paper-with-me

홈 › Papers

Neural Arithmetic Logic Units

2018-08-01 · NeurIPS 2018 12 · Andrew Trask, Felix Hill, Scott Reed, Jack Rae, Chris Dyer, Phil Blunsom

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.

📄 PDF Abstract BibTeX arXiv:1808.00508

Code (36)

1hb6s7t/nalu.ms mindspore
1hb6s7t/snalu.ms mindspore
AnikMallick/NALU_tf tf
FlorianWilhelm/snalu.pytorch pytorch
FrederikWarburg/latent_disagreement pytorch
MindSpore-scientific-2/code-10/tree/main/nalu.ms mindspore
MindSpore-scientific-2/code-11/tree/main/nalu.ms mindspore
MindSpore-scientific-2/code-5/tree/main/nalu.ms mindspore
MindSpore-scientific-2/code-5/tree/main/pixelshuffle1d mindspore
MindSpore-scientific-2/code-8/tree/main/nalu.ms mindspore
MindSpore-scientific/code-1/tree/main/nalu.ms mindspore
MindSpore-scientific/code-10/tree/main/nalu.ms mindspore
MindSpore-scientific/code-11/tree/main/nalu.ms mindspore
MindSpore-scientific/code-11/tree/main/pixelshuffle1d mindspore
MindSpore-scientific/code-13/tree/main/nalu.ms mindspore
MindSpore-scientific/code-4/tree/main/nalu.ms mindspore
MindSpore-scientific/code-8/tree/main/nalu.ms mindspore
MindSpore-scientific/code-8/tree/main/nystrom-attention mindspore
MindSpore-scientific/code-8/tree/main/pixelshuffle1d mindspore
Nilabhra/NALU pytorch
Ragabov/tensorflow-nalu tf
acyrl/NALU tf
ahylton19/simpleNALU-tf tf
akilhylton/simpleNALU-tf tf
arthurdouillard/nalu.pytorch pytorch
bmistry4/nalm-benchmark pytorch
dillondaudert/neural-alu
grananqvist/NALU-tf tf
juliusfrost/Research-Paper-Implementations
llealgt/NALUs tf
merrymercy/NALU tf
nmheim/NeuralArithmetic.jl
pushpulldecoder/Enhanced-NALU pytorch
sayakpaul/NALU pytorch
shshemi/NeuralArithmeticLogicUnit tf
tejank10/NeuralALU.jl pytorch

Similar Papers 제목 키워드 기반

Neural Arithmetic Units

2020-01-14 · ICLR 2020 1 · Andreas Madsen, Alexander Rosenberg Johansen

Neural networks can approximate complex functions, but they struggle to perform exact arithmetic operations over real numbers. The lack of inductive bias for arithmetic operations leaves neural networks without the under…

Inductive Bias

Convolutional Feature Extraction and Neural Arithmetic Logic Units for Stock Prediction

2019-05-18 · Shangeth Rajaa, Jajati Keshari Sahoo

Stock prediction is a topic undergoing intense study for many years. Finance experts and mathematicians have been working on a way to predict the future stock price so as to decide to buy the stock or sell it to make pro…

Sentiment AnalysisStock Prediction

Algoritmos de minería de datos en la industria sanitaria

2021-04-19 · Marta Li Wang

In this paper, we review data mining approaches for health applications. Our focus is on hardware-centric approaches. Modern computers consist of multiple processors, each equipped with multiple cores, each with a set of…

EEGElectroencephalogram (EEG)

Systematically designing better instance counting models on cell images with Neural Arithmetic Logic Units

2020-04-14 · Ashish Rana, Taranveer Singh, Harpreet Singh, Neeraj Kumar 외

The big problem for neural network models which are trained to count instances is that whenever test range goes high training range generalization error increases i.e. they are not good generalizers outside training rang…

Data Augmentationregression

GENIAL: Generative Design Space Exploration via Network Inversion for Low Power Algorithmic Logic Units

2025-07-25 · Maxence Bouvier, Ryan Amaudruz, Felix Arnold, Renzo Andri 외 arxiv

As AI workloads proliferate, optimizing arithmetic units is becoming increasingly important for reducing the footprint of digital systems. Conventional design flows, which often rely on manual or heuristic-based optimiza…