Research Area
Year Published

625 Results

May 24, 2018

A Universal Music Translation Network


We present a method for translating music across musical instruments, genres, and styles. This method is based on a multi-domain wavenet autoencoder, with a shared encoder and a disentangled latent space that is trained end-to-end on waveforms.

By: Noam Mor, Lior Wolf, Adam Polyak, Yaniv Taigman
May 16, 2018

Glow: Graph Lowering Compiler Techniques for Neural Networks


This paper presents the design of Glow, a machine learning compiler for heterogeneous hardware. It is a pragmatic approach to compilation that enables the generation of highly optimized code for multiple targets. Glow lowers the traditional neural network dataflow graph into a two-phase strongly-typed intermediate representation.

By: Saleem Abdulrasool, Summer Deng, Roman Dzhabarov, Jordan Fix, James Hegeman, Roman Levenstein, Bert Maher, Satish Nadathur, Jakob Olesen, Jongsoo Park, Artem Rakhov, Nadav Rotem, Misha Smelyanskiy
May 15, 2018

Network Topology and Communication-Computation Tradeoffs in Decentralized Optimization

Proceedings of the IEEE

Motivated by a variety of applications—decentralized estimation in sensor networks, fitting models to massive data sets, and decentralized control of multi-robot systems, to name a few—significant advances have been made towards the development of robust, practical algorithms with theoretical performance guarantees. This paper presents an overview of recent work in this area.

By: Angelica Nedic, Alex Olshevsky, Mike Rabbat
May 8, 2018

Optimization Methods for Large-Scale Machine Learning

SIAM Review

This paper provides a review and commentary on the past, present, and future of numerical optimization algorithms in the context of machine learning applications.

By: Leon Bottou, Frank E. Curtis, Jorge Nocedal
May 7, 2018

Advances in Pre-Training Distributed Word Representations

Language Resources and Evaluation Conference (LREC)

In this paper, we show how to train high-quality word vector representations by using a combination of known tricks that are however rarely used together.

By: Tomas Mikolov, Edouard Grave, Piotr Bojanowski, Christian Puhrsch, Armand Joulin
May 7, 2018

A Corpus for Multilingual Document Classification in Eight Languages

Language Resources and Evaluation Conference (LREC)

In this paper, we propose a new subset of the Reuters corpus with balanced class priors for eight languages. By adding Italian, Russian, Japanese and Chinese, we cover languages which are very different with respect to syntax, morphology, etc. We provide strong baselines for all language transfer directions using multilingual word and sentence embeddings respectively. Our goal is to offer a freely available framework to evaluate cross-lingual document classification, and we hope to foster by these means, research in this important area.

By: Holger Schwenk, Xian Li
May 2, 2018

Exploring the Limits of Weakly Supervised Pretraining


In this paper, we present a unique study of transfer learning with large convolutional networks trained to predict hashtags on billions of social media images.

By: Dhruv Mahajan, Ross Girshick, Vignesh Ramanathan, Kaiming He, Manohar Paluri, Yixuan Li, Ashwin Bharambe, Laurens van der Maaten
April 30, 2018

Reward Estimation for Variance Reduction in Deep Reinforcement Learning

ICLR workshop

In reinforcement learning (RL), stochastic environments can make learning a policy difficult due to high degrees of variance. As such, variance reduction methods have been investigated in other works, such as advantage estimation and controlvariates estimation. Here, we propose to learn a separate reward estimator to train the value function, to help reduce variance caused by a noisy reward signal.

By: Joshua Romoff, Alexandre Piché, Peter Henderson, Vincent Francois-Lavet, Joelle Pineau
April 30, 2018

Identifying Analogies Across Domains

International Conference on Learning Representations (ICLR)

In this paper, we tackle this very task of finding exact analogies between datasets i.e. for every image from domain A find an analogous image in domain B. We present a matching-by-synthesis approach: AN-GAN, and show that it outperforms current techniques.

By: Yedid Hoshen, Lior Wolf
April 30, 2018

Emergent Communication in a Multi-Modal, Multi-Step Referential Game

International Conference on Learning Representations (ICLR)

Inspired by previous work on emergent communication in referential games, we propose a novel multi-modal, multi-step referential game, where the sender and receiver have access to distinct modalities of an object, and their information exchange is bidirectional and of arbitrary duration.

By: Katrina Evtimova, Andrew Drozdov, Douwe Kiela, Kyunghyun Cho