Pub Date : 2011-10-03DOI: 10.1109/IJCNN.2011.6033416
Ken Saito, K. Okazaki, K. Sakata, T. Ogiwara, Y. Sekine, F. Uchikoba
This paper presents the locomotion generator of MEMS (Micro Electro Mechanical Systems) micro robot. The locomotion generator demonstrates the locomotion of the micro robot, controlled by the P-HINN (Pulse-Type Hardware Inhibitory Neural Networks). P-HINN generates oscillatory patterns of electrical activity such as living organisms. Basic components are the cell body models and inhibitory synaptic models. P-HINN has the same basic features of biological neurons such as threshold, refractory period, spatio-temporal summation characteristics and enables the generation of continuous action potentials. P-HINN was constructed by MOSFETs, can be integrated by CMOS technology. Same as the living organisms P-HINN realized the robot control without using any software programs, or A/D converters. The size of the micro robot fabricated by the MEMS technology was 4×4×3.5 mm. The frame of the robot was made of silicon wafer, equipped with the rotary type actuators, the link mechanisms and 6 legs. The MEMS micro robot emulated the locomotion method and the neural networks of the insect by the rotary type actuators, link mechanisms and P-HINN. As a result, we show that P-HINN can control the forward and backward locomotion of fabricated MEMS micro robot, and also switched the direction by inputting the external trigger pulse. The locomotion speed was 19.5 mm/min and the step width was 1.3 mm.
{"title":"Pulse-Type Hardware Inhibitory Neural Networks for MEMS micro robot using CMOS technology","authors":"Ken Saito, K. Okazaki, K. Sakata, T. Ogiwara, Y. Sekine, F. Uchikoba","doi":"10.1109/IJCNN.2011.6033416","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033416","url":null,"abstract":"This paper presents the locomotion generator of MEMS (Micro Electro Mechanical Systems) micro robot. The locomotion generator demonstrates the locomotion of the micro robot, controlled by the P-HINN (Pulse-Type Hardware Inhibitory Neural Networks). P-HINN generates oscillatory patterns of electrical activity such as living organisms. Basic components are the cell body models and inhibitory synaptic models. P-HINN has the same basic features of biological neurons such as threshold, refractory period, spatio-temporal summation characteristics and enables the generation of continuous action potentials. P-HINN was constructed by MOSFETs, can be integrated by CMOS technology. Same as the living organisms P-HINN realized the robot control without using any software programs, or A/D converters. The size of the micro robot fabricated by the MEMS technology was 4×4×3.5 mm. The frame of the robot was made of silicon wafer, equipped with the rotary type actuators, the link mechanisms and 6 legs. The MEMS micro robot emulated the locomotion method and the neural networks of the insect by the rotary type actuators, link mechanisms and P-HINN. As a result, we show that P-HINN can control the forward and backward locomotion of fabricated MEMS micro robot, and also switched the direction by inputting the external trigger pulse. The locomotion speed was 19.5 mm/min and the step width was 1.3 mm.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115090626","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033394
K. Shibata, Hiroki Utsunomiya
In this paper, by the combination of reinforcement learning and a recurrent neural network, the authors try to provide an explanation for the question: why humans can discover the meaning of patterns and acquire appropriate behaviors based on it. Using a system with a real movable camera, it is demonstrated in a simple task in which the system discovers pattern meaning from delayed rewards by reinforcement learning with a recurrent neural network. When the system moves its camera to the direction of an arrow presented on a display, it can get a reward. One kind of arrow is chosen randomly among four kinds at each episode, and the input of the network is 1,560 visual signals from the camera. After learning, the system could move its camera to the arrow direction. It was found that some hidden neurons represented the arrow direction not depending on the presented arrow pattern and kept it after the arrow disappeared from the image, even though no arrow was seen when it was rewarded and no one told the system that the arrow direction is important to get the reward. Generalization to some new arrow patterns and associative memory function also can be seen to some extent.
