Neural arithmetic logic units

Conference item

Neural networks can learn to represent and manipulate numerical information, but they seldom generalize well outside of the range of numerical values encountered during training. To encourage more systematic numerical extrapolation, we propose an architecture that represents numerical quantities as linear activations which are manipulated using primitive arithmetic operators, controlled by learned gates. We call this module a neural arithmetic logic unit (NALU), by analogy to the arithmetic logic unit in traditional processors. Experiments show that NALU-enhanced neural networks can learn to track time, perform arithmetic over images of numbers, translate numerical language into real-valued scalars, execute computer code, and count objects in images. In contrast to conventional architectures, we obtain substantially better generalization both inside and outside of the range of numerical values encountered during training, often extrapolating orders of magnitude beyond trained numerical ranges.

Authors: Trask, A; Hill, F; Reed, SE; Rae, J; Dyer, C

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Curran Associates
• Host title: Advances in Neural Information Processing Systems 31 (NeurIPS 2018)
• Pages: 8035-8044
• Publication date: 2019-07-01
• Acceptance date: 2018-09-05
• Event title: 32nd Conference on Neural Information Processing Systems (NeurIPS 2018)
• Event location: Palais des Congrès de Montréal, Montréal, Canada
• Event website: https://nips.cc/Conferences/2018
• Event start date: 2018-12-03
• Event end date: 2018-12-08
• ISSN: 1049-5258
• ISBN: 9781510884472
• Language: English