Antoine Nongaillard

A Realistic Approach to Solve the Nash Welfare

The multi-agent resource allocation problem is the negotiation of a set of resources among a population of agents, in order to maximize a social welfare function. The purpose of this study is the definition of the agent behavior which leads, if possible, to an optimal resource allocation at the end of the negotiation process as an emergent phenomenon. This process can be based on any kind of contact networks. Our study focuses on a specific notion: the Nash product, which has not the drawbacks of the other widely used notions. However, centralized approaches cannot handle large instances, since the social function is not linear. After a study of different bilateral transaction types, we underline the most efficient negotiation policy in order to solve the multi-agent resource allocation problem with the Nash product and provide an adaptive, scalable and anytime algorithm.

The Reputation Evaluation Based on Optimized Hidden Markov Model in E-Commerce

Nowadays, a large number of reputation systems have been deployed in practical applications or investigated in the literature to protect buyers from deception and malicious behaviors in online transactions. As an efficient Bayesian analysis tool, Hidden Markov Model (HMM) has been used into e-commerce to describe the dynamic behavior of sellers. Traditional solutions adopt Baum-Welch algorithm to train model parameters which is unstable due to its inability to find a globally optimal solution. Consequently, this paper presents a reputation evaluation mechanism based on the optimized Hidden Markov Model, which is called PSOHMM. The algorithm takes full advantage of the search mechanism in Particle Swarm Optimization (PSO) algorithm to strengthen the learning ability of HMM and PSO has been modified to guarantee interval and normalization constraints in HMM. Furthermore, a simplified reputation evaluation framework based on HMM is developed and applied to analyze the specific behaviors of sellers. The simulation experiments demonstrate that the proposed PSOHMM has better performance to search optimal model parameters than BWHMM, has faster convergence speed, and is more stable than BWHMM. Compared with Average and Beta reputation evaluation mechanism, PSOHMM can reflect the behavior changes of sellers more quickly in e-commerce systems.

Nash Welfare Allocation Problems: Concrete Issues

The allocation of m resources between n agents is an AI problem with a great practical interest for automated trading. The general question is how to configure the behavior of bargaining agents to induce a socially optimal allocation. The literature contains many proposals for calculating a social welfare but the Nash welfare seems to be the one which has the most interesting properties for a fair agent society. It guarantees that all resources are fairly distributed among agents respecting their own preferences. This article shows first that the computation of this welfare is a difficult problem, contrary to common intuition. Many counter-examples describe the pitfalls of this resolution. In a second step, we describe our distributed multi-agent solution based on a specific agent’sbehavior and the results we get on difficult instances. We finally claim that this anytime solution is the only one able to effectively address this problem of obvious practical interest.

A sensitivity analysis approach to identify key environmental performance factors

Life cycle assessment (LCA) is widely used in design phase to reduce the product’s environmental impacts through the whole product life cycle (PLC) during the last two decades. The traditional LCA is restricted to assessing the environmental impacts of a product and the results cannot reflect the effects of changes within the life cycle. In order to improve the quality of ecodesign, it is a growing need to develop an approach which can reflect the changes between the design parameters and product’s environmental impacts. A sensitivity analysis approach based on LCA and ecodesign is proposed in this paper. The key environmental performance factors which have significant influence on the products’ environmental impacts can be identified by analyzing the relationship between environmental impacts and the design parameters. Users without much environmental knowledge can use this approach to determine which design parameter should be first considered when (re)designing a product. A printed circuit board (PCB) case study is conducted; eight design parameters are chosen to be analyzed by our approach. The result shows that the carbon dioxide emission during the PCB manufacture is highly sensitive to the area of PCB panel.

A Multi-agent Resource Negotiation for Social Welfare

This study seeks to provide scalable and distributed algorithms to solve the resource allocation problem within an agent community. We propose an approach that can be applied to any kind of contact network, any range for the utility values, for the most important social welfare notions, avoiding the centralized approach drawbacks. In that purpose, we study various agent behaviors. We show that there exists in each case a simple behavior leading the negotiation process to a socially optimal resource allocation as an emergent phenomenon, or to a socially close solution if the need arises. We give, for each social welfare notion, the agent behavior to implement in order to solve the problem.

