Tutorials

 

Tutorial 1. Criticality in Particle and Robot Swarms

Dr. J. Michael Herrmann and Dr. Thomas Joyce
School of Informatics, The University of Edinburgh, UK


1. Topic


1.1 Topic of the tutorial
The tutorial will present a dynamical systems approach to swarm intelligence with particular emphasis on applications in metaheuristic optimisation and swarm robotics. We will mainly emphasize recent convergent developments that call for a general theoretical framework.


1.2 Interest to the swarm intelligence community
Biological inspiration is much more than a way of popularising certain computational methods by linking them to well known phenomena in nature. We follow the more general idea of Per Bak’s “How nature works” in order to provide a unique approach to swarm intelligence that is based on the principle of criticality. Similar approaches have been employed in various contexts and various optimisation algorithms, such as PSO, DE, and cuckoo search, in the last two decades. Theoretical methods in order to ensure convergence (or non-convergence) of an algorithm are often complex, although mathematically elementary. A slightly more advanced mathematical framework can reveal similarities between otherwise unrelated approaches and is thus able to tame the unbounded number of variants of existing algorithms. As we can show clear advantages for concrete algorithms as well as practical applications, the advancement of the theoretical methods will appear worthwhile to potential participants.

1.3 Issues, methods and opportunities
Criticality has been studied in the context (of modelling) of neural dynamics (Beggs and Plenz, 2003) and examples from a number of real systems (earth quakes, family trees, sand piles and rice piles, coupled physical systems, evolution, biological motor control). Being first studied in natural systems, SOC is an enormous opportunity for fields like metaheuristic optimisation, organic computing, robot collectives, shared control, human-machine interaction etc. We will proceed with the development of the theory specficially with repect to population-based optimisation algorithms, optimal exploration, experimental design, and move on to applications to the interaction among multiple adaptive systems, active learning, and control of autonomous robots and prostheses. In the course of the presentation we will present a number of techniques and experiences such as optimsation algorithms for realistic problems, unbiased estimation of parameters of power law distribution, issues with large-scale simulations, data structures, discretisation, and dimensionality, finite-size scaling, problems with self-averaging
in critical systems, algorithms combining exploratory and goal-oriented learning, etc.

1.4 Outline
The outline is for a 3h version of the tutorial. Adaptations to a shorter time can easily be made by leaving out the robotics part, and to a longer one by extending that part.

1.4.1. Introduction:
(a) Criticality in biological, particle and robot swarms (15 min)
(b) Random dynamical systems (15 min)
(c) Criticality (15 min)
1.4.2. Convergence of metaheuristic optimisation
(a) GA (15 min)
(b) DE (15 min)
(c) PSO (15 min)
(d) ACO (15 min)
1.4.3. Applications
(a) Performance in optimisation problems (10 min)
(b) Artificial (and real) chemistry (10 min)
(c) Swarms for dynamical design (10 min)

1.4.4. Robot swarms
1.4.5. Discussion, outlook, post-tutorial resources

2. Goals

We wish to point researchers and developers from various backgrounds to the advantages of working in marginally stable regimes which provide optimal sensitivity and flexibility as well as predictability and performance at the same time. The general concepts will be explained and presented in various contexts. In this way we can facilitate the transfer between areas which is one of the benefits of a mathematical approach. We will provide the participants with working knowledge to advance the theory of metaheuristic optimisation that is directly useful for applications.

3. Organizers

Dr. J. Michael Herrmann
Edinburgh University, School of Informatics, Inst. for Perception Action and Behaviour,
Informatics Forum, 10 Crichton St., Edinburgh, EH8 9AB, U.K.
Tel: 0044 131 651 7177, Fax: 0044 131 651 3435
E-mail: michael.herrmann@ed.ac.uk, WWW: http://www.inf.ed.ac.uk/

Dr. Thomas Joyce
Edinburgh University, School of Engineering, Alexander Graham Bell building, The
King’s Buildings, Edinburgh EH9 3FG, U.K.
Tel: 0044 131 650 7343, Fax: 0044 131 650 6554
E-mail: t.joyce@ed.ac.uk, WWW: https://www.eng.ed.ac.uk/about/people/dr-thomasjoyce

4. Speaker

J. Michael Herrmann
Edinburgh University, School of Informatics, Inst. for Perception, Action and Behaviour,
Informatics Forum, 10 Crichton St., Edinburgh, EH8 9AB, U.K.
Tel: 0044 131 651 7177, Fax: 0044 131 651 3435
E-mail: michael.herrmann@ed.ac.uk, WWW: http://www.inf.ed.ac.uk

5. Intended audience

We expect a familiarity with practical applications of one or more metaheuristic optimisation algorithms such as PSO, GA or ACO. Also basic understanding of concepts from linear algebra (such as eigenvalues), calculus and probability theory will be useful in order to obtain full benefit from the tutorial. Some of the ideas to be presented here are around in the community for two decades, but have been rarely been considered under a general framework. Therefore the ICSI 2017 provides an optimally timed opportunity for the tutorial.

