ICML 2018oral3 citations
Training Neural Machines with Trace-Based Supervision
Matthew Mirman, Dimitar Dimitrov, Pavle Djordjevic, Timon Gehr, Martin Vechev
Abstract
We investigate the effectiveness of trace-based supervision methods for training existing neural abstract machines. To define the class of neural machines amenable to trace-based supervision, we introduce the concept of a differential neural computational machine (dNCM) and show that several existing architectures (NTMs, NRAMs) can be described as dNCMs. We performed a detailed experimental evaluation with NTM and NRAM machines, showing that additional supervision on the interpretable portions of these architectures leads to better convergence and generalization capabilities of the learning phase than standard training, in both noise-free and noisy scenarios.
BibTeX
@InProceedings{pmlr-v80-mirman18a,
title = {Training Neural Machines with Trace-Based Supervision},
author = {Mirman, