On learning dynamics underlying the evolution of learning rules.

In order to understand the development of non-genetically encoded actions during an animal's lifespan, it is necessary to analyze the dynamics and evolution of learning rules producing behavior. Owing to the intrinsic stochastic and frequency-dependent nature of learning dynamics, these rules are often studied in evolutionary biology via agent-based computer simulations. In this paper, we show that stochastic approximation theory can help to qualitatively understand learning dynamics and formulate analytical models for the evolution of learning rules. We consider a population of individuals repeatedly interacting during their lifespan, and where the stage game faced by the individuals fluctuates according to an environmental stochastic process. Individuals adjust their behavioral actions according to learning rules belonging to the class of experience-weighted attraction learning mechanisms, which includes standard reinforcement and Bayesian learning as special cases. We use stochastic approximation theory in order to derive differential equations governing action play probabilities, which turn out to have qualitative features of mutator-selection equations. We then perform agent-based simulations to find the conditions where the deterministic approximation is closest to the original stochastic learning process for standard 2-action 2-player fluctuating games, where interaction between learning rules and preference reversal may occur. Finally, we analyze a simplified model for the evolution of learning in a producer-scrounger game, which shows that the exploration rate can interact in a non-intuitive way with other features of co-evolving learning rules. Overall, our analyses illustrate the usefulness of applying stochastic approximation theory in the study of animal learning.

Debating the centre : the theory of principal qualificand (mukhyavisesya) in classical Indian philosophy

The theory of language has occupied a special place in the history of Indian thought. Indian philosophers give particular attention to the analysis of the cognition obtained from language, known under the generic name of śābdabodha. This term is used to denote, among other things, the cognition episode of the hearer, the content of which is described in the form of a paraphrase of a sentence represented as a hierarchical structure. Philosophers submit the meaning of the component items of a sentence and their relationship to a thorough examination, and represent the content of the resulting cognition as a paraphrase centred on a meaning element, that is taken as principal qualificand (mukhyaviśesya) which is qualified by the other meaning elements. This analysis is the object of continuous debate over a period of more than a thousand years between the philosophers of the schools of Mimāmsā, Nyāya (mainly in its Navya form) and Vyākarana. While these philosophers are in complete agreement on the idea that the cognition of sentence meaning has a hierarchical structure and share the concept of a single principal qualificand (qualified by other meaning elements), they strongly disagree on the question which meaning element has this role and by which morphological item it is expressed. This disagreement is the central point of their debate and gives rise to competing versions of this theory. The Mïmāmsakas argue that the principal qualificand is what they call bhāvanā ̒bringing into being̒, ̒efficient force̒ or ̒productive operation̒, expressed by the verbal affix, and distinct from the specific procedures signified by the verbal root; the Naiyāyikas generally take it to be the meaning of the word with the first case ending, while the Vaiyākaranas take it to be the operation expressed by the verbal root. All the participants rely on the Pāninian grammar, insofar as the Mimāmsakas and Naiyāyikas do not compose a new grammar of Sanskrit, but use different interpretive strategies in order to justify their views, that are often in overt contradiction with the interpretation of the Pāninian rules accepted by the Vaiyākaranas. In each of the three positions, weakness in one area is compensated by strength in another, and the cumulative force of the total argumentation shows that no position can be declared as correct or overall superior to the others. This book is an attempt to understand this debate, and to show that, to make full sense of the irreconcilable positions of the three schools, one must go beyond linguistic factors and consider the very beginnings of each school's concern with the issue under scrutiny. The texts, and particularly the late texts of each school present very complex versions of the theory, yet the key to understanding why these positions remain irreconcilable seems to lie elsewhere, this in spite of extensive argumentation involving a great deal of linguistic and logical technicalities. Historically, this theory arises in Mimāmsā (with Sabara and Kumārila), then in Nyāya (with Udayana), in a doctrinal and theological context, as a byproduct of the debate over Vedic authority. The Navya-Vaiyākaranas enter this debate last (with Bhattoji Dïksita and Kaunda Bhatta), with the declared aim of refuting the arguments of the Mïmāmsakas and Naiyāyikas by bringing to light the shortcomings in their understanding of Pāninian grammar. The central argument has focused on the capacity of the initial contexts, with the network of issues to which the principal qualificand theory is connected, to render intelligible the presuppositions and aims behind the complex linguistic justification of the classical and late stages of this debate. Reading the debate in this light not only reveals the rationality and internal coherence of each position beyond the linguistic arguments, but makes it possible to understand why the thinkers of the three schools have continued to hold on to three mutually exclusive positions. They are defending not only their version of the principal qualificand theory, but (though not openly acknowledged) the entire network of arguments, linguistic and/or extra-linguistic, to which this theory is connected, as well as the presuppositions and aims underlying these arguments.