{"title":"Discovery of pattern meaning from delayed rewards by reinforcement learning with a recurrent neural network","authors":"K. Shibata, Hiroki Utsunomiya","doi":"10.1109/IJCNN.2011.6033394","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033394","url":null,"abstract":"In this paper, by the combination of reinforcement learning and a recurrent neural network, the authors try to provide an explanation for the question: why humans can discover the meaning of patterns and acquire appropriate behaviors based on it. Using a system with a real movable camera, it is demonstrated in a simple task in which the system discovers pattern meaning from delayed rewards by reinforcement learning with a recurrent neural network. When the system moves its camera to the direction of an arrow presented on a display, it can get a reward. One kind of arrow is chosen randomly among four kinds at each episode, and the input of the network is 1,560 visual signals from the camera. After learning, the system could move its camera to the arrow direction. It was found that some hidden neurons represented the arrow direction not depending on the presented arrow pattern and kept it after the arrow disappeared from the image, even though no arrow was seen when it was rewarded and no one told the system that the arrow direction is important to get the reward. Generalization to some new arrow patterns and associative memory function also can be seen to some extent.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115201820","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033514
F. Corinto, A. Ascoli, M. Gilli
The design of pattern recognition systems based on memristive oscillatory networks need to include a detailed study of the dynamics of the networks and their basic components. A simple two-cell network of this kind, where each cell is made up of a linear circuitry in parallel with a nonlinear memristive element, was found to experience a rich gamut of nonlinear behaviors. In particular, for a synchronization scenario with almost-sinusoidal oscillations, the memristive elements used in the cells exhibited an unusual current-voltage characteristic. This work focuses on the dynamics of the single cell under this synchronization scenario, and, modeling the linear circuitry with a sinusoidal voltage source, analytically derives a rigorous classification of all possible current-voltage characteristics of the periodically-driven memristive element on the basis of amplitude-angular frequency ratio and time hystory of the input source.
{"title":"Class of all i-v dynamics for memristive elements in pattern recognition systems","authors":"F. Corinto, A. Ascoli, M. Gilli","doi":"10.1109/IJCNN.2011.6033514","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033514","url":null,"abstract":"The design of pattern recognition systems based on memristive oscillatory networks need to include a detailed study of the dynamics of the networks and their basic components. A simple two-cell network of this kind, where each cell is made up of a linear circuitry in parallel with a nonlinear memristive element, was found to experience a rich gamut of nonlinear behaviors. In particular, for a synchronization scenario with almost-sinusoidal oscillations, the memristive elements used in the cells exhibited an unusual current-voltage characteristic. This work focuses on the dynamics of the single cell under this synchronization scenario, and, modeling the linear circuitry with a sinusoidal voltage source, analytically derives a rigorous classification of all possible current-voltage characteristics of the periodically-driven memristive element on the basis of amplitude-angular frequency ratio and time hystory of the input source.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115326691","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033383
M. Kimura, T. Isokawa, H. Nishimura, N. Matsui
Properties and performances of associative memories, based on Complex-valued Synergetic Computer (CVSC), are explored in this paper. All the parameters of CVSC are encoded by complex values. CVSC is extended from the conventional Synergetic Computer (RVSC) in which the parameters are real values. Performances of associative memories in CVSC are investigated through a problem of image retrievals where the input images are partially occluded or noise-affected. From the experimental results concerning the retrieval performances related to various sizes of images and different levels of defectiveness of input images, we found that CVSC outperforms RVSC.
