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.
Neural arithmetic logic units
2019-08-16
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
DDC: | 629 |
Ancient Vedic Multiplication Based Optimized High Speed Arithmetic Logic
BASE | 2014
|ARITHMETIC LOGIC DEVICE, CAMERA DEVICE, VEHICLE AND CALIBRATION METHOD
Europäisches Patentamt | 2020
|Design and Implementation of a Reversible Logic based 8-Bit Arithmetic and Logic Unit
British Library Online Contents | 2014
|Design and Implementation of a Reversible Logic Based 8-Bit Arithmetic and Logic Unit
British Library Online Contents | 2014
|