Pub Date : 2011-10-03DOI: 10.1109/IJCNN.2011.6033442
Amirhosein Shantia, Eric Begue, M. Wiering
Real Time Strategy Games are one of the most popular game schemes in PC markets and offer a dynamic environment that involves several interacting agents. The core strategies that need to be developed in these games are unit micro management, building order, resource management, and the game main tactic. Unfortunately, current games only use scripted and fixed behaviors for their artificial intelligence (AI), and the player can easily learn the counter measures to defeat the AI. In this paper, we describe a system based on neural networks that controls a set of units of the same type in the popular game StarCraft. Using the neural networks, the units will either choose a unit to attack or evade from the battlefield. The system uses reinforcement learning combined with neural networks using online Sarsa and neural-fitted Sarsa, both with a short term memory reward function. We also present an incremental learning method for training the units for larger scenarios involving more units using trained neural networks on smaller scenarios. Additionally, we developed a novel sensing system to feed the environment data to the neural networks using separate vision grids. The simulation results show superior performance against the human-made AI scripts in StarCraft.
{"title":"Connectionist reinforcement learning for intelligent unit micro management in StarCraft","authors":"Amirhosein Shantia, Eric Begue, M. Wiering","doi":"10.1109/IJCNN.2011.6033442","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033442","url":null,"abstract":"Real Time Strategy Games are one of the most popular game schemes in PC markets and offer a dynamic environment that involves several interacting agents. The core strategies that need to be developed in these games are unit micro management, building order, resource management, and the game main tactic. Unfortunately, current games only use scripted and fixed behaviors for their artificial intelligence (AI), and the player can easily learn the counter measures to defeat the AI. In this paper, we describe a system based on neural networks that controls a set of units of the same type in the popular game StarCraft. Using the neural networks, the units will either choose a unit to attack or evade from the battlefield. The system uses reinforcement learning combined with neural networks using online Sarsa and neural-fitted Sarsa, both with a short term memory reward function. We also present an incremental learning method for training the units for larger scenarios involving more units using trained neural networks on smaller scenarios. Additionally, we developed a novel sensing system to feed the environment data to the neural networks using separate vision grids. The simulation results show superior performance against the human-made AI scripts in StarCraft.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"81 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":"114559200","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.2011.6033511
C. Qu, Bertram E. Shi
Neurons tuned to binocular disparity in area V1 are hypothesized to be responsible for short latency binocular vergence movements, which align the two eyes on the same object as it moves in depth. Disparity selective neurons in V1 are not only selective to disparity, but also to other visual stimulus dimensions, in particular orientation. In this work, we explore the role of neurons tuned to different orientations in binocular vergence control. We trained an artificial binocular vision system to execute corrective vergence movements based on the outputs of disparity selective neurons tuned to different orientations and scales. As might be expected, we find that neurons tuned to vertical orientations have the strongest effect on the vergence eye movements. The effect of neurons tuned to other orientations decreases as the tuned orientation approaches horizontal. Although adding neurons tuned to non-vertical orientations does not appear to improve vergence tracking accuracy, we find that neurons tuned to non-vertical orientations still play critical roles in binocular vergence control. First, they decrease the time required to learn the vergence control strategy. Second, they also increase the effective range of vergence control.
{"title":"The role of orientation diversity in binocular vergence control","authors":"C. Qu, Bertram E. Shi","doi":"10.1109/IJCNN.2011.6033511","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033511","url":null,"abstract":"Neurons tuned to binocular disparity in area V1 are hypothesized to be responsible for short latency binocular vergence movements, which align the two eyes on the same object as it moves in depth. Disparity selective neurons in V1 are not only selective to disparity, but also to other visual stimulus dimensions, in particular orientation. In this work, we explore the role of neurons tuned to different orientations in binocular vergence control. We trained an artificial binocular vision system to execute corrective vergence movements based on the outputs of disparity selective neurons tuned to different orientations and scales. As might be expected, we find that neurons tuned to vertical orientations have the strongest effect on the vergence eye movements. The effect of neurons tuned to other orientations decreases as the tuned orientation approaches horizontal. Although adding neurons tuned to non-vertical orientations does not appear to improve vergence tracking accuracy, we find that neurons tuned to non-vertical orientations still play critical roles in binocular vergence control. First, they decrease the time required to learn the vergence control strategy. Second, they also increase the effective range of vergence control.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"23 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":"131958522","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.2011.6033607
Yasuki Hirai, M. Kikuchi, H. Kurokawa
The regulatory interaction between gene expressions is considered as a universal mechanism in biological systems and such a mechanism of interactions has been modeled as gene regulatory networks. The gene regulatory networks show a correlation among gene expressions. A lot of methods to describe the gene regulatory network have been developed. Especially, owing to the technologies such as DNA microarrays that provide a number of time course data of gene expressions, the gene regulatory network models described by differential equations have been proposed and developed in recently. To infer such a gene regulatory network using differential equations, it is necessary to approximate many unknown functions from the time course data of gene expressions that is obtained experimentally. One of the successful inference methods of the gene regulatory networks is the method using the neural network. In this study, to improve a performance of the inference, we propose the inferring method of the gene regulatory networks using neural networks adopting a kind of majority rule. Simulation results show the validity of the proposed method.
