Artificial Neural Networks - ICANN 2010: 20th International Conference, Thessaloniki, Greece, Septmeber 15-18, 2010, Proceedings, Part II

This three volume set LNCS 6352, LNCS 6353, and LNCS 6354 constitutes the refereed proceedings of the 20th International Conference on Artificial Neural Networks, ICANN 2010, held in Thessaloniki, Greece, in September 2010. The 102 revised full papers, 68 short papers and 29 posters presented were carefully reviewed and selected from 241 submissions. The second volume is divided in topical sections on Kernel algorithms – support vector machines, knowledge engineering and decision making, recurrent ANN, reinforcement learning, robotics, self organizing ANN, adaptive algorithms – systems, and optimization. Lecture Notes in Computer Science 6353......Page 1 Artificial Neural Networks – ICANN 2010: 20th International Conference / Thessaloniki, Greece, September 15-18, 2010 / Proceedings, Part II......Page 2 Preface......Page 4 Organization......Page 7 Table of Contents – Part II......Page 9 Introduction......Page 15 L2 Support Vector Regressors in the Dual Form......Page 16 Training Methods......Page 18 Convergence Improvement......Page 19 Performance Comparison......Page 20 Conclusions......Page 23 References......Page 24 Introduction......Page 25 Reproducing Kernel Hilbert Spaces......Page 27 Wirtinger's Calculus in Complex RKHS......Page 28 Complex Kernel LMS......Page 30 Experiments......Page 32 References......Page 33 Introduction......Page 35 Material and Methods......Page 36 ε-SV Regression (ε-SVR)......Page 37 Fuzzy Weighted SVR with Fuzzy Partition......Page 38 Application......Page 40 Results for the Data Group with Output in the Interval......Page 41 Conclusions......Page 42 References......Page 43 Introduction......Page 44 First and Second Order SMO......Page 45 Better Directions for Second Order SMO......Page 47 Numerical Experiments......Page 48 Discussion and Conclusions......Page 52 References......Page 53 Introduction......Page 54 Almost Random Projections with Margin Maximization......Page 56 Illustrative Examples......Page 59 Conclusions......Page 61 References......Page 62 Introduction......Page 63 Self-Organizing Maps for Structured Data......Page 64 The Activation Mask Kernel......Page 66 Adding Route Information to the Activation Mask Kernel......Page 67 Data Description and Experimental Setting......Page 69 Results and Discussion......Page 70 Conclusions......Page 71 References......Page 72 Introduction......Page 73 Tensor Product of Euclidean Spaces and Matrices......Page 74 Extension to Hilbert Spaces and Operators......Page 75 2-Way Operatorial Representation......Page 76 A General Class of Supervised Problems......Page 78 The Case of Scalar Outputs......Page 80 Experimental Results......Page 81 References......Page 82 Introduction......Page 84 Tri-Class Support Vector Machines......Page 85 Co-Training for Facial Expressions Annotation......Page 86 Setup......Page 87 Results and Discussion......Page 88 References......Page 89 Introduction......Page 90 Learning Prototypes (LPs)......Page 91 Learning SVs (LSVs)......Page 92 Experiments......Page 93 References......Page 95 Introduction......Page 96 Algorithms for Solving NPP and SVM......Page 97 Convergence of GSK and MDM......Page 98 Convergence of SMO......Page 100 References......Page 101 Introduction......Page 102 Feature Selection by Zero Norm Minimisation......Page 103 Support Feature Machine......Page 104 Experiments......Page 105 Implementation Issues......Page 106 References......Page 107 Introduction......Page 108 The Maze Structure......Page 109 Optimal Agent and Assumptions on Human Model......Page 110 Action Selection Probability......Page 111 Transition Probability......Page 112 Estimation for Belief State......Page 113 Result of Parameter Estimation......Page 114 Result of Belief State Estimation......Page 115 References......Page 116 Introduction......Page 118 Preliminaries: Temporal Learning and Reasoning......Page 119 Representing the Model......Page 120 Learning and Evolving the Model......Page 121 Extracting Knowledge about the Model......Page 122 Validation and Experiments: Case Study......Page 123 Integrating Knowledge Sources......Page 124 Case Study Discussion......Page 125 References......Page 127 Introduction......Page 