arXiv subjects

Adrien Laversanne-Finot

Publications and source records attributed to Adrien Laversanne-Finot.

7 recordsLinked to original sources

arXiv2023

Defending against the nothing-at-stake problem in multi-threaded blockchains

In blockchain systems, the scarcity of a resource is used as a Sybil protection mechanism. In Proof-of-Work blockchains, that resource is computing power. In the event of a fork, the scarcity of this resource theoretically prevents miners from producing blocks on both branches of a fork. In Proof-of-Stake blockchains, because that resource is token stake, the computational cost of creating a block is negligible. In the event of a fork, and if no specific measures have been taken, rational block producers should extend both branches of the fork. In blockchains with sequential block production, a punishment mechanism known as slashing is often cited as a protection against the nothing-at-stake problem. However, in the context of a blockchain with parallel block production, it seems that slashing is not sufficient against the numerous divergence opportunities. In this paper, we propose a novel protection against the nothing-at-stake problem that takes the most out of BFT and Nakamoto-based consensus. By combining those approaches, we wish to scale up blockchains by allowing parallel block production without reconciliation.

cs.CR↗

arXiv2022

GRIMGEP: Learning Progress for Robust Goal Sampling in Visual Deep Reinforcement Learning

Designing agents, capable of learning autonomously a wide range of skills is critical in order to increase the scope of reinforcement learning. It will both increase the diversity of learned skills and reduce the burden of manually designing reward functions for each skill. Self-supervised agents, setting their own goals, and trying to maximize the diversity of those goals have shown great promise towards this end. However, a currently known limitation of agents trying to maximize the diversity of sampled goals is that they tend to get attracted to noise or more generally to parts of the environments that cannot be controlled (distractors). When agents have access to predefined goal features or expert knowledge, absolute Learning Progress (ALP) provides a way to distinguish between regions that can be controlled and those that cannot. However, those methods often fall short when the agents are only provided with raw sensory inputs such as images. In this work we extend those concepts to unsupervised image-based goal exploration. We propose a framework that allows agents to autonomously identify and ignore noisy distracting regions while searching for novelty in the learnable regions to both improve overall performance and avoid catastrophic forgetting. Our framework can be combined with any state-of-the-art novelty seeking goal exploration approaches. We construct a rich 3D image based environment with distractors. Experiments on this environment show that agents using our framework successfully identify interesting regions of the environment, resulting in drastically improved performances. The source code is available at https://sites.google.com/view/grimgep.

cs.LG↗

arXiv2020

Blockclique: scaling blockchains through transaction sharding in a multithreaded block graph

Decentralized crypto-currencies based on the blockchain architecture under-utilize available network bandwidth, making them unable to scale to thousands of transactions per second. We define the Blockclique architecture, that addresses this limitation by sharding transactions in a block graph with a fixed number of threads. The architecture allows the creation of intrinsically compatible blocks in parallel, where each block references one previous block of each thread. The consistency of the Blockclique protocol is formally established in presence of attackers. An experimental evaluation of the architecture's performance in large realistic networks demonstrates an efficient use of available bandwidth and a throughput of thousands of transactions per second.

cs.CR↗

arXiv2019

Autonomous Goal Exploration using Learned Goal Spaces for Visuomotor Skill Acquisition in Robots

The automatic and efficient discovery of skills, without supervision, for long-living autonomous agents, remains a challenge of Artificial Intelligence. Intrinsically Motivated Goal Exploration Processes give learning agents a human-inspired mechanism to sequentially select goals to achieve. This approach gives a new perspective on the lifelong learning problem, with promising results on both simulated and real-world experiments. Until recently, those algorithms were restricted to domains with experimenter-knowledge, since the Goal Space used by the agents was built on engineered feature extractors. The recent advances of deep representation learning, enables new ways of designing those feature extractors, using directly the agent experience. Recent work has shown the potential of those methods on simple yet challenging simulated domains. In this paper, we present recent results showing the applicability of those principles on a real-world robotic setup, where a 6-joint robotic arm learns to manipulate a ball inside an arena, by choosing goals in a space learned from its past experience.