{"title":"Inferring method of the gene regulatory networks using neural networks adopting a majority rule","authors":"Yasuki Hirai, M. Kikuchi, H. Kurokawa","doi":"10.1109/IJCNN.2011.6033607","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033607","url":null,"abstract":"The regulatory interaction between gene expressions is considered as a universal mechanism in biological systems and such a mechanism of interactions has been modeled as gene regulatory networks. The gene regulatory networks show a correlation among gene expressions. A lot of methods to describe the gene regulatory network have been developed. Especially, owing to the technologies such as DNA microarrays that provide a number of time course data of gene expressions, the gene regulatory network models described by differential equations have been proposed and developed in recently. To infer such a gene regulatory network using differential equations, it is necessary to approximate many unknown functions from the time course data of gene expressions that is obtained experimentally. One of the successful inference methods of the gene regulatory networks is the method using the neural network. In this study, to improve a performance of the inference, we propose the inferring method of the gene regulatory networks using neural networks adopting a kind of majority rule. Simulation results show the validity of the proposed method.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"94 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":"134455278","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.2011.6033575
Kenneth L. Rice, T. Taha, K. Iftekharuddin, Keith Anderson, Teddy Salan
At present, a major initiative in the research community is investigating new ways of processing data that capture the efficiency of the human brain in hardware and software. This has resulted in increased interest and development of bio-inspired computing approaches in software and hardware. One such bio-inspired approach is Cellular Simultaneous Recurrent Networks (CSRNs). CSRNs have been demonstrated to be very useful in solving state transition type problems, such as maze traversals. Although powerful in image processing capabilities, CSRNs have high computational demands with increasing input problem size. In this work, we revisit the maze traversal problem to gain an understanding of the general processing of CSRNs. We use a 2.67 GHz Intel Xeon X5550 processor coupled with an NVIDIA Tesla C2050 general purpose graphical processing unit (GPGPU) to create several novel accelerated CSRN implementations as a means of overcoming the high computational cost. Additionally, we explore the use of decoupled extended Kalman filters in the CSRN training phase and find a significant reduction in runtime with negligible change in accuracy. We find in our results that we can achieve average speedups of 21.73 and 3.55 times for the training and testing phases respectively when compared to optimized C implementations. The main bottleneck in training performance was a matrix inversion computation. Therefore, we utilize several methods to reduce the effects of the matrix inversion computation.
