Friday 25 July 2014

IEEE Transactions on Neural Networks and Learning Systems: Volume 25, Issue 8, August 2014

1. GMM-Based Intermediate Matching Kernel for Classification of Varying Length Patterns of Long Duration Speech Using Support Vector Machines
Authors: Aroor Dinesh Dileep; Chellu Chandra Sekhar
Page(s): 1421 - 1432

2. Extensions of Kmeans-Type Algorithms: A New Clustering Framework by Integrating Intracluster Compactness and Intercluster Separation
Authors: Xiaohui Huang; Yunming Ye; Haijun Zhang
Page(s): 1433 - 1446

3. Efficient Kernel Sparse Coding Via First-Order Smooth Optimization
Authors: Minyoung Kim
Page(s): 1447 - 1459

4. Contact-Force Distribution Optimization and Control for Quadruped Robots Using Both Gradient and Adaptive Neural Networks
Authors: Zhijun Li; Shuzhi Sam Ge; Sibang Liu
Page(s): 1460 - 1473

5. On the Capabilities and Computational Costs of Neuron Models
Authors: Michael J. Skocik; Lyle N. Long
Page(s): 1474 - 1483

6. Global Sensitivity Analysis Approach for Input Selection and System Identification Purposes—A New Framework for Feedforward Neural Networks
Authors: Eric Fock
Page(s): 1484 - 1495

7. Cooperative Tracking Control of Nonlinear Multiagent Systems Using Self-Structuring Neural Networks
Authors: Gang Chen; Yong-Duan Song
Page(s): 1496 - 1507

8. Distributed Neural Network Control for Adaptive Synchronization of Uncertain Dynamical Multiagent Systems
Authors: Zhouhua Peng; Dan Wang; Hongwei Zhang; Gang Sun
Page(s): 1508 - 1519

9. Instance-Level Constraint-Based Semisupervised Learning With Imposed Space-Partitioning
Authors: Jayaram Raghuram; David J. Miller; George Kesidis
Page(s): 1520 - 1537

10. Modified Principal Component Analysis: An Integration of Multiple Similarity Subspace Models
Authors: Zizhu Fan; Yong Xu; Wangmeng Zuo; Jian Yang; Jinhui Tang; Zhihui Lai; David Zhang
Page(s): 1538 - 1552

11. On the Complexity of Neural Network Classifiers: A Comparison Between Shallow and Deep Architectures
Authors: Monica Bianchini; Franco Scarselli
Page(s): 1553 - 1565

12. A Minimum Resource Neural Network Framework for Solving Multiconstraint Shortest Path Problems
Authors: Junying Zhang; Xiaoxue Zhao; Xiaotao He
Page(s): 1566 - 1582

13. Simulating Dynamic Plastic Continuous Neural Networks by Finite Elements
Authors: Abdolreza Joghataie; Omid Oliyan Torghabehi
Page(s): 1583 - 1587

14. Minimizing Nearest Neighbor Classification Error for Nonparametric Dimension Reduction
Authors: Wei Bian; Tianyi Zhou; Aleix M. Martinez; George Baciu; Dacheng Tao
Page(s): 1588 - 1594

15. Correction to “Convergence and Rate Analysis of Neural Networks for Sparse Approximation”
Authors: Aurele Balavoine; Justin Romberg; Christopher J. Rozell
Page(s): 1595 - 1596

Monday 21 July 2014

IEEE TNNLS Special issue on “Neurodynamic Systems for Optimization and Applications”

Recurrent neural networks, as dynamical systems, are usually used as models for solving computationally intensive problems. Because of their inherent nature of parallel and distributed information processing, recurrent neural networks are promising computational models for real-time applications. Constrained optimization problems arise in a wide variety of scientific and engineering applications, including signal and image processing, system identification, robot control, process control, pattern recognition, etc. Since the Hopfield neural network was introduced for solving optimization problems, significant progress has been made in theory, algorithms and applications. A number of neurodynamic models have been proposed for solving different problems ranging from discrete optimization to continuous optimization, linear programming to nonlinear optimization, convex optimization to non-convex optimization, smooth optimization to non-smooth optimization, numerical software to analog hardware implementations, etc. Some of them have been successfully applied to robot control, process control, signal and image processing, pattern recognition and classification, economic prediction and so on. In addition, as a kind of neuromorphic systems, they are potentially useful for simulating the brain functions, which is an important topic in neuroscience.

