2025Activity reportProject-Team​​​‌MNEMOSYNE

2.1​​​‌ Summary

At the frontier‌ between integrative and computational‌​‌ neuroscience, we propose to​​ model the brain as​​​‌ a system of active‌ memories in synergy and‌​‌ in interaction with the​​ internal and external world​​​‌ and to simulate it‌ as a whole and‌​‌ in situation.

In​​ integrative and cognitive neuroscience​​​‌ (cf. § 3.1‌), on the basis‌​‌ of current knowledge and​​ experimental data, we develop​​​‌ models of the main‌ cerebral structures, taking a‌​‌ specific care of the​​ kind of mnemonic function​​​‌ they implement and of‌ their interface with other‌​‌ cerebral and external structures.​​ Then, in a systemic​​​‌ approach, we build the‌ main behavioral loops involving‌​‌ these cerebral structures, connecting​​ a wide spectrum of​​​‌ actions to various kinds‌ of sensations. We observe‌​‌ at the behavioral level​​ the properties emerging from​​​‌ the interaction between these‌ loops.

We claim that‌​‌ this approach is particularly​​ fruitful for investigating cerebral​​​‌ structures like the basal‌ ganglia and the prefrontal‌​‌ cortex, difficult to comprehend​​ today because of the​​​‌ rich and multimodal information‌ flows they integrate. We‌​‌ expect to cope with​​ the high complexity of​​​‌ such systems, inspired by‌ behavioral and developmental sciences,‌​‌ explaining how behavioral loops​​ gradually incorporate in the​​​‌ system various kinds of‌ information and associated mnesic‌​‌ representations. As a consequence,​​ the underlying cognitive architecture,​​​‌ emerging from the interplay‌ between these sensations-actions loops,‌​‌ results from a mnemonic​​ synergy.

In computational​​​‌ neuroscience (cf. §‌ 3.2), we concentrate‌​‌ on the efficiency of​​ local mechanisms and on​​​‌ the effectiveness of the‌ distributed computations at the‌​‌ level of the system.​​ We also take care​​​‌ of the analysis of‌ their dynamic properties, at‌​‌ different time scales. These​​ fundamental properties are of​​​‌ high importance to allow‌ the deployment of very‌​‌ large systems and their​​ simulation in a framework​​​‌ of high performance computing‌

Running simulations at a‌​‌ large scale is particularly​​ interesting to evaluate over​​​‌ a long period a‌ consistent and relatively complete‌​‌ network of cerebral structures​​ in realistic interaction with​​​‌ the external and internal‌ world. We face this‌​‌ problem in the domain​​ of autonomous robotics (​​​‌cf. § 3.4)‌ and ensure a real‌​‌ autonomy by the design​​ of an artificial physiology​​​‌ and convenient learning protocoles.‌

We are convinced that‌​‌ this original approach also​​ permits to revisit and​​​‌ enrich algorithms and methodologies‌ in machine learning (‌​‌cf. § 3.3)​​ and in autonomous robotics​​​‌ (cf. § 3.4‌), in addition to‌​‌ elaborate hypotheses to be​​ tested in neuroscience and​​​‌ medicine, while offering to‌ these latter domains a‌​‌ new ground of experimentation​​ similar to their daily​​​‌ experimental studies.

3.1 Integrative and‌​‌ Cognitive Neuroscience

The human​​ brain is often considered​​​‌ as the most complex‌ system dedicated to information‌​‌ processing. This multi-scale complexity,​​​‌ described from the metabolic​ to the network level,​‌ is particularly studied in​​ integrative neuroscience, the goal​​​‌ of which is to​ explain how cognitive functions​‌ (ranging from sensorimotor coordination​​ to executive functions) emerge​​​‌ from (are the result​ of the interaction of)​‌ distributed and adaptive computations​​ of processing units, displayed​​​‌ along neural structures and​ information flows. Indeed, beyond​‌ the astounding complexity reported​​ in physiological studies, integrative​​​‌ neuroscience aims at extracting,​ in simplifying models, regularities​‌ at various levels of​​ description. From a mesoscopic​​​‌ point of view, most​ neuronal structures (and particularly​‌ some of primary importance​​ like the cortex, cerebellum,​​​‌ striatum, hippocampus) can be​ described through a regular​‌ organization of information flows​​ and homogenous learning rules,​​​‌ whatever the nature of​ the processed information. From​‌ a macroscopic point of​​ view, the arrangement in​​​‌ space of neuronal structures​ within the cerebral architecture​‌ also obeys a functional​​ logic, the sketch of​​​‌ which is captured in​ models describing the main​‌ information flows in the​​ brain, the corresponding loops​​​‌ built in interaction with​ the external and internal​‌ (bodily and hormonal) world​​ and the developmental steps​​​‌ leading to the acquisition​ of elementary sensorimotor skills​‌ up to the most​​ complex executive functions.

In​​​‌ summary, integrative neuroscience builds,​ on an overwhelming quantity​‌ of data, a simplifying​​ and interpretative grid suggesting​​​‌ homogenous local computations and​ a structured and logical​‌ plan for the development​​ of cognitive functions. They​​​‌ arise from interactions and​ information exchange between neuronal​‌ structures and the external​​ and internal world and​​​‌ also within the network​ of structures.

This domain​‌ is today very active​​ and stimulating because it​​​‌ proposes, of course at​ the price of simplifications,​‌ global views of cerebral​​ functioning and more local​​​‌ hypotheses on the role​ of subsets of neuronal​‌ structures in cognition. In​​ the global approaches, the​​​‌ integration of data from​ experimental psychology and clinical​‌ studies leads to an​​ overview of the brain​​​‌ as a set of​ interacting memories, each devoted​‌ to a specific kind​​ of information processing 55​​​‌. It results also​ in longstanding and very​‌ ambitious studies for the​​ design of cognitive architectures​​​‌ aiming at embracing the​ whole cognition. With the​‌ notable exception of works​​ initiated by 47,​​​‌ most of these frameworks​ (e.g. Soar, ACT-R), though​‌ sometimes justified on biological​​ grounds, do not go​​​‌ up to a connectionist​ neuronal implementation. Furthermore, because​‌ of the complexity of​​ the resulting frameworks, they​​​‌ are restricted to simple​ symbolic interfaces with the​‌ internal and external world​​ and to (relatively) small-sized​​​‌ internal structures. Our main​ research objective is undoubtly​‌ to build such a​​ general purpose cognitive architecture​​​‌ (to model the brain​ as a whole in​‌ a systemic way), using​​ a connectionist implementation and​​​‌ able to cope with​ a realistic environment.

3.2​‌ Computational Neuroscience

From a​​ general point of view,​​​‌ computational neuroscience can be​ defined as the development​‌ of methods from computer​​ science and applied mathematics,​​​‌ to explore more technically​ and theoretically the relations​‌ between structures and functions​​ in the brain 57​​, 44. During​​​‌ the recent years this‌ domain has gained an‌​‌ increasing interest in neuroscience​​ and has become an​​​‌ essential tool for scientific‌ developments in most fields‌​‌ in neuroscience, from the​​ molecule to the system.​​​‌ In this view, all‌ the objectives of our‌​‌ team can be described​​ as possible progresses in​​​‌ computational neuroscience. Accordingly, it‌ can be underlined that‌​‌ the systemic view that​​ we promote can offer​​​‌ original contributions in the‌ sense that, whereas most‌​‌ classical models in computational​​ neuroscience focus on the​​​‌ better understanding of the‌ structure/function relationship for isolated‌​‌ specific structures, we aim​​ at exploring synergies between​​​‌ structures. Consequently, we target‌ interfaces and interplay between‌​‌ heterogenous modes of computing,​​ which is rarely addressed​​​‌ in classical computational neuroscience.‌

We also insist on‌​‌ another aspect of computational​​ neuroscience which is, in​​​‌ our opinion, at the‌ core of the involvement‌​‌ of computer scientists and​​ mathematicians in the domain​​​‌ and on which we‌ think we could particularly‌​‌ contribute. Indeed, we think​​ that our primary abilities​​​‌ in numerical sciences imply‌ that our developments are‌​‌ characterized above all by​​ the effectiveness of the​​​‌ corresponding computations: we provide‌ biologically inspired architectures with‌​‌ effective computational properties, such​​ as robustness to noise,​​​‌ self-organization, on-line learning. We‌ more generally underline the‌​‌ requirement that our models​​ must also mimick biology​​​‌ through its most general‌ law of homeostasis and‌​‌ self-adaptability in an unknown​​ and changing environment. This​​​‌ means that we propose‌ to numerically experiment such‌​‌ models and thus provide​​ effective methods to falsify​​​‌ them.

Here, computational neuroscience‌ means mimicking original computations‌​‌ made by the neuronal​​ substratum and mastering their​​​‌ corresponding properties: computations are‌ distributed and adaptive; they‌​‌ are performed without an​​ homonculus or any central​​​‌ clock. Numerical schemes developed‌ for distributed dynamical systems‌​‌ and algorithms elaborated for​​ distributed computations are of​​​‌ central interest here 41‌, 46 and were‌​‌ the basis for several​​ contributions in our group​​​‌ 56, 54,‌ 58. Ensuring such‌​‌ a rigor in the​​ computations associated to our​​​‌ systemic and large scale‌ approach is of central‌​‌ importance.

Equally important is​​ the choice for the​​​‌ formalism of computation, extensively‌ discussed in the connectionist‌​‌ domain. Spiking neurons are​​ today widely recognized of​​​‌ central interest to study‌ synchronization mechanisms and neuronal‌​‌ coupling at the microscopic​​ level 42; the​​​‌ associated formalism 45 can‌ be possibly considered for‌​‌ local studies or for​​ relating our results with​​​‌ this important domain in‌ connectionism. Nevertheless, we remain‌​‌ mainly at the mesoscopic​​ level of modeling, the​​​‌ level of the neuronal‌ population, and consequently interested‌​‌ in the formalism developed​​ for dynamic neural fields​​​‌ 40, that demonstrated‌ a richness of behavior‌​‌ 43 adapted to the​​ kind of phenomena we​​​‌ wish to manipulate at‌ this level of description.‌​‌ Our group has a​​ long experience in the​​​‌ study and adaptation of‌ the properties of neural‌​‌ fields 54, 53​​ and their use for​​​‌ observing the emergence of‌ typical cortical properties 51‌​‌. In the envisioned​​​‌ development of more complex​ architectures and interplay between​‌ structures, the exploration of​​ mathematical properties such as​​​‌ stability and boundedness and​ the observation of emerging​‌ phenomena is one important​​ objective. This objective is​​​‌ also associated with that​ of capitalizing our experience​‌ and promoting good practices​​ in our software production.​​​‌

In summary, we think​ that this systemic approach​‌ also brings to computational​​ neuroscience new case studies​​​‌ where heterogenous and adaptive​ models with various time​‌ scales and parameters have​​ to be considered jointly​​​‌ to obtain a mastered​ substratum of computation. This​‌ is particularly critical for​​ large scale deployments.

3.3​​​‌ Machine Learning

The adaptive​ properties of the nervous​‌ system are certainly among​​ its most fascinating characteristics,​​​‌ with a high impact​ on our cognitive functions.​‌ Accordingly, machine learning is​​ a domain 52 that​​​‌ aims at giving such​ characteristics to artificial systems,​‌ using a mathematical framework​​ (probabilities, statistics, data analysis,​​​‌ etc.). Some of its​ most famous algorithms are​‌ directly inspired from neuroscience,​​ at different levels. Connectionist​​​‌ learning algorithms implement, in​ various neuronal architectures, weight​‌ update rules, generally derived​​ from the hebbian rule,​​​‌ performing non supervised (e.g.​ Kohonen self-organizing maps), supervised​‌ (e.g. layered perceptrons) or​​ associative (e.g. Hopfield recurrent​​​‌ network) learning. Other algorithms,​ not necessarily connectionist, perform​‌ other kinds of learning,​​ like reinforcement learning. Machine​​​‌ learning is a very​ mature domain today and​‌ all these algorithms have​​ been extensively studied, at​​​‌ both the theoretical and​ practical levels, with much​‌ success. They have also​​ been related to many​​​‌ functions (in the living​ and artificial domains) like​‌ discrimination, categorisation, sensorimotor coordination,​​ planning, etc. and several​​​‌ neuronal structures have been​ proposed as the substratum​‌ for these kinds of​​ learning 50, 48​​​‌. Nevertheless, we believe​ that, as for previous​‌ models, machine learning algorithms​​ remain isolated tools, whereas​​​‌ our systemic approach can​ bring original views on​‌ these problems.

