Acquisition du langage | EPFL Graph Search

L'acquisition du langage est un domaine de recherche pluridisciplinaire, relevant notamment de la recherche en psychologie et en sciences du langage, qui vise à décrire et comprendre comment l'enfant acquiert le langage, oral ou gestuel, du milieu qui l'entoure. L'acquisition du langage d'un jeune enfant est rapide et s'effectue sans apprentissage formel. Le langage se développe toute la vie, mais c'est surtout entre la naissance et l'âge de cinq ans, que les apprentissages essentiels sont observés. Durant cette période, les transformations de la communication verbale orale sont les plus remarquables tant en compréhension qu'en production. Il s'agit surtout d'un apprentissage implicite démarrant bien avant la scolarisation. Cet apprentissage procède par étapes, commençant par du babillage, pour passer au mot, puis à l'utilisation de suites de mots et à l'acquisition progressive des règles de grammaire. La scolarisation permet l'apprentissage du langage écrit et la prise de conscience

Theory of representation learning in cortical neural networks

Carlos Stein Naves de Brito

Our brain continuously self-organizes to construct and maintain an internal representation of the world based on the information arriving through sensory stimuli. Remarkably, cortical areas related to different sensory modalities appear to share the same functional unit, the neuron, and develop through the same learning mechanism, synaptic plasticity. It motivates the conjecture of a unifying theory to explain cortical representational learning across sensory modalities. In this thesis we present theories and computational models of learning and optimization in neural networks, postulating functional properties of synaptic plasticity that support the apparent universal learning capacity of cortical networks. In the past decades, a variety of theories and models have been proposed to describe receptive field formation in sensory areas. They include normative models such as sparse coding, and bottom-up models such as spike-timing dependent plasticity. We bring together candidate explanations by demonstrating that in fact a single principle is sufficient to explain receptive field development. First, we show that many representative models of sensory development are in fact implementing variations of a common principle: nonlinear Hebbian learning. Second, we reveal that nonlinear Hebbian learning is sufficient for receptive field formation through sensory inputs. A surprising result is that our findings are independent of specific details, and allow for robust predictions of the learned receptive fields. Thus nonlinear Hebbian learning and natural statistics can account for many aspects of receptive field formation across models and sensory modalities. The Hebbian learning theory substantiates that synaptic plasticity can be interpreted as an optimization procedure, implementing stochastic gradient descent. In stochastic gradient descent inputs arrive sequentially, as in sensory streams. However, individual data samples have very little information about the correct learning signal, and it becomes a fundamental problem to know how many samples are required for reliable synaptic changes. Through estimation theory, we develop a novel adaptive learning rate model, that adapts the magnitude of synaptic changes based on the statistics of the learning signal, enabling an optimal use of data samples. Our model has a simple implementation and demonstrates improved learning speed, making this a promising candidate for large artificial neural network applications. The model also makes predictions on how cortical plasticity may modulate synaptic plasticity for optimal learning. The optimal sampling size for reliable learning allows us to estimate optimal learning times for a given model. We apply this theory to derive analytical bounds on times for the optimization of synaptic connections. First, we show this optimization problem to have exponentially many saddle-nodes, which lead to small gradients and slow learning. Second, we show that the number of input synapses to a neuron modulates the magnitude of the initial gradient, determining the duration of learning. Our final result reveals that the learning duration increases supra-linearly with the number of synapses, suggesting an effective limit on synaptic connections and receptive field sizes in developing neural networks.

EPFL2016

The aging brain as a model to understand motor skill acquisition in humans and its restoration using non-invasive brain stimulation

Pablo Maceira Elvira

Motor skill acquisition is essential to our survival, as it enables an efficient interaction with the changing world around us. The acquisition of novel sequential tasks, of particular relevance due to their pervasiveness in everyday life activities, can become more challenging at an advanced age, with structural and functional changes in the brain often resulting in diminished learning abilities. During the pursuit of my doctoral degree, I studied some of the processes leading to the acquisition of a novel sequential motor task, and how some of the mechanisms underlying these processes differ between healthy young and older adults. Young adults acquired the sequential task effectively by prioritizing its accurate execution early in training and focusing on increasing their speed thereafter, resulting in the generation of mechanically efficient execution patterns in the replication of the sequence. In contrast, older adults improved both the accuracy and the speed simultaneously and gradually after more extensive practice, which resulted in an overall decreased performance of the task. However, anodal direct current stimulation applied over the motor cortex partially restored skill acquisition in older adults by facilitating the early improvement of the accuracy, enabling an accelerated generation of efficient execution patterns in the sequence, similar to those seen in young adults. Further investigations into the effects of stimulation to improve skill acquisition in healthy older adults showed age not to be a determinant factor for an individual's proneness to benefit from stimulation; rather, it is the state of the neural system of each individual what likely determines the potential benefits to be had from stimulation. Therefore, a better understanding of the mechanisms targeted by stimulation and the identification of parameters signaling an individual's likelihood to benefit from it are needed to use these techniques to their full potential. Leveraging this knowledge and taking advantage of the portability, robustness and accessibility of this technique could represent an attractive option for applications involving extensive motor training, such as rehabilitative training provided after stroke.

EPFL2022

Theory of representation learning in cortical neural networks

Carlos Stein Naves de Brito

EPFL2016

The aging brain as a model to understand motor skill acquisition in humans and its restoration using non-invasive brain stimulation

Pablo Maceira Elvira

EPFL2022