cs.LG↗

arXiv2018

Curiosity Driven Exploration of Learned Disentangled Goal Spaces

Intrinsically motivated goal exploration processes enable agents to autonomously sample goals to explore efficiently complex environments with high-dimensional continuous actions. They have been applied successfully to real world robots to discover repertoires of policies producing a wide diversity of effects. Often these algorithms relied on engineered goal spaces but it was recently shown that one can use deep representation learning algorithms to learn an adequate goal space in simple environments. However, in the case of more complex environments containing multiple objects or distractors, an efficient exploration requires that the structure of the goal space reflects the one of the environment. In this paper we show that using a disentangled goal space leads to better exploration performances than an entangled goal space. We further show that when the representation is disentangled, one can leverage it by sampling goals that maximize learning progress in a modular manner. Finally, we show that the measure of learning progress, used to drive curiosity-driven exploration, can be used simultaneously to discover abstract independently controllable features of the environment.

cs.LG↗

arXiv2018

Continuous-variable supraquantum nonlocality

Supraquantum nonlocality refers to correlations that are more nonlocal than allowed by quantum theory but still physically conceivable in post-quantum theories, in the sense of respecting the basic no-faster-than-light communication principle. While supraquantum correlations are relatively well understood for finite-dimensional systems, little is known in the infinite-dimensional case. Here, we study supraquantum nonlocality for bipartite systems with two measurement settings and infinitely many outcomes per subsystem. We develop a formalism for generic no-signaling black-box measurement devices with continuous outputs in terms of probability measures, instead of probability distributions, which involves a few technical subtleties. We show the existence of a class of supraquantum Gaussian correlations, which violate the Tsirelson bound of an adequate continuous-variable Bell inequality. We then introduce the continuous-variable version of the celebrated Popescu-Rohrlich (PR) boxes, as a limiting case of the above-mentioned Gaussian ones. Finally, we perform a characterisation of the geometry of the set of continuous-variable no-signaling correlations. Namely, we show that that the convex hull of the continuous-variable PR boxes is dense in the no-signaling set. We also show that these boxes are extreme in the set of no-signaling behaviours and provide evidence suggesting that they are indeed the only extreme points of the no-signaling set. Our results lay the grounds for studying generalized-probability theories in continuous-variable systems.

quant-ph↗

arXiv2014

Effects of coordination and pressure on sound attenuation, boson peak and elasticity in amorphous solids

Connectedness and applied stress strongly affect elasticity in solids. In various amorphous solids, mechanical stability can be lost either by reducing connectedness or by increasing pressure. We present an effective medium theory of elasticity that extends previous approaches by incorporating the effect of compression, of amplitude $e$, allowing one to describe quantitative features of sound propagation, transport, the boson peak, and elastic moduli near the elastic instability occurring at a compression $e_c$. The theory disentangles several frequencies characterizing the vibrational spectrum: the onset frequency $ω_0\sim \sqrt{e_c-e}$ where strongly-scattered modes appear in the vibrational spectrum, the pressure-independent frequency $ω_*$ where the density of states displays a plateau, the boson peak frequency $ω_{BP}$, and the Ioffe-Regel frequency $ω_{IR}$ where scattering length and wavelength become equal. We predict that sound attenuation crosses over from $ω^4$ to $ω^2$ behaviour at $ω_0$. We predict that a frequency-dependent length scale $l_s(ω)$ and speed of sound $ν(ω)$ characterize vibrational modes, and could be extracted from scattering data. One key result is the prediction of a flat diffusivity above $ω_0$, in agreement with previously unexplained observations. We find that the shear modulus does not vanish at the elastic instability, but drops by a factor of 2. We check our predictions in packings of soft particles and study the case of covalent networks and silica. Overall, our approach unifies sound attenuation, transport and length scales entering elasticity in a single framework where disorder is not the main parameter controlling the boson peak, in agreement with observations. This framework leads to a phase diagram where various glasses can be placed, connecting microscopic structure to vibrational properties.

cond-mat.soft↗