At the​​ cognitive level, most of​​​‌ the problems we face​ do not rely on​‌ only one kind of​​ learning and require instead​​​‌ skills that have to​ be learned in preliminary​‌ steps. That is the​​ reason why cognitive architectures​​​‌ are often referred to​ as systems of memory,​‌ communicating and sharing information​​ for problem solving. Instead​​​‌ of the classical view​ in machine learning of​‌ a flat architecture, a​​ more complex network of​​​‌ modules must be considered​ here, as it is​‌ the case in the​​ domain of deep learning.​​​‌ In addition, our systemic​ approach brings the question​‌ of incrementally building such​​ a system, with a​​​‌ clear inspiration from developmental​ sciences. In this perspective,​‌ modules can generate internal​​ signals corresponding to internal​​​‌ goals, predictions, error signals,​ able to supervise the​‌ learning of other modules​​ (possibly endowed with a​​​‌ different learning rule), supposed​ to become autonomous after​‌ an instructing period. A​​ typical example is that​​​‌ of episodic learning (in​ the hippocampus), storing declarative​‌ memory about a collection​​ of past episods and​​​‌ supervising the training of​ a procedural memory in​‌ the cortex.

At the​​ behavioral level, as mentioned​​ above, our systemic approach​​​‌ underlines the fundamental links‌ between the adaptive system‌​‌ and the internal and​​ external world. The internal​​​‌ world includes proprioception and‌ interoception, giving information about‌​‌ the body and its​​ needs for integrity and​​​‌ other fundamental programs. The‌ external world includes physical‌​‌ laws that have to​​ be learned and possibly​​​‌ intelligent agents for more‌ complex interactions. Both involve‌​‌ sensors and actuators that​​ are the interfaces with​​​‌ these worlds and close‌ the loops. Within this‌​‌ rich picture, machine learning​​ generally selects one situation​​​‌ that defines useful sensors‌ and actuators and a‌​‌ corpus with properly segmented​​ data and time, and​​​‌ builds a specific architecture‌ and its corresponding criteria‌​‌ to be satisfied. In​​ our approach however, the​​​‌ first question to be‌ raised is to discover‌​‌ what is the goal,​​ where attention must be​​​‌ focused on and which‌ previous skills must be‌​‌ exploited, with the help​​ of a dynamic architecture​​​‌ and possibly other partners.‌ In this domain, the‌​‌ behavioral and the developmental​​ sciences, observing how and​​​‌ along which stages an‌ agent learns, are of‌​‌ great help to bring​​ some structure to this​​​‌ high dimensional problem.

At‌ the implementation level, this‌​‌ analysis opens many fundamental​​ challenges, hardly considered in​​​‌ machine learning : stability‌ must be preserved despite‌​‌ on-line continuous learning; criteria​​ to be satisfied often​​​‌ refer to behavioral and‌ global measurements but they‌​‌ must be translated to​​ control the local circuit​​​‌ level; in an incremental‌ or developmental approach, how‌​‌ will the development of​​ new functions preserve the​​​‌ integrity and stability of‌ others? In addition, this‌​‌ continous re-arrangement is supposed​​ to involve several kinds​​​‌ of learning, at different‌ time scales (from msec‌​‌ to years in humans)​​ and to interfer with​​​‌ other phenomena like variability‌ and meta-plasticity.

In summary,‌​‌ our main objective in​​ machine learning is to​​​‌ propose on-line learning systems,‌ where several modes of‌​‌ learning have to collaborate​​ and where the protocoles​​​‌ of training are realistic.‌ We promote here a‌​‌ really autonomous learning, where​​ the agent must select​​​‌ by itself internal resources‌ (and build them if‌​‌ not available) to evolve​​ at the best in​​​‌ an unknown world, without‌ the help of any‌​‌ deus-ex-machina to define parameters,​​ build corpus and define​​​‌ training sessions, as it‌ is generally the case‌​‌ in machine learning. To​​ that end, autonomous robotics​​​‌ (cf. § 3.4‌) is a perfect‌​‌ testbed.

3.4 Autonomous Robotics​​

Autonomous robots are not​​​‌ only convenient platforms to‌ implement our algorithms; the‌​‌ choice of such platforms​​ is also motivated by​​​‌ theories in cognitive science‌ and neuroscience indicating that‌​‌ cognition emerges from interactions​​ of the body in​​​‌ direct loops with the‌ world (embodiment of‌​‌ cognition49). In​​ addition to real robotic​​​‌ platforms, software implementations of‌ autonomous robotic systems including‌​‌ components dedicated to their​​ body and their environment​​​‌ will be also possibly‌ exploited, considering that they‌​‌ are also a tool​​ for studying conditions for​​​‌ a real autonomous learning.‌

A real autonomy can‌​‌ be obtained only if​​​‌ the robot is able​ to define its goal​‌ by itself, without the​​ specification of any high​​​‌ level and abstract cost​ function or rewarding state.​‌ To ensure such a​​ capability, we propose to​​​‌ endow the robot with​ an artificial physiology, corresponding​‌ to perceiving some kind​​ of pain and pleasure.​​​‌ It may consequently discriminate​ internal and external goals​‌ (or situations to be​​ avoided). This will mimick​​​‌ circuits related to fundamental​ needs (e.g. hunger and​‌ thirst) and to the​​ preservation of bodily integrity.​​​‌ An important objective is​ to show that more​‌ abstract planning capabilities can​​ arise from these basic​​​‌ goals.