The objective of this special issue is to bring together recent advances in the field of neurodynamic systems for solving optimization problems. We invite original and unpublished research contributions in all relevant areas. We will encourage submissions of papers with new models and applications which would further promote research activities in this area.

Topics of interest include, but are not limited to:
  • Neurodynamic models for constrained optimization
  • Neurodynamic models for multi-objective optimization
  • Neurodynamic models for large-scale optimization problems
  • Neurodynamic models for deep learning
  • Neurodynamic models for optimal control
  • Neurodynamic models for tensor decomposition
  • Analysis of neurodynamic optimization systems
  • Neurodynamic optimization in the brain
  • Neurodynamic optimization for process control
  • Neurodynamic optimization for robot control
  • Neurodynamic optimization for biomedical engineering problems
  • Neurodynamic optimization for signal processing
  • Neurodynamic optimization for image processing
  • Neurodynamic optimization for support vector machine learning
  • Neurodynamic optimization for pattern recognition
  • Neurodynamic optimization for other applications

IMPORTANT DATES

Aug. 15, 2014 – Deadline for manuscript submission
Dec. 31, 2014 – Notification to authors
Feb. 15, 2015 – Deadline for submission of revised manuscripts
Mar.1, 2015 – Final decision
May/June 2015 – Special issue publication in the IEEE TNNLS.

SUBMISSION INSTRUCTIONS

  1. Read the information for authors at http://cis.ieee.org/tnnls
  2. Submit the manuscript by August 15, 2014 at the IEEE-TNNLS webpage http://mc.manuscriptcentral.com/tnnls and follow the submission procedure. Please indicate clearly on the first page of the manuscript and the Author’s Cover Letter that the manuscript has been submitted to the Special Issue on Neurodynamic Systems for Optimization and Applications. Send also an e-mail to chenglong@compsys.ia.ac.cn with subject “TNNLS special issue submission” to notify the editors of your submission.

GUEST EDITORS

Zhigang Zeng
Huazhong University of Science and Technology, China
zgzeng@hust.edu.cn
http://auto.hust.edu.cn/zhigangzeng/

Andrzej Cichocki
Brain Science Institute, RIKEN, Japan
cia@braiin.riken.jp
http://www.bsp.brain.riken.jp/~cia/

Long Cheng
Institute of Automation, Chinese Academy of Sciences, China
long.cheng@ia.ac.cn
http://compsys.ia.ac.cn/~chenglong

Yousheng Xia
Fuzhou University, China
ysxia@fzu.edu.cn
http://cmcs.fzu.edu.cn/action-model-name-teacher-itemid-34

Xiaolin Hu
Tsinghua University, China
xlhu@tsinghua.edu.cn
www.xlhu.cn

Thursday 10 July 2014

WCCI 2014 Day 4

The social media subcommittee has been active again on day four of WCCI 2014, tweeting sessions and making notes for today's post.

The first event was a panel session on Big Data and Computational Intelligence, chaired by Jerry Mendel. Jerry gave an overview of big data, and called for innovative approaches to solve big data problems.

Jose Lazano made the point that big data problems are the same as we have been solving in computational intelligence for years, but that the approaches have to be different. He described the characteristics of big data as the three Vs: Volume, as in the scale of the data; Velocity, the speed the data arrives; and Variety, the wide scope of what the data represents.

Nitesh Chawla added a fourth V, Veracity. How much confidence do we have in the data and its value? He also noted that while companies like Facebook have no problems getting big data sets, it is difficult for academics. Tim Havens echoed this, adding that there is a need for good benchmark data sets for big data. He also pointed out that there are always trade-offs how you validate algorithms for big data.

Xiaodong Li gave a brief overview of the computational intelligence techniques for big data. He especially listed deep learning, parallelized machines and robustness techniques for dealing with volume, and online learning methods for dealing with velocity. He also gave an excellent definition of big data: if you can fit it into memory, it's not big data.

