/torchNALU

PyTorch implementation of 'Neural Arithmetic Logic Units'

Primary LanguagePython

torchNALU

Code in this repository provides a PyTorch implementation of the Neural Arithmetic Logic Units paper from Deepmind. The results produced align with those described within the paper, demonstrating the effectiveness of the NALU/NAC architectures.

Experiments:

To reproduce the results for a static task, run:

python3 network.py

The graphs below show the MSE normalised with respect to the results obtained from a random MLP network such that 100.0 is equivalent to random, 0.0 is perfect accuracy, and >100.0 is worse than a randomly initialised model.

References:

@misc{trask2018neural,
    title={Neural Arithmetic Logic Units},
    author={Andrew Trask and Felix Hill and Scott Reed and Jack Rae and Chris Dyer and Phil Blunsom},
    year={2018},
    eprint={1808.00508},
    archivePrefix={arXiv},
    primaryClass={cs.NE}
}
PyTorch implementation of Neural Arithmetic Logic Units
Made by Daniel Kelshaw