A real autonomy​ with an on-line continuous​‌ learning as described in​​  § 3.3 will be​​​‌ made possible by the​ elaboration of protocols of​‌ learning, as it is​​ the case, in animal​​​‌ conditioning, for experimental studies​ where performance on a​‌ task can be obtained​​ only after a shaping​​​‌ in increasingly complex tasks.​ Similarly, developmental sciences can​‌ teach us about the​​ ordered elaboration of skills​​​‌ and their association in​ more complex schemes. An​‌ important challenge here is​​ to translate these hints​​​‌ at the level of​ the cerebral architecture.

As​‌ a whole, autonomous robotics​​ provide a way to​​​‌ assess the consistency of​ our models in realistic​‌ conditions of use and​​ offer our colleagues in​​​‌ behavioral sciences an object​ of study and comparison,​‌ regarding behavioral dynamics emerging​​ from interactions with the​​​‌ environment, also observable at​ the neuronal level.

In​‌ summary, our main contribution​​ in autonomous robotics is​​​‌ to make autonomy possible,​ by various means corresponding​‌ to endow robots with​​ an artificial physiology, to​​​‌ give instructions in a​ natural and incremental way​‌ and to prioritize the​​ synergy between reactive and​​​‌ robust schemes over complex​ planning structures.

4.1 Overview

Modeling​​ the brain to emulate​​​‌ cognitive functions offers direct​ and indirect application domains.​‌ Our models are designed​​ to be confronted to​​​‌ the reality of life​ sciences and to make​‌ predictions in neuroscience and​​ in the medical domain.​​​‌ Our models also have​ an impact in digital​‌ sciences; their performances can​​ be questioned in informatics,​​​‌ their algorithms can be​ compared with models in​‌ machine learning and artificial​​ intelligence, their behavior can​​​‌ be analysed in human-robot​ interaction. But since what​‌ they produce is related​​ to human thinking and​​​‌ behavior, applications will be​ also possible in various​‌ domains of social sciences​​ and humanities.

4.2 Applications​​​‌ in life sciences

One​ of the most original​‌ specificity of our team​​ is that it is​​​‌ part of a laboratory​ in Neuroscience (with a​‌ large spectrum of activity​​ from the molecule to​​​‌ the behavior), focused on​ neurodegenerative diseases and consequently​‌ working in tight collaboration​​ with the medical domain.​​​‌ Beyond data and signal​ analysis where our expertise​‌ in machine learning may​​ be possibly useful, our​​​‌ interactions are mainly centered​ on the exploitation of​‌ our models. They will​​ be classically regarded as​​​‌ a way to validate​ biological assumptions and to​‌ generate new hypotheses to​​ be investigated in the​​ living. Our macroscopic models​​​‌ and their implementation in‌ autonomous robots will allow‌​‌ an analysis at the​​ behavioral level and will​​​‌ propose a systemic framework,‌ the interpretation of which‌​‌ will meet aetiological analysis​​ in the medical domain​​​‌ and interpretation of intelligent‌ behavior in cognitive neuroscience‌​‌ and related domains like​​ for example educational science.​​​‌

The study of neurodegenerative‌ diseases is targeted because‌​‌ they match the phenomena​​ we model. Particularly, the​​​‌ Parkinson disease results from‌ the death of dopaminergic‌​‌ cells in the basal​​ ganglia, one of the​​​‌ main systems that we‌ are modeling. The Alzheimer‌​‌ disease also results from​​ the loss of neurons,​​​‌ in several cortical and‌ extracortical regions. The variety‌​‌ of these regions, together​​ with large mnesic and​​​‌ cognitive deficits, require a‌ systemic view of the‌​‌ cerebral architecture and associated​​ functions, very consistent with​​​‌ our approach.

4.3 Application‌ in digital sciences

Of‌​‌ course, digital sciences are​​ also impacted by our​​​‌ researches, at several levels.‌ At a global level,‌​‌ we will propose new​​ control architectures aimed at​​​‌ providing a higher degree‌ of autonomy to robots,‌​‌ as well as machine​​ learning algorithms working in​​​‌ more realistic environment. More‌ specifically, our focus on‌​‌ some cognitive functions in​​ closed loop with a​​​‌ real environment will address‌ currently open problems. This‌​‌ is obviously the case​​ for planning and decision​​​‌ making; this is particularly‌ the case for the‌​‌ domain of affective computing,​​ since motivational characteristics arising​​​‌ from the design of‌ an artificial physiology allow‌​‌ to consider not only​​ cold rational cognition but​​​‌ also hot emotional cognition.‌ The association of both‌​‌ kinds of cognition is​​ undoubtly an innovative way​​​‌ to create more realistic‌ intelligent systems but also‌​‌ to elaborate more natural​​ interfaces between these systems​​​‌ and human users.

At‌ last, we think that‌​‌ our activities in well-founded​​ distributed computations and high​​​‌ performance computing are not‌ just intended to help‌​‌ us design large scale​​ systems. We also think​​​‌ that we are working‌ here at the core‌​‌ of informatics and, accordingly,​​ that we could transfer​​​‌ some fundamental results in‌ this domain.

4.4 Applications‌​‌ in human sciences

Because​​ we model specific aspects​​​‌ of cognition such as‌ learning, language and decision,‌​‌ our models could be​​ directly analysed from the​​​‌ perspective of educational sciences,‌ linguistics, economy, philosophy and‌​‌ ethics.

Futhermore, our implication​​ in science outreach actions,​​​‌ including computer science teaching‌ in secondary and primary‌​‌ school, with the will​​ to analyse and evaluate​​​‌ the outcomes of these‌ actions, is at the‌​‌ origin of building a​​ link between our research​​​‌ in computational learning and‌ human learning, providing not‌​‌ only tools but also​​ new modeling paradigms.

5.1 Footprint of research‌​‌ activities

As part of​​ the Institute of Neurodegenerative​​​‌ Diseases that developed a‌ strong commitment to the‌​‌ environment, we take our​​ share in the reduction​​​‌ of our carbon footprint‌ by deciding to reduce‌​‌ our commuting footprint and​​ the number of yearly​​​‌ travels to conference.