{"title":"On retrieval performance of associative memory by Complex-valued Synergetic Computer","authors":"M. Kimura, T. Isokawa, H. Nishimura, N. Matsui","doi":"10.1109/IJCNN.2011.6033383","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033383","url":null,"abstract":"Properties and performances of associative memories, based on Complex-valued Synergetic Computer (CVSC), are explored in this paper. All the parameters of CVSC are encoded by complex values. CVSC is extended from the conventional Synergetic Computer (RVSC) in which the parameters are real values. Performances of associative memories in CVSC are investigated through a problem of image retrievals where the input images are partially occluded or noise-affected. From the experimental results concerning the retrieval performances related to various sizes of images and different levels of defectiveness of input images, we found that CVSC outperforms RVSC.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"98 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122621531","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033657
S. Crone, N. Kourentzes
Forecasting future electricity load represents one of the most prominent areas of electrical engineering, in which artificial neural networks (NN) are routinely applied in practice. The common approach to overcome the complexity of building NNs for high-frequency load data is to segment the time series into simpler and more homogeneous subseries, e.g. seven subseries of hourly loads of only Mondays, Tuesdays etc. These are forecasted independently, using a separate NN model, and then recombined to provide a complete trace forecast for the next days ahead. Despite the empirical importance of load forecasting, and the high operational cost associated with forecast errors, the potential benefits of segmenting time series into subseries have not been evaluated in an empirical comparison. This paper assesses the accuracy of segmenting continuous time series into daily subseries, versus forecasting the original, continuous time series with NNs. Accuracy on hourly UK load data is provided in a valid experimental design, using multiple rolling time origins and robust error metrics in comparison to statistical benchmark algorithms. Results indicate the superior performance of NN on continuous, non-segmented time series, in contrast to best practices in research, practice and software implementations.
{"title":"The impact of preprocessing on forecasting electrical load: An empirical evaluation of segmenting time series into subseries","authors":"S. Crone, N. Kourentzes","doi":"10.1109/IJCNN.2011.6033657","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033657","url":null,"abstract":"Forecasting future electricity load represents one of the most prominent areas of electrical engineering, in which artificial neural networks (NN) are routinely applied in practice. The common approach to overcome the complexity of building NNs for high-frequency load data is to segment the time series into simpler and more homogeneous subseries, e.g. seven subseries of hourly loads of only Mondays, Tuesdays etc. These are forecasted independently, using a separate NN model, and then recombined to provide a complete trace forecast for the next days ahead. Despite the empirical importance of load forecasting, and the high operational cost associated with forecast errors, the potential benefits of segmenting time series into subseries have not been evaluated in an empirical comparison. This paper assesses the accuracy of segmenting continuous time series into daily subseries, versus forecasting the original, continuous time series with NNs. Accuracy on hourly UK load data is provided in a valid experimental design, using multiple rolling time origins and robust error metrics in comparison to statistical benchmark algorithms. Results indicate the superior performance of NN on continuous, non-segmented time series, in contrast to best practices in research, practice and software implementations.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121874768","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033438
J. H. Barrón-Zambrano, C. Torres-Huitzil
Nowadays, the locomotion control research field has been pretty active and has produced different approaches for legged robots. From biological studies, it is known that fundamental rhythmic periodical signals for locomotion are produced by Central Pattern Generator (CPG) and the main part of the coordination takes place in the central nervous system. In spite of the CPG-utility, there are few training methodologies to generate the rhythmic signals based in CPG models. In this paper, an automatic method to find the synaptic weights to generate three basic gaits using Genetic Algorithms (GA) is presented. The method is based on the analysis of the oscillator behavior and its interactions with other oscillators, in a network. The oscillator model used in this work is the proposed by Van Der Pol (VDP). A two-phase GA is adapted: (i) to find the parameter values to produce oscillations and (ii) to generate the weight values of the interconnections between oscillators. The results show the feasibility of the presented method to find the parameters to generate different gaits. The implementation takes advantage that the fitness function works directly with the oscillator and the network. So, knowledge about the robot dynamic is not necessary. The GA based approach uses small population and limited numbers of generations, ideal to be processed on either computers with reduced resources or hardware implementations.