Evaluation model for e-tourism product: a hidden Markov model-based algorithm

Nowadays, e-tourism is widely pronounced as a kind of web marketing for tourism businesses. Tourist agencies and tourism service providers are able to access their customers directly in a cost-effective way. In this context, due to the high uncertainty of business using the internet, reputation systems play a key role in e-tourism business. After a literature review, and an analysis of the limits of the existing models, we propose a new reputation evaluation model based on optimisation hidden Markov model with particle swarm optimisation (PSOHMM). In order to apply this algorithm into e-tourism services/products evaluation, a number of experiments are presented. These simulation experiments demonstrate the efficiency of the PSOHMM, and the fact that it is more stable than others algorithm to search quickly optimal solutions and to detect any change in the behaviour of the e-tourism providers.

A Technique for Large Automated Mechanism Design Problems

Automated mechanism design (AMD) seeks to find, using algorithms, the optimal rules of interaction (a mechanism) between selfish and rational agents in order to get the best outcome. Here optimal is defined by the objective function of the designer of the mechanism where the function has usually some desirable properties (e.g., Pareto optimal). A difficulty with AMD lies in the size of the optimization problem that one needs to solve in order to select the best mechanism: there is a huge number of variables (and constraints but to a lesser extent) even for AMD instances of relatively small size. We study how to adapt the column generation techniques in order to solve the linear programming UP formulation of the AMD problem and compare its efficiency with the classical simplex algorithm for linear programs, on a bartering of goods example. We show that the resulting column generation algorithm is very quickly faster than the simplex algorithm for a fixed number of types (i.e., preference relations) on the goods as the number of goods increases, and then for a fixed number of goods as the number of types increases. Moreover, we show that, as the number of goods increases, the percentage of variables that need to be explicitly considered by the column generation techniques comes down very fast while the simplex algorithm must always consider explicitly all variables.

Multilevel Agent-Based Modelling for Assignment or Matching Problems.

Assignment or matching problems have been addressed by various multi-agent methods, focused on enhancing privacy and distribution. Nevertheless, they little rely on the organisational structure provided by Multi-Agent Systems (MAS). We rather start from the intrinsic ability of multilevel MAS to represent intermediate points of view between the individual and the collective levels, to express matching or assignment problems in a homogeneous formalism. This model allows to define relevant metrics to assess the satisfaction of agent groups and allow them to build solutions that improve the overall well-being without disclosing all their individual information.

A word sense disambiguation method for feature level sentiment analysis

Sentiment analysis is an automatic method used to determine that the opinion of a person about a subject is positive or negative. One of the most important tasks in sentiment analysis is to disambiguate the sense of words according to context. Most errors in sentiment analysis are because of improper sense disambiguation. Few methods for this purpose have been proposed in literature. However, they are not able to properly determine the context of word in a sentence. In addition, the lexicon dictionaries used by these methods lack word senses and also do not provide a context matching technique. These issues need to be addressed in order to improve the performance of sentiment analysis so that it can be used by customers and manufacturers for decision making. In this paper, we propose a feature level sentiment analysis system, which produces a summary of opinions about product features. A word sense disambiguation method is introduced which accurately determines the sense of a word within a context while determining the polarity. In addition, a heuristic based method is proposed in order to determine the text where opinion about a product feature is expressed. The results show that the proposed methods achieve better accuracy than existing methods.

La maximisation du bien-être utilitaire des sociétés d'agents

In this study, we consider the resource allocation problem within an entity set. Centralized approaches that are often used to solve such allocation problems have several important drawbacks. For instance, they require that the entities publish their own preferences, which compromise the privacy of these data. At the opposite, the provided decentralized appraoch is more flexible. This approach, which is based on negotiations among agents, can be applied with any kind of contact network and any range for the utility values. For this purpose, we study various agent behaviors in order to identify which one leads to the optimal resource allocation required, thanks to an emergent phenomenon.