128 Go and Capture Game......Page 129 Multi-Dimensional Recurrent Neural Networks......Page 130 Evolution Strategies......Page 131 Policy Gradients with Parameter-Based Exploration......Page 132 Network Topology......Page 133 Results......Page 134 Future Work......Page 135 References......Page 136 Introduction......Page 138 Motion Fields Described by Affine Models......Page 139 Extraction of Motion Layers with a Recurrent Neural Network......Page 140 Combined Segregation and Affine Model Estimation......Page 143 Results......Page 144 References......Page 146 Solution of the Weighted Regression......Page 147 Problem Formulation......Page 148 Illustrative Example – Tunnel Furnace......Page 149 References......Page 151 Introduction......Page 152 Formulation of Statistical Reconstruction Problem......Page 153 Experimental Results......Page 154 References......Page 155 Introduction......Page 156 Knowledge Structure Based on Statistical Language Analysis......Page 157 The Metaphor Evaluation Process......Page 158 Result of the Simulation......Page 159 References......Page 161 Introduction......Page 162 Attractor-Based Computation with Reservoir Networks......Page 163 Key Ingredients of Reservoir Computing......Page 164 On the Distribution of Attractor States......Page 166 References......Page 167 Introduction......Page 168 Feature Extraction for Action Representation......Page 169 Action Classification Using LSTM-RNN......Page 170 Experimental Results......Page 171 References......Page 173 Introduction......Page 174 The Model......Page 176 The Reward Signal......Page 177 Simulation and Results......Page 178 Discussion......Page 181 References......Page 182 Introduction......Page 184 Incremental Probabilistic Neural Network......Page 185 Learning Algorithm......Page 186 Reinforcement Learning......Page 188 Estimating the Outputs of a Complex Plant......Page 189 A Reinforcement Learning Task......Page 190 Predicting the Motor Actions in a Robotic Task......Page 191 References......Page 192 Introduction......Page 194 Model Overview......Page 196 Implementation Details......Page 197 Experiment I......Page 200 Experiment II......Page 201 References......Page 202 Introduction......Page 204 State of the Art......Page 205 Mathematical Foundations......Page 206 Our Algorithm......Page 208 Experiments......Page 211 References......Page 213 Reinforcement Learning and OCF......Page 214 Application......Page 215 References......Page 217 One-Shot Supervised Reinforcement Learning for Multi-targeted Tasks: RL-SAS......Page 218 Decomposition of Expected Rewards by Targets......Page 220 Simulation Results......Page 221 References......Page 223 Neuroanatomy of Birdsong......Page 224 Model Description......Page 225 The Motor Pathway Model......Page 226 The Vocal Filter Model......Page 227 Results......Page 228 References......Page 229 A Model to Demonstrate the Role of BG in Navigation......Page 230 An Integrated Model for Navigation......Page 233 References......Page 235 Introduction......Page 236 Application Domain......Page 237 Algorithm......Page 238 Experimental Evaluation of the Approach......Page 240 References......Page 241 Introduction......Page 242 Implementation of the Model......Page 243 The Simulation Results......Page 245 References......Page 246 Introduction......Page 248 Extracting the ‘Shape’ of a Visually Observed End Effector Movements (of Self and Others)......Page 250 Virtual Trajectory Synthesis and Learning to Shape......Page 252 Motor Command Synthesis: Coupled Interactions between the Virtual Trajectory and Internal Body Model......Page 255 References......Page 257 Introduction......Page 259 The Extended Kuramoto Model......Page 260 Methods......Page 261 Experiment 1......Page 263 Experiment 2......Page 266 Conclusion and Future Work......Page 267 References......Page 268 Introduction......Page 270 Brain Model......Page 272 The Computational Model......Page 273 Results......Page 276 Analysis......Page 277 References......Page 278 Introduction......Page 280 Properties of the Robot Model......Page 281 Artificial Neural Network Robot Model......Page 282 Application of Artificial Neural Network Model in Sliding Mode Control of Robot......Page 283 Experiments and Results......Page 285 References......Page 288 Introduction......Page 290 Quadruped