The last speaker was Yaochu Jin, who pointed out that due to its volume and variety, biological data like microarray data and gene regulatory networks is also big data.

Janusz Kacprzyk gave an invited lecture on 'Fuzzy dynamic programming: a step towards cognitive dynamic programming'. Janusz presented fuzzy dynamic systems modelling of government regional planning over multiple years for improving cognitive perceptions of socio-economic problems and quality of life. Janusz's passion for this shone through as he stated clearly that this is a real fuzzy model, for a real and important problem, for real end users, and for real money. The model contained fuzzy goals and fuzzy constraints that are objective, such as government limits, but also domain knowledge from experts that are subjective, such as seven life quality indicators.

Huaguang Zhang gave an invited lecture on 'Fuzzy Real-time Leakage Supervisory System for Fluid Transportation Pipeline Networks: New Methods and Applications'. Huaguang research focused on identifying weak leakage in long-distance petroleum pipelines the transient flow produces a drop in pressure at the leakage point of 1%, which is a challenging task. The importance of identifying weak leakages was demonstrated with recent loss of life and economic loses estimated to be 4.4 billion yuan RMB. The first stage of Huanguang's system filters noise signals but not the leakage signals. The second stage validated each characteristic of a chaotic system with statistical analysis. The third stage modelled sections of pipe and raised alarms when differences in sections over time met a threshold. The fourth stage modelled the operating model with the generalised fuzzy hyperbolic tangent model.

Wednesday 9 July 2014

WCCI 2014 Day 3

The first event of today was a plenary panel session organised by Nikhil Pal: Is "Publish or Perish" causing "Death" of science?

Nikhil described the academic system as being like a pyramid, with a very few at the very top and many at the bottom. With so many (too many, in his opinion) at the bottom, there is a tremendous pressure to publish more, thanks in part to an emphasis on metrics like impact factor and h-index. This has resulted in a proliferation of conferences and journals, and an increase in publishing misconduct like plagiarism.

Hisao Ishibuchi gave a thoughtful analysis on the implications of quantity vs quality. His analysis looked at the impact each dimension has on an academic's evaluation by administrators.

Chin-Tong Lin spoke about the victims of the pressure caused by Publish or Perish. The main victims are Associate Editors, and Editors in Chief, who all have their workloads increased by an increase in submissions. He demonstrated how, for the journal IEEE Transactions on Fuzzy Systems, the increase over time of impact factor for the journals led to an increase in submissions. Other victims were the readers and authors of journals: the readers who are subjected to more low-quality papers and authors who must produce the articles.

Derong Liu told the audience that the pressure affected quality "moderately" and increased plagiarism "sort of". He didn't think that the "pay to publish" approach would improve publication quality, and that metrics do influence publications.

Simon Lucas offered his opinion that while the average quality of papers may be going down, the average quality of papers being read is going up. Authors can get sucked into publishing too  much quantity, which negatively affects quality and encourages plagiarism. The "pay to publish" approach will give an advantage of wealthy authors but can be made to work. He said that impact factor must be taken seriously, as it does inflluence publications. In his opinion, the publishing system isn't broken but does need to include more social media in the publications process, especially discussions on articles.

Marios Polycarpou spoke next and told the audience that researchers now spend more time reporting research than actually doing it. This has caused the signal to noise ratio to go down due to the pressure to publish. This pressure comes from too much emphasis on formulas such as h-index, impact factor and so on. Technology makes it so easy to detect plagiarism that it just isn't worth the risk, since one case of plagiarism can ruin a researcher's career.

Jun Wang presented data that showed that the emphasis on pubishing has promoted fraudulent research, and that impact factors are causing coercive citations, that is, citations that authors are coerced into including to boost the journal impact factor. Impact factors also have a strong correlation with article retractions.

Xin Yao was the last speaker and was to the point: trying to quantify something that is not quantifiable is a symptom of lazy management. The pressure to publish has a tremendous impact of younger researchers, which leads them to take risks with plagiarism and academic fraud.

