Emmanuel Kieffer

Clustering approaches for visual knowledge exploration in molecular interaction networks

BackgroundBiomedical knowledge grows in complexity, and becomes encoded in network-based repositories, which include focused, expert-drawn diagrams, networks of evidence-based associations and established ontologies. Combining these structured information sources is an important computational challenge, as large graphs are difficult to analyze visually.ResultsWe investigate knowledge discovery in manually curated and annotated molecular interaction diagrams. To evaluate similarity of content we use: i) Euclidean distance in expert-drawn diagrams, ii) shortest path distance using the underlying network and iii) ontology-based distance. We employ clustering with these metrics used separately and in pairwise combinations. We propose a novel bi-level optimization approach together with an evolutionary algorithm for informative combination of distance metrics. We compare the enrichment of the obtained clusters between the solutions and with expert knowledge. We calculate the number of Gene and Disease Ontology terms discovered by different solutions as a measure of cluster quality.Our results show that combining distance metrics can improve clustering accuracy, based on the comparison with expert-provided clusters. Also, the performance of specific combinations of distance functions depends on the clustering depth (number of clusters). By employing bi-level optimization approach we evaluated relative importance of distance functions and we found that indeed the order by which they are combined affects clustering performance.Next, with the enrichment analysis of clustering results we found that both hierarchical and bi-level clustering schemes discovered more Gene and Disease Ontology terms than expert-provided clusters for the same knowledge repository. Moreover, bi-level clustering found more enriched terms than the best hierarchical clustering solution for three distinct distance metric combinations in three different instances of disease maps.ConclusionsIn this work we examined the impact of different distance functions on clustering of a visual biomedical knowledge repository. We found that combining distance functions may be beneficial for clustering, and improve exploration of such repositories. We proposed bi-level optimization to evaluate the importance of order by which the distance functions are combined. Both combination and order of these functions affected clustering quality and knowledge recognition in the considered benchmarks. We propose that multiple dimensions can be utilized simultaneously for visual knowledge exploration.

A new Co-evolutionary Algorithm Based on Constraint Decomposition

Handling constraints is not a trivial task in evolutionary computing. Even if different techniques have been proposed in the literature, very few have considered co-evolution which tends to decompose problems into easier sub-problems. Existing co-evolutionary approaches have been mainly used to separate the decision vector. In this article we propose a different co-evolutionary approach, referred to as co-evolutionary constraint decomposition algorithm (CCDA), that relies on a decomposition of the constraints. Indeed, it is generally the conjunction of some specific constraints which hardens the problems. The proposed CCDA generates one sub-population for each constraint and optimizes its own local fitness. A sub-population will first try to satisfy its assigned constraint, then the remaining constraints from other sub-populations using a cooperative mechanism, and finally the original objective function. Thanks to this approach, sub-populations will have different behaviors and solutions will approach the feasible domain from different sides. An exchange of information is performed using crossover between individuals from different sub-populations while mutation is applied locally. Promising mutated features are then transmitted through mating. The proposed CCDA has been validated on 8 well-known benchmarks from the literature. Experimental results show the relevance of constraint decomposition in the context of co-evolution compared to state-of-the-art algorithms.

Bayesian optimization approach of general bi-level problems

Real-life problems including transportation, planning and management often involve several decision makers whose actions depend on the interaction between each other. When involving two decision makers, such problems are classified as bi-level optimization problems. In terms of mathematical programming, a bi-level program can be described as two nested problems where the second decision problem is part of the first problems constraints. Bi-level problems are NP-hard even if the two levels are linear. Since each solution implies the resolution of the second level to optimality, efficient algorithms at the first level are mandatory. In this work we propose BOBP, a Bayesian Optimization algorithm to solve Bi-level Problems, in order to limit the number of evaluations at the first level by extracting knowledge from the solutions which have been solved at the second level. Bayesian optimization for hyper parameter tuning has been intensively used in supervised learning (e.g., neural networks). Indeed, hyper parameter tuning problems can be considered as bi-level optimization problems where two levels of optimization are involved as well. The advantage of the bayesian approach to tackle multi-level problems over the BLEAQ algorithm, which is a reference in evolutionary bi-level optimization, is empirically demonstrated on a set of bi-level benchmarks.

Hybrid mobility model with pheromones for UAV detection task

Over the last years, the activities related to unmanned aerial vehicle have seen an exponential growth in several application domains. In that context, a great interest has been devoted to search and tracking scenarios, which require the development of novel UAV mobility management solutions. Recent works on mobility models have shown that bio-inspired algorithms such as ant colonies, have a real potential to tackle complex scenarios. Nevertheless, most of these algorithms are either modified path planning algorithms or dynamical algorithms with no a priori knowledge. This paper proposes H3MP, a hybrid model based on Markov chains and pheromones to take advantage of both static and dynamic methods. Markov chains are evolved to generate a global behavior guiding UAVs to promising areas while pheromones allow local and dynamical mobility management thanks to information sharing between UAVs via stigmergy. Experimental results demonstrate the ability of H3MP to rapidly detect and keep watch on targets compared to random and pheromone based models.

A Novel Co-evolutionary Approach for Constrained Genetic Algorithms

In this paper, a novel type of co-evolutionary algorithm based on constraints decomposition (CHCGA) is proposed. Its principle consists in dividing an initial constrained problem into a sufficient number of sub-problems with weak constrained domains where feasible solutions can be easily determined. One sub-population for each sub-problems are then evolved independently and merged when they become compatible with each other, i.e. they contain enough mutually feasible solutions. Experimental results on the Cloud Brokering optimization problem have demonstrated a strong solution quality gain compared to a standard genetic algorithm.

Multi-objective evolutionary approach for the satellite payload power optimization problem

Todays world is a vast network of global communications systems in which satellites provide high-performance and long distance communications. Satellites are able to forward signals after amplification to offer a high level of service to customers. These signals are composed of many different channel frequencies continuously carrying real-time data feeds. Nevertheless, the increasing demands of the market force satellite operators to develop efficient approaches to manage satellite configurations, in which power transmission is one crucial criterion. Not only the signal power sent to the satellite needs to be optimal to avoid large costs but also the power of the downlink signal has to be strong enough to ensure the quality of service. In this work, we tackle for the first time the bi-objective input/output power problem with multi-objective evolutionary algorithms to discover efficient solutions. A problem specific indirect encoding is proposed and the performance of three state-of-the-art multi-objective evolutionary algorithms, i.e. NSGA-II, SPEA2 and MOCell, is compared on real satellite payload instances.