7.1 Latest software​ developments

7.2 Open data‌

N.Rougier has been nominated‌​‌ as the representant for​​ Open Science for the​​​‌ Inria Bordeaux Center and‌ is part of the‌​‌ Software College of the​​ "Comité pour la Science​​​‌ Ouverte".

8.1 Overview

This year‌​‌ we have addressed several​​ important questions related to​​​‌ our scientific positioning. Central‌ to this positioning, we‌​‌ have studied and modeled​​ bio-inspired learning mechanisms and​​​‌ collaborative mnesic functions (‌cf. § 8.2).‌​‌ We have extended our​​ activities of exploration of​​​‌ higher cognitive functions, also‌ called Metacognition (cf.‌​‌ § 8.3) and​​ have considered how important​​​‌ characteristics can be associated‌ to this framework, like‌​‌ symbolic abstract knowledge (​​​‌cf. § 8.4),​ and oscillations (cf.​‌ § 8.5). Endly,​​ we have pursued our​​​‌ work on language processing​ in birds and robots​‌ (cf. § 8.6​​).

8.2 Decision, learning​​​‌ and collaborative mnesic functions​

Participants: Nicolas Rougier,​‌ Xavier Hinaut.

A​​ prominent view of basal​​​‌ ganglia (BG)/cortical interactions relates​ to a form of​‌ reinforcement learning (RL), in​​ which dopamine influx in​​​‌ the BG signals performance​ and approximates a gradient​‌ descent over many trials.​​ However, when applied to​​​‌ complex, continuous and embodied​ sensorimotor tasks, such gradient-based​‌ RL faces major limitations.​​ Extending the PhD work​​​‌ of Remy Sankar, we​ have thus explored an​‌ alternative based on the​​ songbird’s dual-pathway circuitry, where​​​‌ a cortical-like motor pathway​ and a BG-thalamo-cortical circuit​‌ interact during learning, with​​ structured variability, pathway-specific plasticity,​​​‌ and delayed maturation, providing​ a substrate for guided​‌ exploration and consolidation.

We​​ have also extended the​​​‌ PhD work of Naomi​ Chaix-Eichel and investigated the​‌ nature of splitter cells.​​ More specifically, we investigate​​​‌ whether a random recurrent​ structure is sufficient to​‌ allow latent sequences to​​ appear. To do so,​​​‌ we simulated an agent​ with egocentric sensory inputs​‌ that must navigate and​​ alternate choices at intersections.​​​‌ We were subsequently able​ to identify several splitter​‌ cells inside the model.​​ Remarkably, when we systematically​​​‌ lesioned the identified splitter​ cells, the model’s behavioral​‌ performance remained intact, and​​ new splitter cells consistently​​​‌ emerged through network reorganization.​

We further explored the​‌ role of random recurrent​​ architectures in supporting distributed​​​‌ and adaptive mnemonic functions.​ With collaborators at Standford​‌ Medecine 31, we​​ showed that reservoir models​​​‌ could reveal how short-term​ synaptic depression unifies the​‌ generation of MMN (Mismatch​​ Negativity) and P300 event-related​​​‌ potentials, providing a mechanistic​ link between neural dynamics​‌ and cognitive attention. Further,​​ with collaborators 28,​​​‌ we emphasized the "less​ is more" principle in​‌ natural intelligence, showing how​​ biological constraints foster efficiency,​​​‌ with reservoir computing enabling​ rapid learning from sparse​‌ data. Complementing this, 16​​ introduced ReMi, a low-power,​​​‌ data-free music generation approach​ using randomly initialized RNNs,​‌ highlighting the creative potential​​ of minimalistic architectures.

Finally,​​​‌ within the Défi Inria​ LLM4Code project , we​‌ explored how LLMs (Large​​ Language Models) can help​​​‌ to build generic structures​ (thanks to code generation)​‌ that drive autonomous agents.​​ This bridges the emerging​​​‌ field of Software Engineering​ Agents (SWE-Agent) with agents​‌ in embodied control tasks,​​ demonstrating the role of​​​‌ information access in problem-solving​ in simulated environments 26​‌. On a side​​ path, we explored how​​​‌ to enhance an LLM​ and optimised it for​‌ answering Reservoir Computing questions​​ and code using ReservoirPy​​​‌ by building a RAG​ (Retrieved-Augmented Generation) 25:​‌ we obtained better performances​​ than commercial and open​​​‌ source models for advanced​ coding questions.

8.3 Metacognition​‌

Participants: Frédéric Alexandre,​​ Chloé Mercier, Thierry​​​‌ Viéville.

In the​ doctoral work of Lucie​‌ Fontaine and also associated​​ with our associate team​​​‌ MetaBrain (cf. section 9.1.1​), we are studying​‌ the cerebral circuitry underlying​​ fundamental mechanisms of metacognition.​​ More precisely, we are​​​‌ considering the association of‌ a cortical model, built‌​‌ in a predictive coding​​ framework, with an hippocampal​​​‌ model 17, in‌ order to study in‌​‌ more details the mechanisms​​ of the Complementary Learning​​​‌ Systems theory, combining recall,‌ replay and consolidation.

We‌​‌ have also evaluated the​​ level of flexibility and​​​‌ other metacognitive properties in‌ classical generative AI models‌​‌ 13. We have​​ explained in a position​​​‌ paper 18 why episodic‌ memory is an important‌​‌ mechanism to integrate with​​ generative AI.

In the​​​‌ doctoral work of Baptiste‌ Pesquet and also associated‌​‌ with the ANR project​​ Courrier (cf section 9.3.2​​​‌), we have set‌ the bases of a‌​‌ model of metacognitive evaluation,​​ namely confidence, based on​​​‌ accumulation of evidence 19‌.

8.4 Integrating abstract‌​‌ symbolic knowledge

Participants: Frédéric​​ Alexandre, Chloé Mercier​​​‌, Margarida Romero,‌ Thierry Viéville.