目前，研究界的一项主要举措是研究在硬件和软件中捕捉人类大脑效率的数据处理新方法。这导致了对软件和硬件中生物启发计算方法的兴趣和发展的增加。其中一种受生物启发的方法是细胞同步循环网络(CSRNs)。csrn已被证明在解决状态转换类型问题(如迷宫遍历)方面非常有用。尽管CSRNs具有强大的图像处理能力，但随着输入问题规模的增加，其计算需求也越来越高。在这项工作中，我们重新审视迷宫遍历问题，以了解csrn的一般处理。我们使用2.67 GHz Intel Xeon X5550处理器和NVIDIA Tesla C2050通用图形处理单元(GPGPU)来创建几个新的加速CSRN实现，作为克服高计算成本的手段。此外，我们探索了在CSRN训练阶段使用解耦扩展卡尔曼滤波器，并发现运行时间显著减少，精度变化可以忽略不计。我们在结果中发现，与优化的C实现相比，我们可以在训练和测试阶段分别实现21.73倍和3.55倍的平均速度。训练性能的主要瓶颈是矩阵反演计算。因此，我们利用几种方法来减少矩阵反演计算的影响。
{"title":"GPGPU acceleration of Cellular Simultaneous Recurrent Networks adapted for maze traversals","authors":"Kenneth L. Rice, T. Taha, K. Iftekharuddin, Keith Anderson, Teddy Salan","doi":"10.1109/IJCNN.2011.6033575","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033575","url":null,"abstract":"At present, a major initiative in the research community is investigating new ways of processing data that capture the efficiency of the human brain in hardware and software. This has resulted in increased interest and development of bio-inspired computing approaches in software and hardware. One such bio-inspired approach is Cellular Simultaneous Recurrent Networks (CSRNs). CSRNs have been demonstrated to be very useful in solving state transition type problems, such as maze traversals. Although powerful in image processing capabilities, CSRNs have high computational demands with increasing input problem size. In this work, we revisit the maze traversal problem to gain an understanding of the general processing of CSRNs. We use a 2.67 GHz Intel Xeon X5550 processor coupled with an NVIDIA Tesla C2050 general purpose graphical processing unit (GPGPU) to create several novel accelerated CSRN implementations as a means of overcoming the high computational cost. Additionally, we explore the use of decoupled extended Kalman filters in the CSRN training phase and find a significant reduction in runtime with negligible change in accuracy. We find in our results that we can achieve average speedups of 21.73 and 3.55 times for the training and testing phases respectively when compared to optimized C implementations. The main bottleneck in training performance was a matrix inversion computation. Therefore, we utilize several methods to reduce the effects of the matrix inversion computation.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"76 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":"133746384","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.2011.6033244
Chun-Shu Wei, L. Ko, Shang-Wen Chuang, T. Jung, Chin-Teng Lin
Motion sickness is a common symptom that occurs when the brain receives conflicting information about the sensation of movement. Many motion sickness biomarkers have been identified, and electroencephalogram (EEG)-based motion sickness level estimation was found feasible in our previous study. This study employs genetic feature selection to find a subset of EEG features that can further improve estimation performance over the correlation-based method reported in the previous studies. The features selected by genetic feature selection were very different from those obtained by correlation analysis. Results of this study demonstrate that genetic feature selection is a very effective method to optimize the estimation of motion-sickness level. This demonstration could lead to a practical system for noninvasive monitoring of the motion sickness of individuals in real-world environments.
{"title":"Genetic feature selection in EEG-based motion sickness estimation","authors":"Chun-Shu Wei, L. Ko, Shang-Wen Chuang, T. Jung, Chin-Teng Lin","doi":"10.1109/IJCNN.2011.6033244","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033244","url":null,"abstract":"Motion sickness is a common symptom that occurs when the brain receives conflicting information about the sensation of movement. Many motion sickness biomarkers have been identified, and electroencephalogram (EEG)-based motion sickness level estimation was found feasible in our previous study. This study employs genetic feature selection to find a subset of EEG features that can further improve estimation performance over the correlation-based method reported in the previous studies. The features selected by genetic feature selection were very different from those obtained by correlation analysis. Results of this study demonstrate that genetic feature selection is a very effective method to optimize the estimation of motion-sickness level. This demonstration could lead to a practical system for noninvasive monitoring of the motion sickness of individuals in real-world environments.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"30 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":"130588215","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.2011.6033335
A. Lazar, Yiyin Zhou
Video Time Decoding Machines faithfully reconstruct bandlimited stimuli encoded with Video Time Encoding Machines. The key step in recovery calls for the pseudo-inversion of a typically poorly conditioned large scale matrix. We investigate the realization of time decoders employing only neural components. We show that Video Time Decoding Machines can be realized with recurrent neural networks, describe their architecture and evaluate their performance. We provide the first demonstration of recovery of natural and synthetic video scenes encoded in the spike domain with decoders realized with only neural components. The performance in recovery using the latter decoder is not distinguishable from the one based on the pseudo-inversion matrix method.