6. Format
The tutorial will be delivered as a direct presentation which aims at developing some aspects interactively. There will be also 15 min for general question and answers. The speaker will be available after the tutorial for 1-to-1 discussions or smaller groups.

7. Publicity
After the notification of acceptance of this proposal we will ask the conference organisers announce the tutorial on the relevant mailing lists. We will provide a short announcement text. The conference organisers will provide a landing page for the tutorial, from which we can link our content. Further support in publicising and organising the tutorial by the conference organization will be appreciated.

8. Results

8.1 Materials to be made available to attendees
If local facilities can be used, the participants will receive printouts of all slides and an extensive essay on the topic that includes details not covered by the presentation, and a comprehensive list of references.
8.2 Long-term services
We plan to webified the tutorial after the conference and to make it available to the public. Demos used in the tutorial will be downloadable on the webpage.


Tutorial 2: Dynamic Multi-objective Optimization: Concepts, Challenges, and Recent Advances

Dr. Maysam Orouskhani
Science and Research Branch, Islamic Azad University, Tehran, Iran.
Website: https://srbiau.academia.edu/MaysamOrouskhani

Abstract

Many real-world problems not only require the simultaneous optimization of a number of objective functions, but also need to track the changing optimal solutions. These problems are called: Dynamic multi-objective optimization. These optimization problems do not have a single goal to solve, but many goals that are in conflict with one another- improvement in one goal leads to deterioration of another. Most research in multi-objective optimization has been conducted on static problems and most research on dynamic problems has been conducted on single-objective optimization. In dynamic multi-objective optimization problems (DMOOP) where either the objective functions or the constraints change over time, an optimization algorithm should be able to find, obtain and track the changing set of optimal solutions (POS) and the approximated Pareto front as close as true Pareto front (POF). Therefore, the dynamic multi-objective optimization algorithm also has to deal with the problems of a lack of diversity and outdated memory (similar to dynamic single-objective optimization). In order to determine whether an algorithm can work efficiently in changing environments, it should be evaluated on standard benchmark functions. In addition, to measure the performance of the algorithm and compare it to other algorithms, performance metrics are required. This program aims at bringing academic researchers and practitioners together to review the concepts and definitions, algorithms and techniques, standard benchmark functions, performance measures and challenges of dynamic multi-objective optimization.

Goals

1. Introduction to Dynamic Multi-objective Optimization (15 mins)
- Dynamic Multi-objective Optimization Problems (DMOOP)
- Concepts and Definitions
- Changing POS and POF
- Types of DMOO
- What should a DMOO algorithm do?
2. Classification of DMOOP Algorithms(15 mins)
- Diversity handling techniques
- Prediction-based techniques
- Memory-based techniques
3. Challenges of solving a DMOOP Problems: (15 mins)
- Change detection
- Response to change
- Frequency and Severity of change
- Loosing diversity
- Convergence speed
4. Benchmarks: (15 mins)
- Suite and set of DMOO benchmark functions
- Challenges of DMOO test functions
5. Performance Measures: (20 mins)
- Measures based on accuracy
- Measures based on diversity
- Hybrid performance metrics
- Challenges of Performance Measures
6. Application to real-world Engineering Problems (15 mins)

Biography:

Maysam Orouskhani is PhD of computer engineering at the science and research branch of Islamic Azad University, Iran. His PhD thesis is: “Dynamic multi-objective optimization using hybrid algorithm of Cat swarm optimization and Borda count method”. His tutorial about “dynamic multi-objective optimization (DMOO)” has been accepted to present at ICSI2016, AIPR2016, and ICTCK2016. Furthermore, He is the winner of CEC competition on Dynamic Multi-objective Optimization at CEC2015 held in Japan.

 


 

ICSI Call for Tutorial Proposals

ICSI 2017 technical program will include tutorial sessions. Their aim is to provide a fast introduction to some topics of interest to the swarm intelligence community.

Prospective organizers of tutorial sessions should submit proposals indicating:

* Title of the tutorial.
* Rationale of the need for the tutorial at ICSI.
* Short biography and recent photo of the organizers.

Proposals are due on or before Feb 28, 2017 and should be sent via e-mail (in either pdf or plain ASCII text form) to the tutorial sessions chairs (Prof.

Milan Tuba and/or Dr. Andreas Janecek) and forward to ICSI 2017 secretariat at icsi2017@ic-si.org.

Notification of acceptance will be sent to the organizers no later than March 30, 2017.