Special Session on

Large Scale Global Optimization

2010 IEEE World Congress on Computational Intelligence (CEC@WCCI-2010)

July 18-23, 2010, Barcelona, Spain

Call for papers
The Companion Competition
Test suite for our companion competition is available here

Reference: K. Tang, Xiaodong Li, P. N. Suganthan, Z. Yang and T. Weise, "Benchmark Functions for the CEC'2010 Special Session and Competition on Large Scale Global Optimization," Technical Report, Nature Inspired Computation and Applications Laboratory, USTC, China, http://nical.ustc.edu.cn/cec10ss.php, 2009.

To those who may participate in the competition: Please could you inform Ke Tang about your participation, so that we can update you about any correction of bugs or extension of the deadline.

Update Notice:

  1. Jan 9th, 2010: Some preliminary results have been provided, please check HERE.
  2. Jan 8th, 2010: The evaluation criteria have been modified, please check the updated Technical Report.


Call for papers
In the past two decades, different kinds of nature-inspired optimization algorithms have been developed and applied to solve optimization problems, including Simulated Annealing (SA), Evolutionary Algorithms (EAs), Differential Evolution (DE), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Estimation of Distribution Algorithms (EDA), etc. Although these approaches have shown excellent search abilities when applying to some small or medium size problems, many of them will encounter severe difficulties when applying to large scale problems, e.g., problems with up to 1000 variables. The reasons appear to be two-fold. First, the complexity of a problem usually increases with the number of decision variables, number of constraints, or even number of objectives (for multi-objective optimization). This emergent complexity might prevent a previously successful search strategy from finding the optimal solution. Second, the solution space of the problem increases exponentially with the number of decision variables, and a more efficient search strategy is required to explore all the promising regions with limited computational resources.

Historically, scaling up EAs to large scale problems has attracted much interest, including both theoretical and practical studies. However, existing work in the areas of EAs are still limited given the significance of the scalability issue. Due to this fact, this special session is devoted to highlight the recent advances in EAs for large scale optimization problems, involving single objective or multiple objectives, unconstrained or constrained problems, binary or discrete or real or mixed decision variables. Specifically, we encourage interested researchers to submit their latest work on:


The Companion Competition
A competition on High-dimensional Numerical Optimization will also be organized in company with our special session. The competition allows participants to run their own algorithms on 20 benchmark functions, each of which is of 1000 dimensions. The purpose of this competition is to compare different algorithm on the exactly same platform. The experiments will take about 205 hours with the Matlab version on a PC with 2.40GHz CPU, and 104 hours with the Java version on a PC with 2.2GHz CPU. Each participant (or research group) is invited to submit a paper to the special session to present their algorithm as well as the results obtained. Details of the set of scalable functions and requirements on the simulation procedure are available at http://nical.ustc.edu.cn/wcci2010/lsgo_benchmark.zip. Researchers are welcome to apply any kind of computational intelligence approaches (e.g. EAs, Neural Nets, fuzzy-based methods) to the test suite. The results of this competition will be archived on our web pages as done for the CEC 2008 competition on High-dimensional Function Optimization.


Important Dates
Paper Submission: January 31, 2010 February 7, 2010
Acceptance Notification: March 15, 2010
Final Manuscript Due: May 2, 2010

For latest news, please refer to http://www.wcci2010.org.


Paper Submission
Manuscripts should be prepared according to the standard format and page limit specified in CEC 2010. For more submission instructions, please see the WCCI submission page at: http://www.wcci2010.org/submission.

All special session papers will be treated in the same way as regular papers. All papers accepted by the special session will be included in the CEC 2010 conference proceedings and selected authors will be invited to present their results during WCCI 2010.

Notice:
When submitting, please make sure you have chosen "S01: Large Scale Global Optimization" as the "Main Research Topic".


Related Events


Special Session Organizers
Ke Tang
Nature Inspired Computation and Applications Laboratory (NICAL)
School of Computer Science and Technology
University of Science and Technology of China, Hefei, Anhui, China
Email: ketang@ustc.edu.cn, Website: http://staff.ustc.edu.cn/~ketang

Xiaodong Li
School of Computer Science and Information Technology
RMIT University, Australia
Email: xiaodong.li@rmit.edu.au, Website: http://goanna.cs.rmit.edu.au/~xiaodong

P. N. Suganthan
School of Electrical and Electronic Engineering
Nanyang Technological University, Singapore
Email: epnsugan@ntu.edu.sg, Website: http://www.ntu.edu.sg/home/epnsugan


