• Publications
  • Influence
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
TLDR
This paper describes the TensorFlow interface for expressing machine learning algorithms, and an implementation of that interface that we have built at Google. Expand
  • 7,977
  • 902
  • PDF
Generating Sentences from a Continuous Space
TLDR
We introduce and study an RNN-based variational autoencoder generative model that incorporates distributed latent representations of entire sentences that interpolate between known sentences. Expand
  • 1,251
  • 245
  • PDF
Exploring the Limits of Language Modeling
TLDR
In this work we explore recent advances in Recurrent Neural Networks for large scale Language Modeling, a task central to language understanding. Expand
  • 782
  • 93
  • PDF
An Empirical Exploration of Recurrent Network Architectures
TLDR
We evaluated over ten thousand different RNN architectures, and identified an architecture that outperforms both the LSTM and the recently-introduced Gated Recurrent Unit on some but not all tasks. Expand
  • 1,147
  • 89
  • PDF
Revisiting Distributed Synchronous SGD
TLDR
We propose a method of synchronous stochastic optimization with backup workers to mitigate straggler effects without gradient staleness in deep neural network models. Expand
  • 469
  • 69
  • PDF
Learning dexterous in-hand manipulation
TLDR
We use reinforcement learning (RL) to learn dexterous in-hand manipulation policies that can perform vision-based object reorientation on a physical Shadow Dexterous Hand. Expand
  • 552
  • 35
  • PDF
Learning to Generate Reviews and Discovering Sentiment
TLDR
We explore the properties of byte-level recurrent language models learned in an unsupervised manner, achieve state of the art on the binary subset of the Stanford Sentiment Treebank. Expand
  • 300
  • 27
  • PDF
Dota 2 with Large Scale Deep Reinforcement Learning
TLDR
We developed a distributed training system and tools for continual training which allowed us to train a Dota 2-playing agent called OpenAI Five for 10 months. Expand
  • 202
  • 19
  • PDF
Fault diagnosis and fault tolerant control
  • 155
  • 13
  • PDF
Inferring single-trial neural population dynamics using sequential auto-encoders
TLDR
LFADS, a deep learning method for analyzing neural population activity, can extract neural dynamics from single-trial recordings, stitch separate datasets into one model, and infer perturbations, for example, from behavioral choices to these dynamics. Expand
  • 155
  • 8
  • PDF