Learning Simple Algorithms from Examples

Wojciech Zaremba, Tomas Mikolov, Armand Joulin, Rob Fergus
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:421-429, 2016.

Abstract

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 of interfaces, accessed by a controller. Typical interfaces are 1-D tapes or 2-D grids that hold the input and output data. For the controller, we explore a range of neural network-based models which vary in their ability to abstract the underlying algorithm from training instances and generalize to test examples with many thousands of digits. The controller is trained using Q-learning with several enhancements and we show that the bottleneck is in the capabilities of the controller rather than in the search incurred by Q-learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v48-zaremba16, title = {Learning Simple Algorithms from Examples}, author = {Zaremba, Wojciech and Mikolov, Tomas and Joulin, Armand and Fergus, Rob}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {421--429}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/zaremba16.pdf}, url = {https://proceedings.mlr.press/v48/zaremba16.html}, abstract = {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 of interfaces, accessed by a controller. Typical interfaces are 1-D tapes or 2-D grids that hold the input and output data. For the controller, we explore a range of neural network-based models which vary in their ability to abstract the underlying algorithm from training instances and generalize to test examples with many thousands of digits. The controller is trained using Q-learning with several enhancements and we show that the bottleneck is in the capabilities of the controller rather than in the search incurred by Q-learning.} }
Endnote
%0 Conference Paper %T Learning Simple Algorithms from Examples %A Wojciech Zaremba %A Tomas Mikolov %A Armand Joulin %A Rob Fergus %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-zaremba16 %I PMLR %P 421--429 %U https://proceedings.mlr.press/v48/zaremba16.html %V 48 %X 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 of interfaces, accessed by a controller. Typical interfaces are 1-D tapes or 2-D grids that hold the input and output data. For the controller, we explore a range of neural network-based models which vary in their ability to abstract the underlying algorithm from training instances and generalize to test examples with many thousands of digits. The controller is trained using Q-learning with several enhancements and we show that the bottleneck is in the capabilities of the controller rather than in the search incurred by Q-learning.
RIS
TY - CPAPER TI - Learning Simple Algorithms from Examples AU - Wojciech Zaremba AU - Tomas Mikolov AU - Armand Joulin AU - Rob Fergus BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-zaremba16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 421 EP - 429 L1 - http://proceedings.mlr.press/v48/zaremba16.pdf UR - https://proceedings.mlr.press/v48/zaremba16.html AB - 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 of interfaces, accessed by a controller. Typical interfaces are 1-D tapes or 2-D grids that hold the input and output data. For the controller, we explore a range of neural network-based models which vary in their ability to abstract the underlying algorithm from training instances and generalize to test examples with many thousands of digits. The controller is trained using Q-learning with several enhancements and we show that the bottleneck is in the capabilities of the controller rather than in the search incurred by Q-learning. ER -
APA
Zaremba, W., Mikolov, T., Joulin, A. & Fergus, R.. (2016). Learning Simple Algorithms from Examples. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:421-429 Available from https://proceedings.mlr.press/v48/zaremba16.html.

Related Material