling self-adaptive value systems, which develop autonomously starting from an inherited (or designed) innate representation, within a learning control system architecture is formulated. We discuss the relationship of anticipated effects in this control architecture with psychological theory on motivations and contrast our framework with related approaches.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"27 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84809301","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280515
Feng Sha, C. Bae, Guang Liu, XiMeng Zhao, Yuk Ying Chung, W. Yeh, Xiangjian He
Particle Swarm Optimization has been used in many research and application domain popularly since its development and improvement. Due to its fast and accurate solution searching, PSO has become one of the high potential tools to provide better outcomes to solve many practical problems. In image processing and object tracking applications, PSO also indicates to have good performance in both linear and non-linear object moving pattern, many scientists conduct development and research to implement not only basic PSO but also improved methods in enhancing the efficiency of the algorithm to achieve precise object tracking orbit. This paper is aim to propose a new improved PSO by comparing the inertia weight and constriction factor of PSO. It provides faster and more accurate object tracking process since the proposed algorithm can inherit some useful information from the previous solution to perform the dynamic particle movement when other better solution exists. The testing experiments have been done for different types of video, results showed that the proposed algorithm can have better quality of tracking performance and faster object retrieval speed. The proposed approach has been developed in C++ environment and tested against videos and objects with multiple moving patterns to demonstrate the benefits with precise object similarity.
{"title":"A probability-dynamic Particle Swarm Optimization for object tracking","authors":"Feng Sha, C. Bae, Guang Liu, XiMeng Zhao, Yuk Ying Chung, W. Yeh, Xiangjian He","doi":"10.1109/IJCNN.2015.7280515","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280515","url":null,"abstract":"Particle Swarm Optimization has been used in many research and application domain popularly since its development and improvement. Due to its fast and accurate solution searching, PSO has become one of the high potential tools to provide better outcomes to solve many practical problems. In image processing and object tracking applications, PSO also indicates to have good performance in both linear and non-linear object moving pattern, many scientists conduct development and research to implement not only basic PSO but also improved methods in enhancing the efficiency of the algorithm to achieve precise object tracking orbit. This paper is aim to propose a new improved PSO by comparing the inertia weight and constriction factor of PSO. It provides faster and more accurate object tracking process since the proposed algorithm can inherit some useful information from the previous solution to perform the dynamic particle movement when other better solution exists. The testing experiments have been done for different types of video, results showed that the proposed algorithm can have better quality of tracking performance and faster object retrieval speed. The proposed approach has been developed in C++ environment and tested against videos and objects with multiple moving patterns to demonstrate the benefits with precise object similarity.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"61 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84896783","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280617
Cheng Luo, Xiongcai Cai, N. Chowdhury
User preferences for products are constantly drifting over time as product perception and popularity are changing when new fashions or products emerge. Therefore, the ability to model the tendency of both user preferences and product attractiveness is vital to the design of recommender systems (RSs). However, conventional methods in RSs are incapable of modeling such a tendency accordingly, leading to unsatisfactory recommendation performance in many real-world deployments. In this paper, we develop a novel probabilistic temporal bilinear model for RSs, exploiting both temporal properties and dynamic information in user preferences and item attractiveness derived from the users' feedback over items, to simultaneously track latent factors that represent user preferences and item attractiveness. A learning and inference algorithm combining a sequential Monte Carlo method and the EM algorithm for this model is also developed to tackle the top-k recommendation problem over time. The proposed model is evaluated on three benchmark datasets. The experimental results demonstrate that our proposed model significantly outperforms a variety of existing methods for top-k recommendation.
{"title":"Probabilistic temporal bilinear model for temporal dynamic recommender systems","authors":"Cheng Luo, Xiongcai Cai, N. Chowdhury","doi":"10.1109/IJCNN.2015.7280617","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280617","url":null,"abstract":"User preferences for products are constantly drifting over time as product perception and popularity are changing when new fashions or products emerge. Therefore, the ability to model the tendency of both user preferences and product attractiveness is vital to the design of recommender systems (RSs). However, conventional methods in RSs are incapable of modeling such a tendency accordingly, leading to unsatisfactory recommendation performance in many real-world deployments. In this paper, we develop a novel probabilistic temporal bilinear model for RSs, exploiting both temporal properties and dynamic information in user preferences and item attractiveness derived from the users' feedback over items, to simultaneously track latent factors that represent user preferences and item attractiveness. A learning and inference algorithm combining a sequential Monte Carlo method and the EM algorithm for this model is also developed to tackle the top-k recommendation problem over time. The proposed model is evaluated on three benchmark datasets. The experimental results demonstrate that our proposed model significantly outperforms a variety of existing methods for top-k recommendation.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"35 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85007625","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 : 2015-07-12DOI: 10.1109/IJCNN.2015.7280390
Andrew Mundy, James C. Knight, T. Stewart, S. Furber
By building and simulating neural systems we hope to understand how the brain may work and use this knowledge to build neural and cognitive systems to tackle engineering problems. The Neural Engineering Framework (NEF) is a hypothesis about how such systems may be constructed and has recently been used to build the world's first functional brain model, Spaun. However, while the NEF simplifies the design of neural networks, simulating them using standard computer hardware is still computationally expensive - often running far slower than biological real-time and scaling very poorly: problems the SpiNNaker neuromorphic simulator was designed to solve. In this paper we (1) argue that employing the same model of computation used for simulating general purpose spiking neural networks on SpiNNaker for NEF models results in suboptimal use of the architecture, and (2) provide and evaluate an alternative simulation scheme which overcomes the memory and compute challenges posed by the NEF. This proposed method uses factored weight matrices to reduce memory usage by around 90% and, in some cases, simulate 2000 neurons on a processing core - double the SpiNNaker architectural target.
{"title":"An efficient SpiNNaker implementation of the Neural Engineering Framework","authors":"Andrew Mundy, James C. Knight, T. Stewart, S. Furber","doi":"10.1109/IJCNN.2015.7280390","DOIUrl":"https://doi.org/10.1109/IJCNN.2015.7280390","url":null,"abstract":"By building and simulating neural systems we hope to understand how the brain may work and use this knowledge to build neural and cognitive systems to tackle engineering problems. The Neural Engineering Framework (NEF) is a hypothesis about how such systems may be constructed and has recently been used to build the world's first functional brain model, Spaun. However, while the NEF simplifies the design of neural networks, simulating them using standard computer hardware is still computationally expensive - often running far slower than biological real-time and scaling very poorly: problems the SpiNNaker neuromorphic simulator was designed to solve. In this paper we (1) argue that employing the same model of computation used for simulating general purpose spiking neural networks on SpiNNaker for NEF models results in suboptimal use of the architecture, and (2) provide and evaluate an alternative simulation scheme which overcomes the memory and compute challenges posed by the NEF. This proposed method uses factored weight matrices to reduce memory usage by around 90% and, in some cases, simulate 2000 neurons on a processing core - double the SpiNNaker architectural target.","PeriodicalId":6539,"journal":{"name":"2015 International Joint Conference on Neural Networks (IJCNN)","volume":"48 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2015-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85078014","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：