Gaits......Page 291 Basic CPG Model......Page 292 Digital Hardware Implementation......Page 293 Module of Van Der Pol Oscillator......Page 294 Quadruped Gait Network Architecture......Page 295 Implementation Results......Page 296 Conclusions and Future Work......Page 298 References......Page 299 Introduction......Page 300 Evolutionary Strategies......Page 301 Trajectory Tracking Problem......Page 303 Kinematic Model of a WMR......Page 304 Trajectory-Tracking......Page 305 Sliding-Mode Trajectory-Tracking Control......Page 306 Experimental Results......Page 307 Conclusion......Page 308 References......Page 309 Introduction......Page 310 The Cognitive Model......Page 311 The Basic Memory and Encoding Perceptions......Page 312 Place Cells......Page 313 Building a Cognitive Topological Map of Environment......Page 315 Simulations and Results......Page 316 References......Page 319 Introduction......Page 321 The Hybrid Control Structure......Page 322 Leader-Following Formation Models......Page 324 Sliding-Mode Controller Design......Page 326 Simulation Results......Page 327 Conclusions......Page 329 References......Page 330 Introduction......Page 331 Motivated Sensorimotor Navigation......Page 332 Learning a Reinforcement Signal via Stimulation of a Non-specific Sensor......Page 334 Robotic Experiments: Learning Interactively to Reach a Goal When the Robot Is Lost......Page 337 Conclusions and Perspectives......Page 338 References......Page 339 Introduction......Page 341 The Algorithm for Maze Navigation and Topological MapCreation......Page 343 References......Page 346 Introduction......Page 347 First Stage: Minimum Information Learning......Page 348 Second Stage: Maximum Information Learning......Page 350 Third Stage: Maximum Information Relearning......Page 351 Results and Discussion......Page 352 References......Page 355 Introduction......Page 357 Modeling of Dynamics Using SOM on a Parameter Space......Page 358 Selection of a Parametric Model......Page 359 Visualization of Changes in Dynamic Behaviour......Page 360 Tank Level Control Dynamics......Page 361 Isolation of Chatter Effect in Vibration Data of a Rolling Mill......Page 363 References......Page 365 Introduction......Page 367 Structured Flows on Manifolds (SFMs)......Page 369 Functional Architectures......Page 370 Sequential Dynamics......Page 371 Implementation of Cursive Handwriting......Page 372 Discussion......Page 373 References......Page 374 Introduction......Page 376 Experiment Description......Page 378 Saccadic Onset Detection and ST-Data Collection......Page 379 Grouping STs with Neural-Gas acting on EOG-Velocity Patterns......Page 380 Between-Groups Comparison of Brain Dynamics......Page 381 Results......Page 383 References......Page 384 Introduction......Page 386 GNNs and PM–GraphSOMs......Page 387 The General Framework......Page 388 GNN and PM–GraphSOM Peculiarities......Page 389 A Layered Architecture for Web Spam Detection......Page 390 Experimental Results......Page 392 References......Page 394 Introduction......Page 396 Self-Organizing Maps (SOMs) and Contribution......Page 397 Comparison of the Labelling Versions......Page 400 Selection of the Number of Data Samples......Page 401 Selection of the Training Algorithm and its Parameters......Page 402 References......Page 404 Introduction......Page 406 Description of the Technique......Page 407 Definition of Maps of Dynamics for Time-Response Analysis......Page 409 Experiments and Results......Page 410 References......Page 414 Introduction......Page 416 Snap-Drift Algorithm......Page 417 SDSOM......Page 418 Results......Page 419 Conclusion......Page 422 References......Page 423 Introduction......Page 424 SOM Approach for Monitoring Fault Evolution......Page 425 Results and Discussion......Page 426 References......Page 427 Introduction......Page 428 System Description......Page 429 Linear Analysis......Page 430 Experiments......Page 431 Conclusion and Discussion......Page 432 References......Page 433 Introduction......Page 434 STRAGEN Algorithm......Page 435 Experiments......Page 436 Noise......Page 437 Three Gaits one Data Base......Page 438 References......Page 439 Introduction......Page 440 Self-Organising Map and Visualisations......Page 441 Minimum Spanning Tree Visualisation......Page 442 Experimental Evaluation......Page 