Pierre Collet

A Swarm Approach for Automatic Auditing of Pedagogical Planning

This paper describes a possible application of swarm intelligence techniques in e-learning: an auditing tool for pedagogical planning. Swarm intelligence techniques can be applied to a web system thanks to the fact that the available online material can be organized in a graph by means of hyperlinks. In this case, the swarm that moves on the graph is composed of students who unconsciously leave pheromones in the environment depending on their success or failure. The paper presents the system and shows its capacity to serve as an auditing tool for courses designed by a pedagogical team.

Artificial Immune Ecosystems: the role of expert-based learning in artificial cognition

The rapid evolution of IT ecosystems significantly challenges the security models our infrastructures rely on. Beyond the old dichotomy between open and closed systems, it is now necessary to handle securely the interaction between heterogeneous devices building dynamic ecosystems. To this regard, bio-inspired approaches provide a rich set of conceptual tools, but they have failed to lay the basis for robust and efficient solutions. Our research effort intends to revisit the contribution of artificial immune system research to bring immune properties: security, resilience, distribution, memory, into IT infrastructures. Artificial immune ecosystems support a comprehensive model for anomaly detection and characterization, but their cognitive capacity are limited by the state of the art in machine learning and the rapid evolution of cybersecurity threats so far. We therefore propose to enrich the cognitive process with expert-based learning for reinforcement, classification and investigation. Application to system supervision using system logs and supervision time series confirms the relevance and performance of this model.

POEM-COPA Collaborative Open Peer Assessment

This paper describes the Collaborative Open Peer Assessment formative evaluation proposed by the POEM Personalised Open Education for the Masses platform of the Complex Systems Digital Campus UNESCO UniTwin. COPA allows teachers to use open questions and answers to assess the level of the great number of students that is typical of Massive Open Online Courses. Indeed, it offers peer evaluation as well as a highly pedagogic experience: beyond answering questions, students are also asked to actively participate by posing their own questions. However, it has been found during tests that COPA can also be successfully used on much smaller groups ( ∼ 20 students or even less) after face-to-face courses, to help the students revise. COPA has already been used on more than 10 courses, with great feedback from the students. COPA is available as a standalone version. It will also soon be available as an Open-edX plugin, (allowing COPA to be used in any Open-edX based MOOC) as well as under the MOODLE platform, for teachers who would like to use open questions and open answers, rather than the more simple Multiple Choice Questions.

Introducing Pareto-based MOEA to Neurosurgery Preoperative Path Planning

This paper presents the first implementation of NSGA-II in neurosurgery preoperative path planning. Deep Brain Stimulation (DBS) is a surgical treatment of Parkinsons disease that can be regarded as a multi-objective optimization problem, searching for the best compromise between multiple electrode placement rules. Most of the current automatic decision-making processes use aggregative approaches with single objective optimization, even though they are known for their inability to find all Pareto-optimal solutions. Firstly, we show this is the case on 20 datasets of patients by comparing our implementation of NSGA-II to the weighted sum (WS) strategy. Then, we show that it requires about 9 hours to find equivalent results using a deterministic scan of the search space where NSGA-II does it in about 3mn. This paper presents an objective validation that even simple techniques such as NSGA-II should be used by surgeons over more intuitive weighted based methods.

Application of Evolutionary Algorithms for the Optimization of Genetic Regulatory Networks

Synthetic biology aims at reinvesting theoretical knowledge from various do-mains (biology, engineering, microelectronics) for the development of new bio-logical functions. Concerning the design of such functions, the classical trial-error approach is expensive and time consuming. Computer-aided design is therefore of key interest in this field. As for other domains, such as microelectronics or robotics, evolutionary algo-rithms can be used to this end. This article is a first step in this direction: it describes the optimization of an existing artificial gene regulatory network using evolutionary algorithms. Evolutionary algorithms successfully find a good set of parameters (the simu-lated response of the system which fits at 99 % the expected response) in about 200 s (corresponding to 5000 generations) on a standard computer. This is the proof of concept of our approach. Moreover, results analysis allows the biologist not only to save time during the design process but also to study the specificity of a system.

Know-How modelling for e-Learning

By defining Know-How and considering the collection of relevant data, we provide a computational cognitive model that is based on the problem-solving and task-analysis literature. The model is of help for the conception of an automatic assessment system of Skills acquisition related to Know-How. The platform we are building is implementing the method as part of the “Know-How Digital Campus” project (KH-DC).