Yann LeCun, Director of AI at Facebook, gave a plenary on "Deep Learning and the Representation of Natural Data". The third renewal of neural nets has focused mostly on audio and video data. An overview of deep learning explained the case for deep learning over SVM approaches, which are effectively lookup/template methods. Various example applications were discussed, such as face detection and body pose recognition. The most exciting part was the array of demos! A video was shown of a convolutional network taking images from a Kinect and constructing a 3D model of a hand to idenitfy gestures. Yann plugged in a standard webcam and feed the images into a trained convolutional network, and the webcam recognised a variety of objects in the hall, such as Yann's laptop, keyboard, iPod (well, iPhone, but close enough!) sunglasses, coffee mugs etc. Another webcam demo with a convolutional network demonstrated how it can learn a variety of objects by pointing the webcam at an object and clicking 'learn'. The final video showed Yann's students using a pair of stereo vision cameras with a convolutional network to identify traversible and non-traverisble terrain for a robot to navigate.
 

Tuesday 8 July 2014

WCCI 2014 Day 2 Summary

Several members (Min Jiang, Stephen Matthews and Mike Watts) of the Social Media Subcommittee are attending the WCCI 2014 conference in Beijing. We have been Tweeting the plenaries and invited lectures that we are able to attend. Below is a summary of the highlights of Day 2 of WCCI 2014.

The first event was the plenary panel session "CI-related Research Funding", which was chaired by Marios Polycarpou. The first speaker was Marimuthu Palaniswami, who spoke about Australian research networks and their funding structure. An important point he raised was that Australian funding requires reporting of outcomes, such as what the research achieved, and how it raised the national capacity to do research. He also spoke about his seven tips for having a happy career in computational intelligence:

  1. Direction
  2. Be simple and objective
  3. Hunt as a pack
  4. Leverage your resources
  5. Maintain your resistance
  6. Demonstrate your impact
  7. Enjoy the experience


The next speaker was Jose Principe, who talked about his experiences over the last 25 years writing grant proposals in the USA. The messages from his talk were firstly that the rate at which US proposals are funded is so low (15%) that researchers must spend too much of their time writing proposals; secondly, computational intelligence is on the verge of making huge contributions to the world, but it is under-funded.

Paul Werbos of the US NSF followed Dr Principe, and explained that the reason so few neural network related proposals are funded is that their aren't enough proposals on neural networks. His tips for increasing your chances for getting funding from the NSF are:

  •     Include AIS - Adaptive and Intelligent Systems - in the title of your proposal
  • Target the grand challenges in the field: cognitive optimization, prediction and big applications
  • Hot topic proposals do not do well at NSF, because little is lost by not funding them
  • Remember the three questions that will be asked when evaulating your proposal: What exactly will you do? How will you do it? Why is it important?


Xin Yao offered a brief response to one of the chair's first questions: what are the hot topics in CI? Xin Yao's answer: whatever you are working on at the time.

Chenghong Wang also spoke about the rate of funding of research proposals in CI in China.

Kalyan Deb gave an invited lecture about "Evolutionary Multi-Objective Optimization (EMO): Two Eventful Decades and Beyond". A great introduction to EMO was provided with easy-to-understand examples, such as the objectives involved with buying a car, i.e., bueno, bonito, barato (good, nice, cheap). A brief overview of EMO history was presented that included David Goldberg's, 1989, 10-line EMO suggestion and early EMO implementations to Deb's recent incarnation -- NSGA-III. As well as the test benchmark problems, spacecraft trajectory optimisation with University of Illinois at Urbana-Champaign and NASA-JPS and a mine scheduling application in Australia were mentioned. Research problems for the future were discussed.

Donald Wunsch gave an invited lecture about "Innovations and Open Problems in Supervised, Unsupervised and Reinforcement Learning". Donald was very enthusiastic and excited for students because they have the opportunity to embark on exciting challenges in the field. One such challenge is unified learning modalities for automatically combining supervised, reinforcement, and supervised learning without human intervention. Donald was very clear that he did not support the popular view that computers will be more intelligent than humans. In fact, Donald foresees this not even happening during our grandchildren's' lifetime - as he put it "If you expect to see robots as smart as humans in your lifetime, you're at the wrong conference, you should be at a conference on life extension".

There will be more tweets tomorrow of the panel plenary "Is "Publish or Perish" causing "Death" of science" and of several invited lectures. We'll also wrap up with another summary blog post.