As‌​‌ a follow-up to our​​ previous work, proposing to​​​‌ map an ontology onto‌ a Vector Symbolic Architecture‌​‌ (VSA) with a partial​​ implementation into spiking neural​​​‌ networks, we have finalized‌ modeling such a mesoscopic‌​‌ process at a macroscopic​​ scale. This formalism allows​​​‌ to integrate abstract symbolic‌ knowledge in biologically plausible‌​‌ mechanisms considering more complex​​ neural architectures, beyond what​​​‌ is possible at the‌ implementation level with high‌​‌ dimensional vector calculus.

We​​ have also addressed a​​​‌ more formal work on‌ manipulating symbolic knowledge equipped‌​‌ with a metric and​​ on applying this formalism​​​‌ to complex ill-defined problem-solving,‌ allowing to explicitly introduce‌​‌ symbolic knowledge in usal​​ machine learning numerical algorithms.​​​‌

Both issues are in‌ progress with journal articles‌​‌ in review.

8.5 Integrating​​ oscillations

In 2025, as​​​‌ part of an open‌ and reproducible science effort,‌​‌ we have published an​​ article in Rescience C​​​‌ as a follow-up of‌ the work of Mathilde‌​‌ Reynes during her Master​​ internship 14, reproducing​​​‌ a model of the‌ cortex and thalamus, originally‌​‌ developed in 2002 and​​ foundational to many subsequent​​​‌ studies on thalamocortical oscillations,‌ using a modern programming‌​‌ language and modeling paradigm.​​ Our work demonstrates that​​​‌ some initial results were‌ due to programming errors‌​‌ and provides recommendations for​​ improved reproducibility.

At the​​​‌ microscopic scale, as part‌ of the PhD project‌​‌ of Maeva Andriantsoamberomanga, we​​ have finished elaborating our​​​‌ multicompartment model of the‌ hippocampus capable of reproducing‌​‌ theta-nested gamma oscillations, and​​ are currently investigating the​​​‌ effects of extracellular electrical‌ stimulation on hippocampal oscillations.‌​‌ This work was presented​​ at the OCNS conference​​​‌ in July (32‌), and a journal‌​‌ article is currently in​​ preparation.

It should be​​​‌ noted that this year's‌ work on oscillations was‌​‌ initially thought to be​​ part of the new​​​‌ Inria project-team NeuroDTx instead‌ of Mnemosyne. However, even‌​‌ though the initiation of​​ NeuroDTx was encouraged by​​​‌ Inria and CNRS and‌ its project proposal approved‌​‌ about 18 months ago,​​ the team still hasn't​​​‌ been officially created due‌ to CNRS and Inria‌​‌ failing to sign an​​ agreement on intellectual property.​​​‌ This regrettable situation leads‌ to unnecessary complexity regarding‌​‌ budgeting, reduced coherence in​​​‌ Mnemosyne's research project and​ reduced visibility for the​‌ new axis of NeuroDTx,​​ and we sincerely hope​​​‌ it can be resolved​ in 2026.

8.6 Language​‌ processing

In order to​​ bridge the gaps between​​​‌ data-scarce and data-hungry models,​ within the AEx BrainGPT​‌ project (cf. section 9.3.3​​), we started to​​​‌ develop hybrid architectures such​ as Reservoir-Transformer models. While​‌ Transformers are powerful, they​​ exhibit quadratic complexity and​​​‌ lack biological plausibility in​ modeling cognitive functions like​‌ working memory. To address​​ this, we introduce Echo​​​‌ State Transformers (EST), a​ hybrid architecture combining Transformer​‌ attention with Reservoir Computing​​ 33, using trainable​​​‌ reservoirs as dynamic working​ memory units that enable​‌ constant-step complexity. Evaluated on​​ the Time Series Library,​​​‌ EST achieves state-of-the-art performance​ in classification and anomaly​‌ detection, ranking first in​​ two out of five​​​‌ categories, while offering scalable​ and biologically inspired alternatives​‌ to standard Transformers. This​​ direction sets us in​​​‌ relation to the research​ on alternative models to​‌ Transformers (MAMBA, State Space​​ Models, ...). On-going research​​​‌ explores how such hybrid​ models scale to big​‌ language data corpora. Moreover,​​ in the context of​​​‌ studying parallels between human​ language acquisition and songbird​‌ developmental learning, we analysed​​ song syntax in a​​​‌ new way 15:​ we applied subword tokenization​‌ methods, commonly used in​​ NLP, to identify meaningful​​​‌ chunks in canary song​ sequences. We compared these​‌ data-driven segmentations with our​​ expert annotations and found​​​‌ significant alignment. Ongoing analyses​ aim to relate discovered​‌ chunks to neural activity​​ in premotor areas like​​​‌ HVC. This approach not​ only provides a novel​‌ tool for studying birdsong​​ structure but also offers​​​‌ insights into the emergence​ of hierarchical organization in​‌ vocal learning.

9.1 International​​​‌ initiatives

9.2 European​ initiatives

9.3 National initiatives

9.4 Regional initiatives

9.5 Public policy​​​‌ support

Participants: Frederic Alexandre​, Xavier Hinaut,​‌ Nicolas Rougier, Thierry​​ Viéville.

We had​​​‌ some activities related to​ several ministries (in addition​‌ to the ministry of​​ research):

• Health: related to​​​‌ Covid hospitalization forecasting 30​,
• Justice: development with​‌ the department of law​​ of U. Bordeaux of​​​‌ the Observatory of Surveillance​ in Democracy, cf. section​‌ 9.4.1,
• Education: T.​​ Viéville, expert for the​​​‌ OECD about AI Literacy​ Framework for Primary and​‌ Secondary Education,
• Defence: several​​ actions of training and​​​‌ prospective for the ministry​ of Defence.

10.1 Promoting scientific activities​​

10.2​ Teaching - Supervision -​‌ Juries - Educational and​​ pedagogical outreach

Many courses​​​‌ are given in french​ universities and schools of​‌ engineers at different levels​​ (LMD) by most team​​​‌ members, in computer science,​ in applied mathematics, in​‌ neuroscience and in cognitive​​ science.

