25.8k citations · 55.9k across the 29 of their papers we have counts for
5 papers · 1 filter
Learning Simple Algorithms from Examples
Wojciech Zaremba, Tomas Mikolov, Armand Joulin +1
We present an approach for learning simple algorithms such as copying, multi-digit addition and single digit multiplication directly from examples. Our framework consists of a set…
Alternative structures for character-level RNNs
Piotr Bojanowski, Armand Joulin, Tomas Mikolov
Recurrent neural networks are convenient and efficient models for language modeling. However, when applied on the level of characters instead of words, they suffer from several pro…
A Roadmap towards Machine Intelligence
Tomas Mikolov, Armand Joulin, Marco Baroni
The development of intelligent machines is one of the biggest unsolved challenges in computer science. In this paper, we propose some fundamental properties these machines should h…
Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets
Armand Joulin, Tomas Mikolov
Despite the recent achievements in machine learning, we are still very far from achieving real artificial intelligence. In this paper, we discuss the limitations of standard deep l…
Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks
Jason Weston, Antoine Bordes, Sumit Chopra +4
One long-term goal of machine learning research is to produce methods that are applicable to reasoning and natural language, in particular building an intelligent dialogue agent. T…