{"title":"Realizing Video Time Decoding Machines with recurrent neural networks","authors":"A. Lazar, Yiyin Zhou","doi":"10.1109/IJCNN.2011.6033335","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033335","url":null,"abstract":"Video Time Decoding Machines faithfully reconstruct bandlimited stimuli encoded with Video Time Encoding Machines. The key step in recovery calls for the pseudo-inversion of a typically poorly conditioned large scale matrix. We investigate the realization of time decoders employing only neural components. We show that Video Time Decoding Machines can be realized with recurrent neural networks, describe their architecture and evaluate their performance. We provide the first demonstration of recovery of natural and synthetic video scenes encoded in the spike domain with decoders realized with only neural components. The performance in recovery using the latter decoder is not distinguishable from the one based on the pseudo-inversion matrix method.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"599 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":"132456121","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.2011.6033555
Kittipat Kampa, E. Hasanbelliu, J. Príncipe
This paper presents an efficient approach to calculate the difference between two probability density functions (pdfs), each of which is a mixture of Gaussians (MoG). Unlike Kullback-Leibler divergence (DKL), the authors propose that the Cauchy-Schwarz (CS) pdf divergence measure (DCS) can give an analytic, closed-form expression for MoG. This property of the DCS makes fast and efficient calculations possible, which is tremendously desired in real-world applications where the dimensionality of the data/features is very high. We show that DCS follows similar trends to DKL, but can be computed much faster, especially when the dimensionality is high. Moreover, the proposed method is shown to significantly outperform DKL in classifying real-world 2D and 3D objects, and static hand posture recognition based on distances alone.
{"title":"Closed-form cauchy-schwarz PDF divergence for mixture of Gaussians","authors":"Kittipat Kampa, E. Hasanbelliu, J. Príncipe","doi":"10.1109/IJCNN.2011.6033555","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033555","url":null,"abstract":"This paper presents an efficient approach to calculate the difference between two probability density functions (pdfs), each of which is a mixture of Gaussians (MoG). Unlike Kullback-Leibler divergence (DKL), the authors propose that the Cauchy-Schwarz (CS) pdf divergence measure (DCS) can give an analytic, closed-form expression for MoG. This property of the DCS makes fast and efficient calculations possible, which is tremendously desired in real-world applications where the dimensionality of the data/features is very high. We show that DCS follows similar trends to DKL, but can be computed much faster, especially when the dimensionality is high. Moreover, the proposed method is shown to significantly outperform DKL in classifying real-world 2D and 3D objects, and static hand posture recognition based on distances alone.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"40 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":"132579978","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.2011.6033527
Bao Liu, Junhong Wang, Huachao Qu
Inspired by the modulation mechanism of Neuroendocrine-Immune System (NEIs), this paper presents a novel structure of artificial neural network named NEI-NN as well as its evolutionary method. The NEI-NN includes two parts, i.e. positive sub-network (PSN) and negative sub-network (NSN). The increased and decreased secretion functions of hormone are designed as the neuron functions of PSN and NSN, respectively. In order to make the novel neural network learn quickly, we redesign the novel neuron, which is different from those of conventional neural networks. Besides the normal input signals, two control signals are also considered in the proposed solution. One control signal is the enable/disable signal, and the other one is the slope control signal. The former can modify the structure of NEI-NN, and the later can regulate the evolutionary speed of NEI-NN. The NEI-NN can obtain the optimized network structure during the evolutionary process of weights. We chooses a second order with delay model to examine the performance of novel neural network. The experiment results show that the optimized structure and learning speed of NEI-NN are better than the conventional neural network.
{"title":"A novel neural network inspired from Neuroendocrine-Immune System","authors":"Bao Liu, Junhong Wang, Huachao Qu","doi":"10.1109/IJCNN.2011.6033527","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033527","url":null,"abstract":"Inspired by the modulation mechanism of Neuroendocrine-Immune System (NEIs), this paper presents a novel structure of artificial neural network named NEI-NN as well as its evolutionary method. The NEI-NN includes two parts, i.e. positive sub-network (PSN) and negative sub-network (NSN). The increased and decreased secretion functions of hormone are designed as the neuron functions of PSN and NSN, respectively. In order to make the novel neural network learn quickly, we redesign the novel neuron, which is different from those of conventional neural networks. Besides the normal input signals, two control signals are also considered in the proposed solution. One control signal is the enable/disable signal, and the other one is the slope control signal. The former can modify the structure of NEI-NN, and the later can regulate the evolutionary speed of NEI-NN. The NEI-NN can obtain the optimized network structure during the evolutionary process of weights. We chooses a second order with delay model to examine the performance of novel neural network. The experiment results show that the optimized structure and learning speed of NEI-NN are better than the conventional neural network.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"45 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":"116073207","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.2011.6033434
E. Palomo, John North, D. Elizondo, Rafael Marcos Luque Baena, Tim Watson
Digital crimes are a part of modern life but evidence of these crimes can be captured in network traffic data logs. Analysing these logs is a difficult process, this is especially true as the format that different attacks can take can vary tremendously and may be unknown at the time of the analysis. The main objective of the field of network forensics consists of gathering evidence of illegal acts from a networking infrastructure. Therefore, software tools, and techniques, that can help with these digital investigations are in great demand. In this paper, an approach to analysing and visualising network traffic data based upon the use of self-organising maps (SOM) is presented. The self-organising map has been widely used in clustering tasks in the literature; it can enable network clusters to be created and visualised in a manner that makes them immediately more intuitive and understandable and can be performed on high-dimensional input data, transforming this into a much lower dimensional space. In order to show the usefulness of this approach, the self-organising map has been applied to traffic data, for use as a tool in network forensics. Moreover, the proposed SOM takes into account the qualitative features that are present in the traffic data, in addition to the quantitative features. The traffic data was was clustered and visualised and the results were then analysed. The results demonstrate that this technique can be used to aid in the comprehension of digital forensics and to facilitate the search for anomalous behaviour in the network environment.