目前，运动控制的研究领域非常活跃，并产生了不同的方法来控制有腿机器人。从生物学研究可知，运动的基本节律周期信号是由中枢模式发生器(Central Pattern Generator, CPG)产生的，而协调的主要部分发生在中枢神经系统。尽管CPG很实用，但很少有训练方法来生成基于CPG模型的节奏信号。本文提出了一种利用遗传算法自动寻找突触权值以生成三种基本步态的方法。该方法基于对网络中振子行为及其与其他振子相互作用的分析。本工作中使用的振荡器模型是由Van Der Pol (VDP)提出的。采用两相遗传算法:(i)找到产生振荡的参数值，(ii)产生振荡之间互连的权值。实验结果表明，该方法能够有效地找到生成不同步态的参数。该实现利用了适应度函数直接与振荡器和网络工作的优点。所以，关于机器人动力学的知识是不必要的。基于遗传算法的方法使用较少的种群和有限的代数，非常适合在资源较少或硬件实现较少的计算机上进行处理。
{"title":"Two-phase GA parameter tunning method of CPGs for quadruped gaits","authors":"J. H. Barrón-Zambrano, C. Torres-Huitzil","doi":"10.1109/IJCNN.2011.6033438","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033438","url":null,"abstract":"Nowadays, the locomotion control research field has been pretty active and has produced different approaches for legged robots. From biological studies, it is known that fundamental rhythmic periodical signals for locomotion are produced by Central Pattern Generator (CPG) and the main part of the coordination takes place in the central nervous system. In spite of the CPG-utility, there are few training methodologies to generate the rhythmic signals based in CPG models. In this paper, an automatic method to find the synaptic weights to generate three basic gaits using Genetic Algorithms (GA) is presented. The method is based on the analysis of the oscillator behavior and its interactions with other oscillators, in a network. The oscillator model used in this work is the proposed by Van Der Pol (VDP). A two-phase GA is adapted: (i) to find the parameter values to produce oscillations and (ii) to generate the weight values of the interconnections between oscillators. The results show the feasibility of the presented method to find the parameters to generate different gaits. The implementation takes advantage that the fitness function works directly with the oscillator and the network. So, knowledge about the robot dynamic is not necessary. The GA based approach uses small population and limited numbers of generations, ideal to be processed on either computers with reduced resources or hardware implementations.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129843984","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033479
B. Ribeiro, Ningshan Chen
MANY dimensionality reduction algorithms have been proposed easing both tasks of visualization and classification in high dimension problems. Despite the different motivations they can be cast in a graph embedding framework. In this paper we address weighted graph subspace learning methods for bankruptcy analysis. The rationale behind re-embedding the data in a lower dimensional space that would be better filled is twofold: to get the most compact representation (visualization) and to make subsequent processing of data more easy (classification). The approaches used, Graph regularized Non-Negative Matrix Factorization (GNMF) and Spatially Smooth Subspace Learning (SSSL), construct an affinity weight graph matrix to encode geometrical information and to learn in the training set the subspace models that enhance visualization and are able to ease the task of bankruptcy prediction. The experimental results on a real problem of French companies show that from the perspective of financial problem analysis the methodology is quite effective.
{"title":"Graph weighted subspace learning models in bankruptcy","authors":"B. Ribeiro, Ningshan Chen","doi":"10.1109/IJCNN.2011.6033479","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033479","url":null,"abstract":"MANY dimensionality reduction algorithms have been proposed easing both tasks of visualization and classification in high dimension problems. Despite the different motivations they can be cast in a graph embedding framework. In this paper we address weighted graph subspace learning methods for bankruptcy analysis. The rationale behind re-embedding the data in a lower dimensional space that would be better filled is twofold: to get the most compact representation (visualization) and to make subsequent processing of data more easy (classification). The approaches used, Graph regularized Non-Negative Matrix Factorization (GNMF) and Spatially Smooth Subspace Learning (SSSL), construct an affinity weight graph matrix to encode geometrical information and to learn in the training set the subspace models that enhance visualization and are able to ease the task of bankruptcy prediction. The experimental results on a real problem of French companies show that from the perspective of financial problem analysis the methodology is quite effective.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"209 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124234428","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033546
Jin‐Song Pei, J. P. Wright, S. Masri, E. Mai, A. Smyth
This paper reports a continuous development of the work by the authors presented at IJCNN 2005 & 2007 [1, 2]. A series of parsimonious universal approximator architectures with pre-defined values for weights and biases called “neural network prototypes” are proposed and used in a repetitive and systematic manner for the initialization of sigmoidal neural networks in function approximation. This paper provides a more in-depth literature review, presents one training example using laboratory data indicating quick convergence and trained sigmoidal neural networks with stable generalization capability, and discusses the complexity measure in [3, 4]. This study centers on approximating a subset of static nonlinear target functions - mechanical restoring force considered as a function of system states (displacement and velocity) for single-degree-of-freedom systems. We strive for efficient and rigorous constructive methods for sigmoidal neural networks to solve function approximation problems in this engineering mechanics application and beyond. Future work is identified.
