In NMT, how far can we get without attention and without separate encoding and decoding? To answer that question, we introduce a recurrent neural translation model that does not use attention and does not have a separate encoder and decoder. Our eager translation model is low-latency, writing target tokens as soon as it reads the first source token, and uses constant memory during decoding. It performs on par with the standard attention-based model of Bahdanau et al. (2014), and better on long sentences.

**Click to Read Paper****Click to Read Paper**

Language Generation with Recurrent Generative Adversarial Networks without Pre-training

Dec 21, 2017

Ofir Press, Amir Bar, Ben Bogin, Jonathan Berant, Lior Wolf

Dec 21, 2017

Ofir Press, Amir Bar, Ben Bogin, Jonathan Berant, Lior Wolf

**Click to Read Paper**

Fast and Accurate Reconstruction of Compressed Color Light Field

Mar 28, 2018

Ofir Nabati, David Mendlovic, Raja Giryes

Mar 28, 2018

Ofir Nabati, David Mendlovic, Raja Giryes

**Click to Read Paper**

On statistical learning via the lens of compression

Dec 30, 2016

Ofir David, Shay Moran, Amir Yehudayoff

This work continues the study of the relationship between sample compression schemes and statistical learning, which has been mostly investigated within the framework of binary classification. The central theme of this work is establishing equivalences between learnability and compressibility, and utilizing these equivalences in the study of statistical learning theory. We begin with the setting of multiclass categorization (zero/one loss). We prove that in this case learnability is equivalent to compression of logarithmic sample size, and that uniform convergence implies compression of constant size. We then consider Vapnik's general learning setting: we show that in order to extend the compressibility-learnability equivalence to this case, it is necessary to consider an approximate variant of compression. Finally, we provide some applications of the compressibility-learnability equivalences: (i) Agnostic-case learnability and realizable-case learnability are equivalent in multiclass categorization problems (in terms of sample complexity). (ii) This equivalence between agnostic-case learnability and realizable-case learnability does not hold for general learning problems: There exists a learning problem whose loss function takes just three values, under which agnostic-case and realizable-case learnability are not equivalent. (iii) Uniform convergence implies compression of constant size in multiclass categorization problems. Part of the argument includes an analysis of the uniform convergence rate in terms of the graph dimension, in which we improve upon previous bounds. (iv) A dichotomy for sample compression in multiclass categorization problems: If a non-trivial compression exists then a compression of logarithmic size exists. (v) A compactness theorem for multiclass categorization problems.
Dec 30, 2016

Ofir David, Shay Moran, Amir Yehudayoff

**Click to Read Paper**

Near-Optimal Representation Learning for Hierarchical Reinforcement Learning

Oct 02, 2018

Ofir Nachum, Shixiang Gu, Honglak Lee, Sergey Levine

Oct 02, 2018

Ofir Nachum, Shixiang Gu, Honglak Lee, Sergey Levine

**Click to Read Paper**

On Detection of Faint Edges in Noisy Images

Jun 22, 2017

Nati Ofir, Meirav Galun, Sharon Alpert, Achi Brandt, Boaz Nadler, Ronen Basri

Jun 22, 2017

Nati Ofir, Meirav Galun, Sharon Alpert, Achi Brandt, Boaz Nadler, Ronen Basri

**Click to Read Paper**