Apprentissage par problèmes | EPFL Graph Search

Dans l'apprentissage par problèmes (APP), ou apprentissage par résolution de problèmes, les apprenants, regroupés par équipes, travaillent ensemble à résoudre un problème généralement proposé par l'enseignant, problème pour lequel ils n'ont reçu aucune formation particulière, de façon à faire des apprentissages de contenu et de savoir-faire, à découvrir des notions nouvelles de façon active (il s’instruit lui-même) en y étant poussé par les nécessités du problème soumis. La tâche de l'équipe est habituellement d'expliquer les phénomènes sous-jacents au problème et de tenter de le résoudre dans un processus non linéaire. La démarche est guidée par l'enseignant qui joue un rôle de facilitateur. Le concept a été développé par la faculté de médecine de l’Université McMaster de Hamilton en Ontario dès 1970. Ceci est différent d’une étude de cas ou d’une résolution, en ce sens que l’on ne donne pas forcément de solution. Domaines et situations L'apprentissage par problèmes a é

Students' learning strategies, motivation and project-management skills during interdisciplinary projects in Covid times

Julien Delisle, Siara Ruth Isaac, Helena Kovacs, Marc Lafuente Martinez

In this study, we investigated an interdisciplinary project-based learning program. Students were confronted with interdisciplinary challenges in the form of complex and authentic problems, like building a racing car or making a mobile laboratory for genetic analysis. These hands-on educational formats had a clear aim to develop skills required in students’ future careers, as well as bridge the gap between theory and practice. In this paper, we show preliminary results of an on-going mixed method study where the students’ learning strategies, motivation, and project management skills were measured through a survey using a pre and post-test approach. Quantitative results were contrasted with qualitative input from interviews with projects’ coaches and students’ focus groups. The results provided evidence of gains regarding professional skills (e.g. risk assessment in projects), but also shed light on difficulties and needs to implement meaningful experiential learning content within engineering education (e.g. collaboration). On top of this, the research took place during COVID-19 lockdown, hence both students’ and coaches’ reflections accounted for ways in which this situation did impact the projects.

Technische Universität Berlin2021

Visual Augmentation of Deictic Gestures in MOOC Videos

Pierre Dillenbourg, Kshitij Sharma

We present an eye-tracking study to compare different modalities for visual augmentations of the teacher’s explicit deictic gestures on a video lecture. We compared three visualizations: 1) hand gestures with a pointer, 2) gaze overlay, and 3) no-augmentation baseline. We investigate the teacher-student pair in a video-based learning context as an abstraction of an expert-novice pair where the goal is to attain a high level of shared understanding. The key phase of having a shared understanding is to have a common ground between the pair. Previous studies showed that explicit deixis plays a major role in initiating and maintaining common ground. This led us to hypothesize that augmenting videos with teacher’s deictic gestures might help students perform better. We found that augmenting the video with teacher’s gaze results in higher learning gain than no visualization. Moreover, gaze visualization also helped students in maintaining longer attention spans than hand gestures.

2016

Model-based reinforcement learning and navigation in animals and machines

Dane Sterling Corneil

For decades, neuroscientists and psychologists have observed that animal performance on spatial navigation tasks suggests an internal learned map of the environment. More recently, map-based (or model-based) reinforcement learning has become a highly active research area in machine learning. With a learned model of their environment, both animals and artificial agents can generalize between tasks and learn rapidly. In this thesis, I present approaches for developing efficient model--based behaviour in machines and explaining model--based behaviour in animals. From a neuroscience perspective, I focus on the hippocampus, believed to be a major substrate of model-based behaviour in the brain. I consider how hippocampal connectivity enable path--finding between different locations in an environment. The model describes how environments with boundaries and barriers can be represented in recurrent neural networks (i.e. attractor networks), and how the transient activity in these networks, after being stimulated with a goal location, could be used for determining a path to the goal. I also propose how the connectivity of these map--like networks can be learned from the spatial firing patterns observed in the input pathway to the hippocampus (i.e. grid cells and border cells). From a machine learning perspective, I describe a reinforcement learning model that integrates model-based methods and "episodic control", an approach to reinforcement learning based on episodic memory. According to episodic control, the agent learns how to act in the environment by storing snapshot-like memories of its observations, then comparing its current observations to similar snapshot memories where it took an action that resulted in high reward. In our approach, the agent augments these real-world memories with episodes simulated offline using a learned model of the environment. These ``simulated memories'' allow the agent to adapt faster when the reward locations change. Next, I describe Variational State Tabulation (VaST), a model--based method for learning quickly with continuous and high-dimensional observations (like those found in 3D navigation tasks). The VaST agent learns to map its observations to a limited number of discrete abstract states, and build a transition model over those abstract states. The long--term values of different actions in each state are updated continuously and efficiently in the background as the agent explores the environment. I show how the VaST agent can learn faster than other state-of-the-art algorithms, even changing its policy after a single new experience, and how it can respond quickly to changing rewards in complex 3D environments. The models I present allow the agent to rapidly adapt to changing goals and rewards, a key component of intelligence. They use a combination of features attributed to model-based and episodic controllers, suggesting that the division between the two fields is not strict. I therefore also consider the consequences of these findings on theories of model-based learning, episodic control and hippocampal function.

EPFL2018

Model-based reinforcement learning and navigation in animals and machines

Dane Sterling Corneil

EPFL2018