F. Alexandre and​​​‌ T. Viéville have been​ involved in the animation​‌ and online coaching of​​ the "Intelligence Artificielle Intelligente"​​​‌ citizen formation, via​ the creation of a​‌ MOOC, with more than​​ 80,000 participants, allowing everyone​​​‌ to master these disruptive​ technologies by better understanding​‌ ground notions.

T. Vieville​​ is part of teachers​​​‌ and education policy makers​ trainings regarding artificial intelligence,​‌ and is an expert​​ for the OECD about​​​‌ AI Literacy Framework for​ Primary and Secondary Education.​‌

10.3 Popularization

T. Viéville​​ has co-organized and participated​​​‌ at large scale popularization​ actions (more than 500​‌ children impacted) targeting underprivileged​​ educational areas, proposing educational​​​‌ robotics activities in application​ of the previous multidisciplinary​‌ collaborations with learning science​​ research.

T. Viéville has​​​‌ been invited for high-school​ interactive and participative conferences​‌ to explain artificial intelligence​​ and computational thinking (10​​​‌ interventions).

A. Aussel has​ been invited to multiple​‌ high-schools as part of​​ the "Un Scientifique, une​​​‌ classe : Chiche !"​ program. She has participated​‌ in the Circuit Scientifique​​ Bordelais as well as​​​‌ the "Moi Informaticienne, Moi​ Mathématicienne" program.

N.Rougier has​‌ been involved in more​​ than a dozen popularization​​​‌ events in 2025, ranging​ from invited talks, round​‌ tables, interviews, podcast and​​ animations.

X. Hinaut has​​​‌ been involved in several​ popularization events in 2025:​‌ Performances Art & Science​​ events (Jun25, May25, Mar25),​​​‌ "Fête de la Science"​ (Oct25, Cap Sciences, Bdx),​‌ Scientific Circuit "Hors les​​ Murs" (Oct25, Lycée Aiguillon,​​​‌ Lot), Open debate "Rencard​ du savoir « IA​‌ : un état de​​ l’art » (Mar25, Médiathèque​​​‌ Gradignan), "Café IA" (Fev25,​ Le Node, Bdx).

11.1 Major publications​​

• 1 articleF.Frédéric​​​‌ Alexandre. A global​ framework for a systemic​‌ view of brain modeling​​.Brain Informatics8​​​‌1February 2021,​ 22HALDOI
• 2​‌ articleM.Mathieu Bourdenx​​, A.Aurélien Nioche​​​‌, S.Sandra Dovero​, M.-L.Marie-Laure Arotcarena​‌, S. M.Sandrine​​ M. J. Camus,​​​‌ G.Gregory Porras,​ M.-L.Marie-Laure Thiolat,​‌ N. P.Nicolas P.​​ Rougier, A.Alice​​​‌ Prigent, P.Philippe​ Aubert, S.Sylvain​‌ Bohic, C.Christophe​​ Sandt, F.Florent​​​‌ Laferrière, E.Evelyne​ Doudnikoff, N.Niels​‌ Kruse, B.Brit​​ Mollenhauer, S.Salvatore​​​‌ Novello, M.Michele​ Morari, T.Thierry​‌ Leste-Lasserre, I.Ines​​ Trigo Damas, M.​​​‌Michel Goillandeau, C.​Celine Perier, C.​‌Cristina Estrada, N.​​Nuria García Carrillo,​​​‌ A.Ariadna Recasens,​ N. N.Nishant Narayanan​‌ Vaikath, O.Omar​​ El Agnaf, M.​​​‌ T.Maria Trinidad Herrero​, P.Pascal Derkinderen​‌, M.Miquel Vila​​, J. A.Jose​​​‌ A Obeso, B.​Benjamin Dehay and E.​‌Erwan Bezard. Identification​​ of distinct pathological signatures​​​‌ induced by patient-derived $$$$ -synuclein​ structures in nonhuman primates​‌.Science Advances 6​​20May 2020,​​​‌ eaaz9165HALDOI
• 3​ inproceedingsT.Thomas Ferté​‌, D.Dan Dutartre​​, B. P.Boris​​​‌ P. Hejblum, R.​Romain Griffier, V.​‌Vianney Jouhet, R.​​Rodolphe Thiébaut, P.​​​‌Pierrick Legrand and X.​Xavier Hinaut. Reservoir​‌ Computing for Short High-Dimensional​​ Time Series: an Application​​​‌ to SARS-CoV-2 Hospitalization Forecast​.ICML'24: Proceedings of​‌ the 41st International Conference​​ on Machine Learning235​​​‌Proceedings of Machine Learning​ ResearchVienna, AustriaJuly​‌ 2024, 13570--13591HAL​​DOI
• 4 articleA.​​​‌Aurélien Nioche, N.​ P.Nicolas P. Rougier​‌, M.Marc Deffains​​, S.Sacha Bourgeois-Gironde​​​‌, S.Sébastien Ballesta​ and T.Thomas Boraud​‌. The adaptive value​​ of probability distortion and​​​‌ risk-seeking in macaques' decision-making​.Philosophical Transactions of​‌ the Royal Society B:​​ Biological SciencesJanuary 2021​​​‌HALDOI
• 5 article​S.Silvia Pagliarini,​‌ A.Arthur Leblois and​​ X.Xavier Hinaut.​​​‌ Vocal Imitation in Sensorimotor​ Learning Models: a Comparative​‌ Review.IEEE Transactions​​ on Cognitive and Developmental​​​‌ SystemsNovember 2020HAL​DOI
• 6 articleL.​‌Luca Pedrelli and X.​​Xavier Hinaut. Hierarchical-Task​​​‌ Reservoir for Online Semantic​ Analysis from Continuous Speech​‌.IEEE Transactions on​​ Neural Networks and Learning​​​‌ SystemsSeptember 2021HAL​DOI
• 7 articleN.​‌ .Nicolas P. Rougier​​ and G. I.Georgios​​​‌ Is Detorakis. Randomized​ Self Organizing Map.​‌Neural Computation2021HAL​​
• 8 bookN. .​​Nicolas P. Rougier.​​​‌ Scientific Visualization: Python +‌ Matplotlib.November 2021‌​‌HAL
• 9 articleR.​​Remya Sankar, N.​​​‌ .Nicolas P. Rougier‌ and A.Arthur Leblois‌​‌. Computational benefits of​​ structural plasticity, illustrated in​​​‌ songbirds.Neuroscience &‌ Biobehavioral Reviews2021HAL‌​‌
• 10 articleA.Anthony​​ Strock, X.Xavier​​​‌ Hinaut and N. P.‌Nicolas P. Rougier.‌​‌ A Robust Model of​​ Gated Working Memory.​​​‌Neural ComputationNovember 2019‌, 1-29HALDOI‌​‌

