Ilya Sutskever

Mastering the game of Go with deep neural networks and tree search

The game of Go has long been viewed as the most challenging of classic games for artificial intelligence owing to its enormous search space and the difficulty of evaluating board positions and moves. Here we introduce a new approach to computer Go that uses ‘value networks’ to evaluate board positions and ‘policy networks’ to select moves. These deep neural networks are trained by a novel combination of supervised learning from human expert games, and reinforcement learning from games of self-play. Without any lookahead search, the neural networks play Go at the level of state-of-the-art Monte Carlo tree search programs that simulate thousands of random games of self-play. We also introduce a new search algorithm that combines Monte Carlo simulation with value and policy networks. Using this search algorithm, our program AlphaGo achieved a 99.8% winning rate against other Go programs, and defeated the human European Go champion by 5 games to 0. This is the first time that a computer program has defeated a human professional player in the full-sized game of Go, a feat previously thought to be at least a decade away.

Addressing the Rare Word Problem in Neural Machine Translation

Neural Machine Translation (NMT) is a new approach to machine translation that has shown promising results that are comparable to traditional approaches. A significant weakness in conventional NMT systems is their inability to correctly translate very rare words: end-to-end NMTs tend to have relatively small vocabularies with a single unk symbol that represents every possible out-of-vocabulary (OOV) word. In this paper, we propose and implement an effective technique to address this problem. We train an NMT system on data that is augmented by the output of a word alignment algorithm, allowing the NMT system to emit, for each OOV word in the target sentence, the position of its corresponding word in the source sentence. This information is later utilized in a post-processing step that translates every OOV word using a dictionary. Our experiments on the WMT’14 English to French translation task show that this method provides a substantial improvement of up to 2.8 BLEU points over an equivalent NMT system that does not use this technique. With 37.5 BLEU points, our NMT system is the first to surpass the best result achieved on a WMT’14 contest task.

Deep, narrow sigmoid belief networks are universal approximators

In this note, we show that exponentially deep belief networks can approximate any distribution over binary vectors to arbitrary accuracy, even when the width of each layer is limited to the dimensionality of the data. We further show that such networks can be greedily learned in an easy yet impractical way.

Training Deep and Recurrent Networks with Hessian-Free Optimization

In this chapter we will first describe the basic HF approach, and then examine well-known performance-improving techniques such as preconditioning which we have found to be beneficial for neural network training, as well as others of a more heuristic nature which are harder to justify, but which we have found to work well in practice. We will also provide practical tips for creating efficient and bug-free implementations and discuss various pitfalls which may arise when designing and using an HF-type approach in a particular application.

Temporal-Kernel Recurrent Neural Networks

A Recurrent Neural Network (RNN) is a powerful connectionist model that can be applied to many challenging sequential problems, including problems that naturally arise in language and speech. However, RNNs are extremely hard to train on problems that have long-term dependencies, where it is necessary to remember events for many timesteps before using them to make a prediction. In this paper we consider the problem of training RNNs to predict sequences that exhibit significant long-term dependencies, focusing on a serial recall task where the RNN needs to remember a sequence of characters for a large number of steps before reconstructing it. We introduce the Temporal-Kernel Recurrent Neural Network (TKRNN), which is a variant of the RNN that can cope with long-term dependencies much more easily than a standard RNN, and show that the TKRNN develops short-term memory that successfully solves the serial recall task by representing the input string with a stable state of its hidden units.

Learning online alignments with continuous rewards policy gradient

Sequence-to-sequence models with soft attention had significant success in machine translation, speech recognition, and question answering. Though capable and easy to use, they require that the entirety of the input sequence is available at the beginning of inference, an assumption that is not valid for instantaneous translation and speech recognition. To address this problem, we present a new method for solving sequence-to-sequence problems using hard online alignments instead of soft offline alignments. The online alignments model is able to start producing outputs without the need to first process the entire input sequence. A highly accurate online sequence-to-sequence model is useful because it can be used to build an accurate voice-based instantaneous translator. Our model uses hard binary stochastic decisions to select the timesteps at which outputs will be produced. The model is trained to produce these stochastic decisions using a standard policy gradient method. In our experiments, we show that this model achieves encouraging performance on TIMIT and Wall Street Journal (WSJ) speech recognition datasets.

A simpler unified analysis of budget perceptrons

The kernel Perceptron is an appealing online learning algorithm that has a drawback: whenever it makes an error it must increase its support set, which slows training and testing if the number of errors is large. The Forgetron and the Randomized Budget Perceptron algorithms overcome this problem by restricting the number of support vectors the Perceptron is allowed to have. These algorithms have regret bounds whose proofs are dissimilar. In this paper we propose a unified analysis of both of these algorithms by observing that the way in which they remove support vectors can be seen as types of L2-regularization. By casting these algorithms as instances of online convex optimization problems and applying a variant of Zinkevichs theorem for noisy and incorrect gradient, we can bound the regret of these algorithms more easily than before. Our bounds are similar to the existing ones, but the proofs are less technical.