Preliminary results for your reference
Algorithms Quality F1 F2 F3 F4 F5 F6 F7 F8 F9 F10
DECC-G Best 1.63e-07 1.25e+03 1.20e+00 7.78e+12 1.50e+08 3.89e+06 4.26e+07 6.37e+06 2.66e+08 1.03e+04
Median 2.86e-07 1.31e+03 1.39e+00 1.51e+13 2.38e+08 4.80e+06 1.07e+08 6.70e+07 3.18e+08 1.07e+04
Worst 4.84e-07 1.40e+03 1.68e+00 2.65e+13 4.12e+08 7.73e+06 6.23e+08 9.22e+07 3.87e+08 1.17e+04
Mean 2.93e-07 1.31e+03 1.39e+00 1.70e+13 2.63e+08 4.96e+06 1.63e+08 6.44e+07 3.21e+08 1.06e+04
Std 8.62e-08 3.26e+01 9.73e-02 5.37e+12 8.44e+07 8.02e+05 1.37e+08 2.89e+07 3.38e+07 2.95e+02
DECC-G* Best 6.33e-12 4.21e+02 2.23e-08 9.76e+11 2.08e+08 5.07e-03 3.45e+06 2.79e+07 1.18e+07 2.33e+03
Median 8.97e-12 4.43e+02 3.30e-08 1.96e+12 2.49e+08 8.85e-03 1.04e+07 4.07e+07 1.41e+07 2.49e+03
Worst 1.31e-11 4.57e+02 4.16e-08 5.39e+12 2.72e+08 1.40e-02 2.28e+07 1.50e+08 1.77e+07 2.64e+03
Mean 8.81e-12 4.42e+02 3.30e-08 2.29e+12 2.45e+08 8.77e-03 1.10e+07 6.14e+07 1.41e+07 2.48e+03
Std 1.49e-12 9.94e+00 5.20e-09 9.97e+11 1.64e+07 2.46e-03 5.44e+06 3.24e+07 1.39e+06 7.63e+01
MLCC Best 0.00e+00 1.73e-11 1.28e-13 4.27e+12 2.15e+08 5.85e+06 4.16e+04 4.51e+04 8.96e+07 2.52e+03
Median 0.00e+00 6.43e-11 1.46e-13 1.03e+13 3.92e+08 1.95e+07 5.15e+05 4.67e+07 1.24e+08 3.16e+03
Worst 3.83e-26 1.09e+01 1.86e-11 1.62e+13 4.87e+08 1.98e+07 2.78e+06 9.06e+07 1.46e+08 5.90e+03
Mean 1.53e-27 5.57e-01 9.88e-13 9.61e+12 3.84e+08 1.62e+07 6.89e+05 4.38e+07 1.23e+08 3.43e+03
Std 7.66e-27 2.21e+00 3.70e-12 3.43e+12 6.93e+07 4.97e+06 7.37e+05 3.45e+07 1.33e+07 8.72e+02
Algorithms Quality F11 F12 F13 F14 F15 F16 F17 F18 F19 F20
DECC-G Best 2.06e+01 7.78e+04 1.78e+03 6.96e+08 1.09e+04 5.97e+01 2.50e+05 5.61e+03 1.02e+06 3.59e+03
Median 2.33e+01 8.87e+04 3.00e+03 8.07e+08 1.18e+04 7.51e+01 2.89e+05 2.30e+04 1.11e+06 3.98e+03
Worst 2.79e+01 1.07e+05 1.66e+04 9.06e+08 1.39e+04 9.24e+01 3.26e+05 4.71e+04 1.20e+06 5.32e+03
Mean 2.34e+01 8.93e+04 5.12e+03 8.08e+08 1.22e+04 7.66e+01 2.87e+05 2.46e+04 1.11e+06 4.06e+03
Std 1.78e+00 6.87e+03 3.95e+03 6.07e+07 8.97e+02 8.14e+00 1.98e+04 1.05e+04 5.15e+04 3.66e+02
DECC-G* Best 5.82e-08 6.16e+01 3.78e+02 2.46e+07 3.62e+03 7.04e-08 8.09e+01 8.37e+02 9.90e+05 2.83e+03
Median 7.52e-08 7.72e+01 5.40e+02 2.90e+07 3.88e+03 1.04e-07 1.03e+02 1.08e+03 1.15e+06 3.21e+03
Worst 8.79e-01 1.19e+02 7.55e+02 3.56e+07 4.25e+03 2.18e+00 1.33e+02 1.53e+03 1.23e+06 6.23e+03
Mean 3.52e-02 7.87e+01 5.50e+02 2.91e+07 3.88e+03 4.01e-01 1.03e+02 1.08e+03 1.14e+06 3.33e+03
Std 1.76e-01 1.41e+01 9.78e+01 2.91e+06 1.76e+02 6.59e-01 1.38e+01 1.61e+02 5.85e+04 6.63e+02
MLCC Best 1.96e+02 2.42e+04 1.01e+03 2.62e+08 5.30e+03 2.08e+02 1.38e+05 2.51e+03 1.21e+06 1.70e+03
Median 1.98e+02 3.47e+04 1.91e+03 3.16e+08 6.89e+03 3.95e+02 1.59e+05 4.17e+03 1.36e+06 2.04e+03
Worst 1.98e+02 4.25e+04 3.47e+03 3.77e+08 1.04e+04 3.97e+02 1.86e+05 1.62e+04 1.54e+06 2.34e+03
Mean 1.98e+02 3.49e+04 2.08e+03 3.16e+08 7.11e+03 3.76e+02 1.59e+05 7.09e+03 1.36e+06 2.05e+03
Std 6.98e-01 4.92e+03 7.27e+02 2.77e+07 1.34e+03 4.71e+01 1.43e+04 4.77e+03 7.35e+04 1.80e+02
  1. DECC-G: The algorithm proposed in:
    • Zhenyu Yang, Ke Tang and X. Yao: "Large Scale Evolutionary Optimization Using Cooperative Coevolution", Information Sciences, 178(15):2985-2999, August 2008.
      Available as a PDF here.
    The parameter group size was set to s=100.
  2. DECC-G*: The same as DECC-G, except that the grouping structure was used as prior knownledge. The parameter group size was set to s=50, and the adaptive weighting strategy of DECC-G was not used.
    Note: DECC-G* is only used for reference purpose, please do not use the grouping structure information to design your own algorithm.
  3. MLCC: The algorithm proposed in:
    • Zhenyu Yang, Ke Tang and Xin Yao: "Multilevel Cooperative Coevolution for Large Scale Optimization", in Proceedings of the 2008 IEEE Congress on Evolutionary Computation (CEC2008), Hongkong, China, 2008, pp. 1663-1670.
      Available as a PDF here
  4. The results were collected when all the 3 million FEs were used up. The results provided here will NOT be included in the final ranking of the competition.