Simplicité | EPFL Graph Search

La simplicité (du latin simplicitas) est la propriété, la condition ou la qualité d'être simple, naturel, et non-combiné. Elle dénote souvent la beauté, pureté, clarté ou spontanéité. Les choses simples sont par définition souvent plus faciles à expliquer et à comprendre que les choses compliquées, la complexité étant l'opposé de la simplicité. La simplicité signifie aussi le fait d'être libéré des efforts et de la confusion, et des choses non indispensables. La simplicité en tant que mode de vie peut aussi référer à la simplicité volontaire. D'après le rasoir d'Occam, toutes les autres choses étant égales, la théorie la plus simple est la plus sûrement vraie, d'où l'importance du concept de simplicité en épistémologie. D'après Thomas d'Aquin, Dieu est Simplicité divine. Les membres de la Société religieuse des Amis (quakers) pratiquent le témoignage de simplicité, qui est la simplification de la vie d'un individu dans le but de se concentrer sur les choses les plus importantes et év

CauseOccam: Learning Interpretable Abstract Representations in Reinforcement Learning Environments via Model Sparsity

Sergei Volodin

"I choose this restaurant because they have vegan sandwiches" could be a typical explanation we would expect from a human. However, current Reinforcement Learning (RL) techniques are not able to provide such explanations, when trained on raw pixels. RL algorithms for state-of-the-art benchmark environments are based on neural networks, which lack interpretability, because of the very factor that makes them so versatile – they have many parameters and intermediate representations. Enforcing safety guarantees is important when deploying RL agents in the real world, and guarantees require interpretability of the agent. Humans use short explanations that capture only the essential parts and often contain few causes to explain an effect. In our thesis, we address the problem of making RL agents understandable by humans. In addition to the safety concerns, the quest to mimic human-like reasoning is of general scientific interest, as it sheds light on the easy problem of consciousness. The problem of providing interpretable and simple causal explanations of agent's behavior is connected to the problem of learning good state representations. If we lack such a representation, any reasoning algorithm's outputs would be useless for interpretability, since even the "referents" of the "thoughts" of such a system would be obscure to us. One way to define simplicity of causal explanations via the sparsity of the Causal Model that describes the environment: the causal graph has the fewest edges connecting causes to their effects. For example, a model for choosing the restaurant that only depends on the cause "vegan" is simpler and more interpretable than a model that looks at each pixel of a photo of the menu of a restaurant, and possibly relies as well on spurious correlations, such the style of the menu. In this thesis, we propose a framework "CauseOccam" for model-based Reinforcement Learning where the model is regularized for simplicity in terms of sparsity of the causal graph it corresponds to. The framework contains a learned mapping from observations to latent features, and a model predicting latent features at the next time-steps given ones from the current time-step. The latent features are regularized with the sparsity of the model, compared to a more traditional regularization on the features themselves, or via a hand-crafted interpretability loss. To achieve sparsity, we use discrete Bernoulli variables with gradient estimation, and to find the best parameters, we use the primal-dual constrained formulation to achieve a target model quality. The novelty of this work is in learning jointly a sparse causal graph and the representation taking pixels as the input on RL environments. We test this framework on benchmark environments with non-trivial high-dimensional dynamics and show that it can uncover the causal graph with the fewest edges in the latent space. We describe the implications of our work for developing priors enforcing interpretability.

2021

Exponential Natural Evolution Strategies

Tom Schaul, Daniël Pieter Wierstra

The family of natural evolution strategies (NES) offers a principled approach to real-valued evolutionary optimization by following the natural gradient of the expected fitness. Like the well-known CMA-ES, the most competitive algorithm in the field, NES comes with important invariance properties. In this paper, we introduce a number of elegant and efficient improvements of the basic NES algorithm. First, we propose to parameterize the positive definite covariance matrix using the exponential map, which allows the covariance matrix to be updated in a vector space. This new technique makes the algorithm completely invariant under linear transformations of the underlying search space, which was previously achieved only in the limit of small step sizes. Second, we compute all updates in the natural coordinate system, such that the natural gradient coincides with the vanilla gradient. This way we avoid the computation of the inverse Fisher information matrix, which is the main computational bottleneck of the original NES algorithm. Our new algorithm, exponential NES (xNES), is significantly simpler than its predecessors. We show that the various update rules in CMA-ES are closely related to the natural gradient updates of xNES. However, xNES is more principled than CMA-ES, as all the update rules needed for covariance matrix adaptation are derived from a single principle. We empirically assess the performance of the new algorithm on standard benchmark functions

2010

Adaptive Self-Sealing Suction-Based Soft Robotic Gripper

Metin Sitti, Sukho Song

While suction cups prevail as common gripping tools for a wide range of real-world parts and surfaces, they often fail to seal the contact interface when engaging with irregular shapes and textured surfaces. In this work, the authors propose a suction-based soft robotic gripper where suction is created inside a self-sealing, highly conformable and thin flat elastic membrane contacting a given part surface. Such soft gripper can self-adapt the size of its effective suction area with respect to the applied load. The elastomeric membrane covering edge of the soft gripper can develop an air-tight self-sealing with parts even smaller than the gripper diameter. Such gripper shows 4 times higher adhesion than the one without the membrane on various textured surfaces. The two major advantages, underactuated self-adaptability and enhanced suction performance, allow the membrane-based suction mechanism to grip various three-dimensional (3D) geometries and delicate parts, such as egg, lime, apple, and even hydrogels without noticeable damage, which can have not been gripped with the previous adhesive microstructures-based and active suction-based soft grippers. The structural and material simplicity of the proposed soft gripper design can have a broad use in diverse fields, such as digital manufacturing, robotic manipulation, transfer printing, and medical gripping.

WILEY2021

CauseOccam: Learning Interpretable Abstract Representations in Reinforcement Learning Environments via Model Sparsity

Sergei Volodin

2021