443 References......Page 445 Introduction......Page 446 World Knowledge Representation......Page 447 Sentence Comprehension......Page 449 Experiments......Page 450 References......Page 451 Introduction......Page 452 ACD Approach......Page 453 Echo State Networks......Page 454 PHB Production Process......Page 455 Results and Discussion......Page 456 Conclusions......Page 459 References......Page 460 Introduction......Page 462 Background......Page 463 Description of the Data......Page 464 Representation of the Data......Page 465 Visualization Using Principal Component Analysis (PCA)......Page 466 Classifier Used......Page 467 Results for Channel 1......Page 468 Results for Channel 2......Page 469 References......Page 470 Introduction......Page 472 Detecting Changes Using the ICI rule......Page 473 Change-Detection Refinement Procedure......Page 474 ICI-Based Adaptive Classifier......Page 476 Experiments......Page 478 References......Page 480 Introduction......Page 482 Related Work......Page 483 Adaptive Metric Learning for Initialization Graph......Page 484 Graph Based Semi-Supervised Classification......Page 485 Experiments......Page 487 Experimental Results......Page 488 Discussion of lp Estimation......Page 490 Conclusion......Page 491 References......Page 492 Mathematical Model of the Longitudinal Dynamics......Page 493 Speed vs. Braking Distance......Page 494 Control System......Page 495 Genetic Optimization of the Controller......Page 496 Results......Page 498 References......Page 499 Introduction......Page 500 Local Fusion with Neural Networks......Page 501 Experimental Results......Page 503 Conclusions......Page 504 References......Page 505 Introduction......Page 506 System Architecture......Page 507 Selection of Relevant Visual and Auditory Segments......Page 508 Results......Page 509 References......Page 511 Introduction......Page 512 Problem Formulation and Model Description......Page 513 Global Convergence......Page 514 Simulation Results......Page 516 References......Page 518 Introduction......Page 520 HTM Formalism......Page 521 Proposed Extension for the HTM Formalism......Page 524 Example Application: Sign Language Recognition......Page 525 Discussion......Page 529 References......Page 531 Introduction......Page 533 Mutual Information......Page 534 Particle Swarm Optimization for ICA......Page 535 Results and Discussion......Page 536 References......Page 538 Introduction......Page 539 Surface Transformation......Page 540 Experimental Results and Discussion......Page 543 References......Page 544 Artificial Immune Networks......Page 545 Pareto Dominance......Page 546 The Pareto Cloud......Page 547 Evolution of the Network......Page 548 Experiments......Page 549 References......Page 550 Author Index......Page 551 th This volume is part of the three-volume proceedings of the 20 International Conference on Arti?cial Neural Networks (ICANN 2010) that was held in Th- saloniki, Greece during September 15–18, 2010. ICANN is an annual meeting sponsored by the European Neural Network Society (ENNS) in cooperation with the International Neural Network So- ety (INNS) and the Japanese Neural Network Society (JNNS). This series of conferences has been held annually since 1991 in Europe, covering the?eld of neurocomputing, learning systems and other related areas. As in the past 19 events, ICANN 2010 provided a distinguished, lively and interdisciplinary discussion forum for researches and scientists from around the globe. Ito?eredagoodchanceto discussthe latestadvancesofresearchandalso all the developments and applications in the area of Arti?cial Neural Networks (ANNs). ANNs provide an information processing structure inspired by biolo- cal nervous systems and they consist of a large number of highly interconnected processing elements (neurons). Each neuron is a simple processor with a limited computing capacity typically restricted to a rule for combining input signals (utilizing an activation function) in order to calculate the output one. Output signalsmaybesenttootherunitsalongconnectionsknownasweightsthatexcite or inhibit the signal being communicated. ANNs have the ability “to learn” by example (a large volume of cases) through several iterations without requiring a priori?xed knowledge of the relationships between process parameters.