{"title":"Toward constructive methods for sigmoidal neural networks - function approximation in engineering mechanics applications","authors":"Jin‐Song Pei, J. P. Wright, S. Masri, E. Mai, A. Smyth","doi":"10.1109/IJCNN.2011.6033546","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033546","url":null,"abstract":"This paper reports a continuous development of the work by the authors presented at IJCNN 2005 & 2007 [1, 2]. A series of parsimonious universal approximator architectures with pre-defined values for weights and biases called “neural network prototypes” are proposed and used in a repetitive and systematic manner for the initialization of sigmoidal neural networks in function approximation. This paper provides a more in-depth literature review, presents one training example using laboratory data indicating quick convergence and trained sigmoidal neural networks with stable generalization capability, and discusses the complexity measure in [3, 4]. This study centers on approximating a subset of static nonlinear target functions - mechanical restoring force considered as a function of system states (displacement and velocity) for single-degree-of-freedom systems. We strive for efficient and rigorous constructive methods for sigmoidal neural networks to solve function approximation problems in this engineering mechanics application and beyond. Future work is identified.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121438073","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033311
O. Bichler, D. Querlioz, S. Thorpe, J. Bourgoin, C. Gamrat
In this paper, we present a novel approach to extract complex and overlapping temporally correlated features directly from spike-based dynamic vision sensors. A spiking neural network capable of performing multilayer unsupervised learning through Spike-Timing-Dependent Plasticity is introduced. It shows exceptional performances at detecting cars passing on a freeway recorded with a dynamic vision sensor, after only 10 minutes of fully unsupervised learning. Our methodology is thoroughly explained and first applied to a simpler example of ball trajectory learning. Two unsupervised learning strategies are investigated for advanced features learning. Robustness of our network to synaptic and neuron variability is assessed and virtual immunity to noise and jitter is demonstrated.
{"title":"Unsupervised features extraction from asynchronous silicon retina through Spike-Timing-Dependent Plasticity","authors":"O. Bichler, D. Querlioz, S. Thorpe, J. Bourgoin, C. Gamrat","doi":"10.1109/IJCNN.2011.6033311","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033311","url":null,"abstract":"In this paper, we present a novel approach to extract complex and overlapping temporally correlated features directly from spike-based dynamic vision sensors. A spiking neural network capable of performing multilayer unsupervised learning through Spike-Timing-Dependent Plasticity is introduced. It shows exceptional performances at detecting cars passing on a freeway recorded with a dynamic vision sensor, after only 10 minutes of fully unsupervised learning. Our methodology is thoroughly explained and first applied to a simpler example of ball trajectory learning. Two unsupervised learning strategies are investigated for advanced features learning. Robustness of our network to synaptic and neuron variability is assessed and virtual immunity to noise and jitter is demonstrated.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121664035","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 : 2011-10-03DOI: 10.1109/IJCNN.2011.6033209
Ting Huang, Derong Liu
In this paper, we apply adaptive dynamic programming to the residential energy system control and management, with an emphasis on home battery use connected to power grids. The proposed scheme is built upon a self-learning architecture with only a single critic module instead of the action-critic dual module architecture. The novelty of the present scheme is its ability to improve the performance as it learns and gains more experience in real-time operations under uncertain changes of the environment. Simulation results demonstrate that the proposed scheme can achieve the minimum electricity cost for residential customers.
{"title":"Residential energy system control and management using adaptive dynamic programming","authors":"Ting Huang, Derong Liu","doi":"10.1109/IJCNN.2011.6033209","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033209","url":null,"abstract":"In this paper, we apply adaptive dynamic programming to the residential energy system control and management, with an emphasis on home battery use connected to power grids. The proposed scheme is built upon a self-learning architecture with only a single critic module instead of the action-critic dual module architecture. The novelty of the present scheme is its ability to improve the performance as it learns and gains more experience in real-time operations under uncertain changes of the environment. Simulation results demonstrate that the proposed scheme can achieve the minimum electricity cost for residential customers.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"11 8","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113972026","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：