{"title":"Visualisation of network forensics traffic data with a self-organising map for qualitative features","authors":"E. Palomo, John North, D. Elizondo, Rafael Marcos Luque Baena, Tim Watson","doi":"10.1109/IJCNN.2011.6033434","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033434","url":null,"abstract":"Digital crimes are a part of modern life but evidence of these crimes can be captured in network traffic data logs. Analysing these logs is a difficult process, this is especially true as the format that different attacks can take can vary tremendously and may be unknown at the time of the analysis. The main objective of the field of network forensics consists of gathering evidence of illegal acts from a networking infrastructure. Therefore, software tools, and techniques, that can help with these digital investigations are in great demand. In this paper, an approach to analysing and visualising network traffic data based upon the use of self-organising maps (SOM) is presented. The self-organising map has been widely used in clustering tasks in the literature; it can enable network clusters to be created and visualised in a manner that makes them immediately more intuitive and understandable and can be performed on high-dimensional input data, transforming this into a much lower dimensional space. In order to show the usefulness of this approach, the self-organising map has been applied to traffic data, for use as a tool in network forensics. Moreover, the proposed SOM takes into account the qualitative features that are present in the traffic data, in addition to the quantitative features. The traffic data was was clustered and visualised and the results were then analysed. The results demonstrate that this technique can be used to aid in the comprehension of digital forensics and to facilitate the search for anomalous behaviour in the network environment.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"39 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":"116199033","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.2011.6033377
Weike Pan, J. Kwok
Clustering is an invaluable data analysis tool in a variety of applications. However, existing algorithms often assume that the clusters do not have any structural relationship. Hence, they may not work well in situations where such structural relationships are present (e.g., it may be given that the document clusters are residing in a hierarchy). Recently, the development of the kernel-based structured clustering algorithm CLUHSIC [9] tries to alleviate this problem. But since the input kernel matrix is defined purely based on the feature vectors of the input data, it does not take the output clustering structure into account. Consequently, a direct alignment of the input and output kernel matrices may not assure good performance. In this paper, we reduce this mismatch by learning a better input kernel matrix using techniques from semi-supervised kernel learning. We combine manifold information and output structure information with pairwise clustering constraints that are automatically generated during the clustering process. Experiments on a number of data sets show that the proposed method outperforms existing structured clustering algorithms.
{"title":"Structured clustering with automatic kernel adaptation","authors":"Weike Pan, J. Kwok","doi":"10.1109/IJCNN.2011.6033377","DOIUrl":"https://doi.org/10.1109/IJCNN.2011.6033377","url":null,"abstract":"Clustering is an invaluable data analysis tool in a variety of applications. However, existing algorithms often assume that the clusters do not have any structural relationship. Hence, they may not work well in situations where such structural relationships are present (e.g., it may be given that the document clusters are residing in a hierarchy). Recently, the development of the kernel-based structured clustering algorithm CLUHSIC [9] tries to alleviate this problem. But since the input kernel matrix is defined purely based on the feature vectors of the input data, it does not take the output clustering structure into account. Consequently, a direct alignment of the input and output kernel matrices may not assure good performance. In this paper, we reduce this mismatch by learning a better input kernel matrix using techniques from semi-supervised kernel learning. We combine manifold information and output structure information with pairwise clustering constraints that are automatically generated during the clustering process. Experiments on a number of data sets show that the proposed method outperforms existing structured clustering algorithms.","PeriodicalId":415833,"journal":{"name":"The 2011 International Joint Conference on Neural Networks","volume":"19 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":"125407999","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：