Intuition | EPFL Graph Search

L'intuition est un mode de connaissance, de pensée ou de jugement, conçu comme immédiat (au sens de direct) ; selon les acceptions, c'est un processus ou une faculté de l'esprit. Définie de plusieurs manières en philosophie ainsi qu'en psychologie, l’intuition serait le fait de pressentir ou comprendre quelque chose sans analyse ni raisonnement. L'intuition peut être supra-rationnelle ou infra-rationnelle. Son domaine est large : il concerne aussi bien la connaissance proprement dite (métaphysique ou représentation du monde) que les sentiments (sur les choses) ou les motivations (à agir). Elle est parfois dénommée, de façon plus familière, flair. Cette dernière appellation ne concerne que la connaissance infra-rationnelle. Étymologie Du latin intuitio, , dérivé de intueri . Fonction L'intuition semble être immédiate du fait qu'elle paraît opérer sans user de la raison, et est généralement perçue comme inconsciente : seule sa conclusion es

Learning strategies and representations for intuitive robot learning from demonstration

Thibaut Antoine Kulak

Robots are becoming more and more present around us, both in industries and in our homes. One key capability of robots is their adaptability to various situations that might appear in the real world. Robot skill learning is therefore a crucial aspect of robotics aiming to provide robots with programs enabling them to perform one or several tasks successfully. While such programming is usually done by an engineer or a developer, making robot programming available to anyone would dramatically increase the range of applications currently feasible for robots. Learning from Demonstration (LfD) is a robot skill learning paradigm addressing this aim by developing intuitive frameworks for non-expert users to easily (re)program robots. While Learning from Demonstration has emerged as a successful way to program robots, several limitations remain to be addressed. Typical approaches still require some forms of preprocessing, such as the alignment of the demonstrations, or the choice of the movement representation. Also, the algorithms have to run with a relatively low number of demonstrations that human users are typically willing to give, while being performant, adaptable and generalizable to new situations. In this thesis, we propose to address these shortcomings with methods that make Learning from Demonstration more intuitive and user-friendly. We notably propose a novel movement representation requiring no demonstration alignment, and active learning strategies that permit to learn complex skills from fewer demonstrations.

EPFL2021

Using reinforcement learning for optimizing the reproduction of tasks in robot programming by demonstration

Florent Günter

As robots start pervading human environments, the need for new interfaces that would simplify human-robot interaction has become more pressing. Robot Programming by Demonstration (RbD) develops intuitive ways of programming robots, taking inspiration in strategies used by humans to transmit knowledge to apprentices. The user-friendliness of RbD is meant to allow lay users with no prior knowledge in computer science, electronics or mechanics to train robots to accomplish tasks the same way as they would with a co-worker. When a trainer teaches a task to a robot, he/she shows a particular way of fulfilling the task. For a robot to be able to learn from observing the trainer, it must be able to learn what the task entails (i.e. answer the so-called "What-to-imitate?" question), by inferring the user's intentions. But most importantly, the robot must be able to adapt its own controller to fit at best the demonstration (the so-called "How-to-imitate?" question) despite different setups and embodiments. The latter is the question that interested us in this thesis. It relates to the problem of optimizing the reproduction of the task under environmental constraints. The "How-to-imitate?" question is subdivided into two problems. The first problem, also known as the "correspondence problem", relates to resolving the discrepancy between the human demonstrator and robot's body that prevent the robot from doing an identical reproduction of the task. Even though we helped ourselves by considering solely humanoid platforms, that is platforms that have a joint configuration similar to that of the human, discrepancies in the number of degrees of freedom and range of motion remained. We resolved these by exploiting the redundant information conveyed through the demonstrations by collecting data through different frames of reference. By exploiting these redundancies in an algorithm comparable to the damped least square algorithm, we are able to reproduce a trajectory that minimizes the error between the desired trajectory and the reproduced trajectory across each frame of reference. The second problem consists in reproducing a trajectory in an unknown setup while respecting the task constraints learned during training. When the information learned from the demonstration no longer suffice to generalize the task constraints to a new set-up, the robot must re-learn the task; this time through trial-and-error. Here we considered the combination of trial-and-error learning to complement RbD. By adding a trial-and-error module to the original Imitation Learning algorithm, the robot can find a solution that is more adapted to the context and to its embodiment than the solution found using RbD. Specifically, we compared Reinforcement Learning (RL) – to other classical optimization techniques. We show that the system is advantageous in that: a) learning is more robust to unexpected events that have not been encountered during the demonstrations and b) the robot is able to optimize its own model of the task according to its own embodiment.

EPFL2009

Portfolio construction under information asymmetry

Lionel Coulot

We introduce in this thesis the idea of a variable lookback model, i.e., a model whose predictions are based on a variable portion of the information set. We verify the intuition of this model in the context of experimental finance. We also propose a novel algorithm to estimate it, the variable lookback algorithm, and apply the latter to build investment strategies. Financial markets under information asymmetry are characterized by the presence of better-informed investors, also called insiders. The literature in finance has so far concentrated on theoretical models describing such markets, in particular on the role played by the price in conveying information from informed to uninformed investors. However, the implications of these theories have not yet been incorporated into processing methods to extract information from past prices and this is the aim of this thesis. More specifically, the presence of a time-varying number of insiders induces a time-varying predictability in the price process, which calls for models that use a variable lookback window. Moreover, although our initial motivation comes from the study of markets under information asymmetry, the problem is more general, as it touches several issues in statistical modeling. The first one concerns the structure of the model. Existing methods use a fixed model structure despite evidences from data, which support an adaptive one. The second one concerns the improper handling of the nonstationarity in data. The stationarity assumption facilitates the mathematical treatment. Hence, existing methods relies on some form of stationarity, for example, by assuming local stationary, as in the windowing approach, or by modeling the underlying switching process, for example, with a Markov chain of order 1. However, these suffer from certain limitations and more advanced methods that take explicitly into account the nonstationariry of the signal are desirable. In summary, there is a need to develop a method that constantly monitors what is the appropriate structure, when a certain model works and when not or when are the underlying assumptions of the model violated. We verify our initial intuition in the context of experimental finance. In particular, we highlight the diffusion of information in the market. We give a precise definition to the notion of the time of maximally informative price and verify, in line with existing theories, that the time of maximally informative price is inversely proportional to the number of insiders in the market. This supports the idea of a variable lookback model. Then, we develop an estimation algorithm that selects simultaneously the order of the process and the lookback window based on the minimum description length principle. The algorithm maintains a series of estimators, each based on a different order and/or information set. The selection is based on an information theoretic criterion, that accounts for the ability of the model to fit the data, penalized by the model complexity and the amount of switching between models. Finally, we put the algorithm at work and build investment strategies. We devise a method to draw dynamically the trend line for the time-series of log-prices and propose an adaptive version of the well-known momentum strategy. The latter outperforms standard benchmarks, in particular during the 2009 momentum crash.

EPFL2013