Detection in urine of 4-methyl-2-hexaneamine, a doping agent.

Stimulants are banned in-competition for all categories of sports by the World Anti-Doping Agency. A simple liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay employing electrospray ionisation in positive mode was developed in that work for the quantification in urine specimens of 4-methyl-2-hexaneamine, a primary amine exhibiting sympathomimetic properties. Following a simple pretreatment procedure, the analyte was separated using a gradient mobile phase on reverse phase C8 column. Selected reaction monitoring m/z 116.2-->57.3 was specific for detection of 4-methyl-2-hexaneamine and the assay exhibited a linear dynamic range of 50-700 ng/mL. The validated method has been successfully applied to analyze the target compound in food supplements as well as in urine specimens. The administered drug (40 mg) was detected at the level of 350 ng/mL in the urine up to 4 days.

ADME pharmacogenetics: investigation of the pharmacokinetics of the antiretroviral agent lopinavir coformulated with ritonavir.

BACKGROUND: An ADME (absorption, distribution, metabolism and excretion)-pharmacogenetics association study may identify functional variants relevant to the pharmacokinetics of lopinavir co-formulated with ritonavir (LPV/r), a first-line anti-HIV agent. METHODS: An extensive search of literature and web resources helped select ADME genes and single nucleotide polymorphisms (SNPs, functional and HapMap tagging SNPs) with a proven or potentially relevant role in LPV/r pharmacokinetics. The study followed a two-stage design. Stage 1 (discovery) considered a Caucasian population (n=638) receiving LPV/r, where we selected 117 individuals with low LPV clearance (cases) and 90 individuals with high clearance (controls). Genotyping was performed by a 1536-SNP customized GoldenGate Illumina BeadArray. Stage 2 (confirmation) represented a replication study of candidate SNPs from the stage 1 in 148 individuals receiving LPV/r. The analysis led to formal population pharmacokinetic-pharmacogenetic modeling of demographic, environmental and candidate SNP effects. RESULTS: One thousand three hundred and eighty SNPs were successfully genotyped. Nine SNPs prioritized by the stage 1 analysis were brought to replication. Stage 2 confirmed the contribution of two functional SNPs in SLCO1B1, one functional SNP in ABCC2 and a tag SNP of the CYP3A locus in addition to body weight effect and ritonavir coadministration. According to the population pharmacokinetic-pharmacogenetic model, genetic variants explained 5% of LPV variability. Individuals homozygous rs11045819 (SLCO1B1*4) had a clearance of 12.6 l/h, compared with 5.4 l/h in the reference group, and 3.9 l/h in individuals with two or more variant alleles of rs4149056 (SLCO1B1*5), rs717620 (ABCC2) or rs6945984 (CYP3A). A subanalysis confirmed that although a significant part of the variance in LPV clearance was attributed to fluctuation in ritonavir levels, genetic variants had an additional effect on LPV clearance. CONCLUSION: The two-stage strategy successfully identified genetic variants affecting LPV/r pharmacokinetics. Such a general approach of ADME pharmacogenetics should be generalized to other drugs.

Étude et implémentation d'un modèle de conscience d'agent

La thèse essaie de montrer comment il est possible d'offrir une implémentation fonctionnelle d'un agent doté d'une conscience (psychologique). Une première étape étudie les différentes approches, définitions et théories de la conscience proposées par la littérature. Cette étude dégage plus particulièrement un modèle psychologique qui offre une modélisation des fonctionnalités de la conscience, de ses éléments constitutifs et des relations entre ces éléments. Cet effort de formalisation permet d'identifier les corrélations computionnelles du modèle ouvrant ainsi la voie à une implémentation fonctionnelle de la conscience. Une seconde étape réuni les outils et méthodes informatiques existants en vue de procéder à une telle implémentation. En particulier, celle-ci repose sur un modèle de communication permettant d'élaborer une machine virtuelle basée sur des processus concurrents synchronisés. La troisième étape consiste à implémenter les corrélations computationnelles dont l'une est une fonction de délibération qui, après une analyse itérative de son état et de son environnement (machine à état), aboutit à la sélection d'une action. Une deuxième fonction est la formation de contextes, autrement dit l'apprentissage d'automatismes, consistant à compiler la délibération. Cette compilation s'opère grâce à un processus concurrent reflétant le processus de délibération, dotant ainsi l'agent de la capacité d'observer son propre fonctionnement. La thèse se conclut en proposant quelques axes de recherches et d'applications futures susceptibles de prolonger le travail.