Interview with Prof. Ponnuthurai Suganthan of Nanyang Technological University in Singapore, chairman of the 2017 IEEE Swarm Intelligence Symposium (Interviewed by Dr. David Fogel, co-chair, 2017 IEEE Symposium Series on Computational Intelligence.)

Prof. Suganthan has been researching areas for computational intelligence since at least 1993, focusing first on Hopfield neural networks, then moving to fuzzy systems, self-organizing maps, genetic algorithms, swarm intelligence, multiobjective optimization, differential evolution, evolutionary programming, and, more recently, classification, and forecasting problems. Prof. Suganthan serves as chairman of the IEEE 2017 Swarm Intelligence Symposium, to be held as part of the 2017 IEEE Symposium Series on Computational Intelligence, held in Honolulu, HI, Nov. 27-Dec. 1. I asked Prof. Suganthan to provide insight into this area.

DF: You are the chairman of the 2017 IEEE Swarm Intelligence Symposium. Let’s start by asking “What is swarm intelligence and how does it differ from other types of ‘intelligence’?”

PS: Intelligence is perhaps difficult to define, or difficult for people to agree on a definition in any case. But swarm intelligence was motivated by observing the behavior of collectives of organisms in nature. How do they move together? How do they communicate to achieve an end that no individual of the collective could accomplish on its own? So you can think of flocking birds, schooling fish, ants finding food, honey bees finding food. These swarming behaviors have evolved over millions of years and have survived natural selection. A natural question is: “Would it be possible to emulate these time-tested and successful natural swarm intelligence strategies to solve challenging real-world problems?” So figuring out the potential uses for these models and better understanding their theoretical properties is what the field of swarm intelligence is all about as far as I’m concerned.

DF: What are some examples of swarm intelligence algorithms?

PS: The prominent examples are ant colony optimization, particle swarm optimization, and honey bee colony optimization. This link describes additional swarm intelligence methods:

DF: What examples are there are of successes with these algorithms?

PS: There are literally thousands. So in choosing some here I don’t mean to overlook others. I suppose it’s easiest to focus on work that I’m most familiar with. Power systems have benefited substantially from the applications of swarm intelligence methods. Communication engineering has benefited from ant colony optimization and particle swarm optimization for solving routing problems, as well as problems such as antenna design. One example is the work of Ribeiro et al. (2016) that used particle swarm methods to improve continuous descent sequences in aircraft, validating on real data from Brasilia International Airport. In addition, biomedical as well as image analysis researchers have applied swarm algorithms in real-world problems. I recommend that anyone interested in pursuing more applications have a look at the main journals and IEEE conferences and search for “swarm” and then follow up directly with the authors of papers that appear to be in their area of interest.

DF: We hear a lot about deep learning in the news. Does swarm intelligence contribute to deep learning?

PS: Not yet I think. Deep learning algorithms have been focused on back propagation. This means that the learning algorithms in deep learning nets are using gradient information rather than searching a space stochastically. Doing it stochastically can be done. There are many examples of using evolutionary or swarm algorithms to optimize neural networks. So, we may see a synergy between swarm methods and deep learning in the future. The challenge here is the number of connection weights running into several million. Hence, the conventional large population-based approaches may not be effective. Micro-populatiProf.ons could be considered first.

DF: What do you think some of the most important contributions of swarm intelligence have been over the years?

PS: Always the pioneering works come to my mind. These are ant colony optimization (also known as ACO) offered originally by Marco Dorigo, and particle swarm optimization (also known as PSO) offered by Russ Eberhart and James Kennedy. ACO has demonstrated excellent performance in solving combinatorial problems while PSO has impressed researchers in the continuous numerical optimization domain. Bee colony methods have also subsequently become popular among researchers. In recent years, numerous nature-inspired methods have been proposed. Although these methods have been inspired by different natural phenomena, the equation forms are strikingly similar. Hence, in the future, it is desirable for researchers to pay attention to similarities at the equation level and performance level. And I should also mention the original work of Ken Price and Rainer Storn on differential evolution which has been dominating the IEEE Congress on Evolutionary Computation benchmarking competitions since 2005.

DF: What are you working on in this area?

PS: Instead of proposing yet another swarm-inspired algorithm with strong similarities at the equation level to existing algorithms, I think it’s beneficial to strive to improve performance on specific problems. Hence, we are looking at utilizing things like adaptation, ensemble learning, neighborhood topologies, novel mutation-crossover operators, dynamic population size adjustments, and so on, and attempting to determine when various aspects of these can be used beneficially in problem solving.

DF: What led you to this work?

PS: The “no free lunch” theorem (David Wolpert and William Macready, IEEE Transactions on Evolutionary Computation, Vol. 1:1, pp. 67-82, 1996) states that it is impossible to develop a single algorithm that’s the most effective for solving all classes of problems. This is common sense, somewhat like a theorem similar to there’s no panacea in medicine, no silver bullet in politics, and so on. In other words, different problems require specially designed solutions. Hence, developing strategies that enhance swarm methods will enable us to devise methods to address specific problems with more success.

DF: What do you think someone not working directly in this area of computational intelligence would gain by attending your symposium?

PS: Swarm intelligence algorithms are able to solve diverse classes of optimization problems, such as constrained, multiobjective, dynamic, combinatorial, discrete, and so on. In addition, by emulating swarm behaviors, they are able to solve problems involving multi-agents such as robotic soccer, surveillance by unmanned vehicles, and so on. These problems are frequently encountered in diverse fields. Hence, academics, researchers, and practitioners from various domains will find something exciting to them when attending our meeting.


Contact information:
Ponnithurai Suganthan: Nanyang Technological University, Singapore,
David Fogel: Natural Selection, Inc., 6480 Weathers Pl., Suite 350, San Diego, CA 92121, (858) 455-6449.

The URL for the 2017 IEEE Symposium Series on Computational Intelligence is forthcoming.

To read more:

A. Rajasekhar, N. Lynn, S. Das, and P.N. Suganthan, “Computing with the collective intelligence of honey bees – a survey,” Swarm and Evolutionary Computation, Doi: 10.1016/j.swervo.2016.06.01, Feb, 2017.

V. F. Ribeiro, D.A. Pamplona, J.A.T.G. Fregnani, I.R. de Oliveira, and L. Weigang, “Modeling the swarm optimization to build effective continuous descent arrival sequences,” 2016 IEEE 19th International Conference on Intelligent Transportation Systems, Doi: 10.1109/ITSC.2016.7795640.

S. Das, S.S. Mullick, and P.N. Suganthan, “Recent advances in differential evolution – An updated survey,” Swarm and Evolutionary Computation, Vol. 27, 1-30, 2016.

James M. Keller, Derong Liu, and David B. Fogel, Fundamentals of Computational Intelligence: Neural Networks, Fuzzy Logic, and Evolutionary Computation, John Wiley, NY, 2016