11.2 Publications of the​​ year

11.3 Cited‌​‌ publications

• 40 articleS.​​S .Amari. Dynamic​​​‌ of pattern formation in‌ lateral-inhibition type neural fields‌​‌.Biological Cybernetics27​​1977, 77--88back​​​‌ to text
• 41 book‌D.D.P .Bertsekas and‌​‌ J.J.N .Tsitsiklis.​​ Parallel and Distributed Computation:​​​‌ Numerical Methods.Athena‌ Scientific1997back to‌​‌ text
• 42 articleR.​​R .Brette, M.​​​‌M .Rudolph, T.‌T .Carnevale, M.‌​‌M .Hines, D.​​D .Beeman, J.​​​‌ ..J.M . Bower‌, M.M .Diesmann‌​‌, A.A. Morrison​​​‌, P. H.P.​ H. Goodman, F.​‌ C.F. C. Jr.​​ Harris, M.M.​​​‌ Zirpe, T.T​ .Natschläger, D.D​‌ .Pecevski, B.B​​ .Ermentrout, M.M​​​‌ .Djurfeldt, A.A​ .Lansner, O.O​‌ .Rochel, T.T​​ .Viéville, E.E​​​‌ .Muller, A.A.P​ .Davison, S. ..​‌S .El Boustani and​​ A.A .Destexhe.​​​‌ Simulation of networks of​ spiking neurons: a review​‌ of tools and strategies​​.Journal of Computational​​​‌ Neuroscience2332007​, 349--398back to​‌ text
• 43 articleS.​​S .Coombes. Waves,​​​‌ bumps and patterns in​ neural field theories.​‌Biol. Cybern.932005​​, 91-108back to​​​‌ text
• 44 bookP.​P .Dayan and L.​‌L.F .Abbott. Theoretical​​ Neuroscience : Computational and​​​‌ Mathematical Modeling of Neural​ Systems.MIT Press​‌2001back to text​​
• 45 bookW.W​​​‌ .Gerstner and W.W.M​ .Kistler. Spiking Neuron​‌ Models: Single Neurons, Populations,​​ Plasticity.Cambridge University​​​‌ PressCambridge University Press​2002back to text​‌
• 46 articleD.D​​ .Mitra. Asynchronous relaxations​​​‌ for the numerical solution​ of differential equations by​‌ parallel processors.SIAM​​ J. Sci. Stat. Comput.​​​‌811987,​ 43--58back to text​‌
• 47 bookR.R.C​​ .O'Reilly and Y.Y​​​‌ .Munakata. Computational Explorations​ in Cognitive Neuroscience: Understanding​‌ the Mind by Simulating​​ the Brain.Cambridge,​​​‌ MA, USAMIT Press​2000back to text​‌
• 48 inproceedingsF.Frédéric​​ Alexandre. Biological Inspiration​​​‌ for Multiple Memories Implementation​ and Cooperation.International​‌ Conference on Computational Intelligence​​2000back to text​​​‌
• 49 articleD. H.​Dana H. Ballard,​‌ M. M.Mary M.​​ Hayhoe, P. K.​​​‌Polly K. Pook and​ R. P.Rajesh P.​‌ N. Rao. Deictic​​ codes for the embodiment​​​‌ of cognition.Behavioral​ and Brain Sciences20​‌041997, 723--742​​URL: http://dx.doi.org/10.1017/S0140525X97001611back to​​​‌ text
• 50 articleK.​Kenji Doya. What​‌ are the computations of​​ the cerebellum, the basal​​​‌ ganglia and the cerebral​ cortex?Neural Networks12​‌1999, 961--974back​​ to text
• 51 article​​​‌J.Jérémy Fix,​ N. P.Nicolas P.​‌ Rougier and F.Frédéric​​ Alexandre. A dynamic​​​‌ neural field approach to​ the covert and overt​‌ deployment of spatial attention​​.Cognitive Computation3​​​‌12011, 279-293​URL: http://hal.inria.fr/inria-00536374/enDOIback​‌ to text
• 52 book​​T.Tom Mitchell.​​​‌ Machine Learning.Mac​ Graw-Hill Press1997back​‌ to text
• 53 article​​N. P.Nicolas P.​​​‌ Rougier. Dynamic Neural​ Field with Local Inhibition​‌.Biological Cybernetics94​​32006, 169-179​​​‌back to text
• 54​ articleN. P.Nicolas​‌ P. Rougier and A.​​Axel Hutt. Synchronous​​​‌ and Asynchronous Evaluation of​ Dynamic Neural Fields.​‌J. Diff. Eq. Appl.​​2009back to text​​​‌back to text
• 55​ articleL.L.R. Squire​‌. Memory systems of​​ the brain: a brief​​​‌ history and current perspective​.Neurobiol. Learn. Mem.​‌822004, 171-177​​back to text
• 56​​ articleW.Wahiba Taouali​​​‌, T.Thierry Viéville‌, N. P.Nicolas‌​‌ P. Rougier and F.​​Frédéric Alexandre. No​​​‌ clock to rule them‌ all.Journal of‌​‌ Physiology1051-32011​​, 83-90back to​​​‌ text
• 57 bookT.‌T.P. Trappenberg. Fundamentals‌​‌ of Computational Neuroscience.​​Oxford University Press2002​​​‌back to text
• 58‌ articleT.Thierry Viéville‌​‌. An unbiased implementation​​ of regularization mechanisms.​​​‌Image and Vision Computing‌23112005,‌​‌ 981--998URL: http://authors.elsevier.com/sd/article/S0262885605000909back​​ to text