Personne
Aymeric Daphnis Kévin Dieuleveut
Cette personne n’est plus à l’EPFL
Publications associées (5)
Context Mover's Distance & Barycenters: Optimal Transport of Contexts for Building Representations
Martin Jaggi, Aymeric Daphnis Kévin Dieuleveut, Sidak Pal Singh, Andreas Hug
We present a framework for building unsupervised representations of entities and their compositions, where each entity is viewed as a probability distribution rather than a vector embedding. In particular, this distribution is supported over the contexts w ...
ADDISON-WESLEY PUBL CO2020On Convergence-Diagnostic based Step Sizes for Stochastic Gradient Descent
Nicolas Henri Bernard Flammarion, Scott William Pesme, Aymeric Daphnis Kévin Dieuleveut
Constant step-size Stochastic Gradient Descent exhibits two phases: a transient phase during which iterates make fast progress towards the optimum, followed by a stationary phase during which iterates oscillate around the optimal point. In this paper, we s ...
2020Communication trade-offs for Local-SGD with large step size
Aymeric Daphnis Kévin Dieuleveut
Synchronous mini-batch SGD is state-of-the-art for large-scale distributed machine learning. However, in practice, its convergence is bottlenecked by slow communication rounds between worker nodes. A natural solution to reduce communication is to use the " ...
NEURAL INFORMATION PROCESSING SYSTEMS (NIPS)2019Unsupervised Scalable Representation Learning for Multivariate Time Series
Martin Jaggi, Aymeric Daphnis Kévin Dieuleveut
Time series constitute a challenging data type for machine learning algorithms, due to their highly variable lengths and sparse labeling in practice. In this paper, we tackle this challenge by proposing an unsupervised method to learn universal embeddings ...
NEURAL INFORMATION PROCESSING SYSTEMS (NIPS)2019Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regression
Nicolas Henri Bernard Flammarion, Aymeric Daphnis Kévin Dieuleveut
We consider the optimization of a quadratic objective function whose gradients are only accessible through a stochastic oracle that returns the gradient at any given point plus a zero-mean finite variance random error. We present the first algorithm that a ...
2017