About the course
The course will cover many different aspects of Artificial Intelligence with many different approaches. Here is a short explanation of the topics you will discuss.
Table Of Contents :
1- General Introduction to Artificial Intelligence
2- Swarm Intelligence and Ant Colony Optimization
4- Optimisation and metaheuristics
6- Data Mining in BioInformatics
7- Text Mining
1- General Introduction to Artificial Intelligence
Teacher(s) : Hugues Bersini
IRIDIA, headed by Hugues Bersini, is the Artificial Intelligence Research Laboratory of ULB. It is deeply involved in theoretical and applied research on the building of competent software for complex applications. The objective of artificial intelligence is to render our computer even more competent, more convivial, more invisible, everywhere and indispensable at the same time. This artificial intelligence is indeed implemented everywhere: in the functioning of GPS and what the car tells you when she speaks to you, in the functioning of Google, in Hollywood pictures, in medical diagnoses and video game, in systems exploiting automatic speech recognition or synthesis of word, in robots that construct cars, execute heart surgery or play football.
Our major domains of study are artificial intelligence techniques for process control and classification, nature inspired heuristics for the solution of combinatorial and continuous space optimisation problems, data mining and object oriented technologies. Research activities at IRIDIA are closely connected to fundamental research on the development of soft computing and the reduction of complexity by emergent and adaptive computing. Many of the software solutions developed by IRIDIA are operational on industrial sites such as automatic default recognition for Glaverbel, medical automatic diagnosis for Erasme Hospital, time series prediction for Master Food and Dieteren, process control for Fafer, Honeywell and Siemens. This introductory course will sketch the main ideas and technologies that compose Artificial Intelligence.
2- Swarm Intelligence and Ant Colony Optimization
Teacher(s) : Marco Dorigo and Mauro Birratari
Swarm intelligence is the discipline that deals with natural and artificial systems composed of many individuals that coordinate using decentralized control and self-organization. In particular, it focuses on the collective behaviors that result from the local interactions of the individuals with each other and with their environment. The characterizing property of a swarm intelligence system is its ability to act in a coordinated way without the presence of a coordinator or of an external controller. The course will first give a general introduction to swarm intelligence and then briefly introduce the ant colony optimization metaheuristic.
3- Swarm Robotics
Teacher(s) : Marco Dorigo and Mauro Birratari
Swarm robotics could be defined as the application of swarm intelligence principles to the control of groups of robots. In this course I will discuss results of Swarm-bots, an experiment in swarm robotics. A swarm-bot is an artifact composed of a swarm of assembled s-bots. The s-bots are mobile robots capable of connecting to, and disconnecting from, others-bots. In the swarm-bot form, the s-bots are attached to each other and, when needed, become a single robotic system that can move and change its shape. S-bots have relatively simple sensors and motors and limited computational capabilities. A swarm-bot can solve problems that cannot be solved by s-bots alone. In the talk, I will shortly describe the s-bots hardware and the methodology we followed to develop algorithms for their control. Then I will focus on the capabilities of the swarm-bot robotic system by showing video recordings of some of the many experiments we performed to study coordinated movement, path formation, self-assembly, collective transport, shape formation, and other collective behaviors.
4- Optimisation and metaheuristics
Teacher(s) : Thomas Stuezle
When searching for the global optimum solution to complex problems, one is generally faced with a fundamental conflict between precision, reliability and computing time: Each optimization method represents a particular compromise. Genetic Algorithms (GAs) and Evolutionary Strategies (ESs), which are new promising optimisation methods originaly influenced by Darwinnian Selectionist theory, present unquestionable advantages and original principles such as : population-based search, recombination and stochastic mechanisms (especially the ability to 'rough out' a problem reliably by finding the most promising regions of the entire search space). However, they often represent an unsatisfactory compromise: indeed, they suffer from a certain inefficiency, characterized by a slow convergence and a lack of accuracy when a high-quality solution is required.
In contrast, classical hill-climbing and iterated local search methods appear to realize another extreme compromise in solving the conflict: they focus solely on precision and computation time at the expense of reliability and rush to the first discovered local extremum. In the course, we will explain why these considerations, highlighting the complementary properties of GAs and hill-climbing methods, suggest that hybridization between both approaches may lead to improved performances. As a consequence, this has been a center of focus and a constant line of research for IRIDIA in the field of optimization.
5- Evolutionary Game Theory
Teacher(s) : Francisco Santos
Evolutionary game theory (EGT) is the application of interaction dependent strategy drift in populations to game theory. It originated in 1973 with John Maynard Smith and George R. Price's formalization of evolutionary stable strategies as an application of the mathematical theory of games to biological contexts, arising from the realization that frequency dependent fitness introduces a strategic aspect to evolution. EGT differs from classical game theory by focusing on the dynamics of strategy change more than the properties of strategy equilibria. Despite its name, evolutionary game theory has become of increased interest to economists, sociologists, anthropologists, and philosophers.
The common methodology to study the evolutionary dynamics in games is through replicator equations. Continuous replicator equations assume infinite populations, continuous time, complete mixing and that strategies breed true. The attractors (stable fixed points) of the equations are equivalent with evolutionarily stable states. Departing from the prisoner dilemma, they help to explain how cooperators can emerge in a multi-agent system despite the fact that each agent could more profitably choose to adopt a defective attitude.
6- Data Mining in BioInformatics
Teacher(s) : David Venet and David Weiss
Microarrays are a revolutionary new technology with great potential to provide accurate medical diagnosis, help find the right treatment and cure for many diseases and provide a detailed genome-wide molecular portrait of cellular states. They appear today very promising and extend the possibilities of applying computational analysis and data mining to aid research in biology and medicine.
There is little doubt about the potential of computational and statistical analysis of molecular probes to improve the understanding of the cell and the possibilities of molecular medicine. Finding new insights into the molecular basis of biological processes and searching for new drugs and treatments is a problem of high complexity and where the techniques of molecular biology has been applied for many decades. The process is analogous to a large search of a few molecular entities, connections or relationships in a large sea of possibilities.
We will organize a visit in a medical laboratory research where they are deeply involved in the use of microarray technologies for cancer diagnosis and treatment.
7- Text Mining
Teacher(s) : Nicolas Van Zeebroek, Mohammed Ben-Hadou and Pierre Sener
Presentation of Mentis and Master, Food Society for Data Mining, and of the IRIS society for Text Mining
Mentis, a recent spin off of IRIDIA, has a large set of tools in Data Mining and in Text Mining. Text mining is a set of algorithmic recipes aiming at extracting relevant information out of a large corpus of unstructured texts. With this set of tools, Mentis can deliver ad-hoc studies and services for his customers. The main basic services provided by Mentis are the following:
-
Data Mining Services:
- data cleaning
- rules and pattern discovery
- forecasting
- profiling
- pattern recognition
-
Text Mining Services:
- Automatic classification of text-based items
- Specific Information extraction
- Automatic matching of text-based items
In addition to these basic services, Mentis can provide tailored services in Data Mining and in Text Mining.
Some examples of consultancy services provided by Mentis:
- For Masterfoods : evaluation of the success of advertising campaigns
- For SupplyChange : matching of hospital?s buying descriptions against high quality suppliers catalogue systems.
Mentis will organize visits at the companies they work for.
