M
A
Last name: MUZY
First name: Alexandre
Institution: UOM
Links
International audience
Conditions under which compositions of component systems form a well-defined system-of-systems are here formulated at a fundamental level. Statement of what defines a well-defined composition and sufficient conditions guaranteeing such a result offers insight into exemplars that can be found in special cases such as differential equation and...
Learning dynamics at cognitive process level is difficult to study and emulate because of the complexity of intricate psychological and neuronal mechanisms and dynamics. When considering the parallel processing of a task, the difficulty relies on the execution concurrency making the process contributions indistinguishable. We present here a...
We present here a system morphism methodology to give insight into the lumping process of networks of linear systems. Lumping networks allows reducing the number of components and states to obtain simulatable models. Such lumped networks can be connected together through their input/output interfaces, using an engineering approach...
We consider the ability of Discrete Event System Speci-cation (DEVS) to provide the concepts and formalisms needed for modeling and simulation of emergent behavior. We show that DEVS provides systems components and coupling for models of systems of systems with emergent behavior. Further, DEVS coupling supports dynamic structure for adaptive...
International audience
Sequential machines can be realized by series and parallel connections of components. These connections can be automatically learned by simulation, rating and selecting component models. Our goal here is threefold: (i) discuss implications for model reconstruction and reconfiguration in model engineering, (ii) present mathematical analyses of...
Synthesis of systems constitutes a vast class of problems. Although machine learning techniques operate at the functional level, little attention has been paid to system synthesis using a hierarchical model-base. This paper develops an original approach for automatically rating component systems and composing them according to the experimental...
In Discrete Event System Specification (DEVS), the dynamics of a network is constituted only by the dynamics of its basic components. The state of each component is fully encapsulated. Control in the network is fully decentralized to each component. At dynamic structure level, DEVS should permit the same level of decentralization. However, it...
International audience
There exists usually a gap between bio-inspired computational techniques and what biologists can do with these techniques in their current researches. Although biology is the root of systems-theory and artificial neural networks, computer scientists are tempted to build their own systems independently of biological issues. This publication is a...
International audience
This paper addresses the question of optimal phenotypic plasticity as a response to environmental fluctuations while optimizing the cost/benefit ratio, where the cost is energetic expense of plasticity, and benefit is fitness. The dispersion matrix Σ of the genes' response (H = ln|Σ|) is used: (i) in a numerical model as a metric of the...
We consider the situation in which cooperating agents learn to achieve a common goal based solely on a global return that results from all agents' behavior. The method proposed is based on taking into account the agents' activity , which can be any additional information to help solving multi-agent decentralized learning problems. We propose a...
Dynamics of biological-ecological systems is strongly depending on spatial dimensions. Most of powerful simulators in ecology take into account for system spatiality thus embedding stochastic processes. Due to the difficulty of researching particular trajectories, biologists and computer scientists aim at predicting the most probable...
Activity metrics can be used to profile DEVS models before and during the simulation. It is critical to get good activity metrics of models before and during their simulation. Having a means to compute a-priori activity of components (analytic activity) may be worth when simulating a model (or parts of it) for the first time. After, during the...
No description
In reinforcement learning, credit assignment with historydependent reward is a key problem to solve for being able to model agents: (i) associating the returns from their environment with their past (series of) actions, and (ii) figuring out which past decisions are responsible for the current achievement of their goal. Usual approaches...
No description
This study introduces a novel framework for understanding the cognitive underpinnings of individual behavior which often deviates from rational decision-making aimed at maximizing rewards in real-life scenarios. We propose a structure learning approach to infer an agent's internal model, composed of a learning rule and an internal...
International audience
International audience
Event-scheduling algorithms can compute in continuous time the next occurrence of points (as events) of a counting process based on their current conditional intensity. In particular event-scheduling algorithms can be adapted to perform the simulation of finite neuronal networks activity. These algorithms are based on Ogata's thinning strategy...
We propose a simple network of Hawkes processes as a cognitive model capable of learning to classify objects. Our learning algorithm, named EWAK for Exponentially Weighted Average and Kalikow decomposition, is based on a local synaptic learning rule based on firing rates at each output node. We were able to use local regret bounds to prove...
No description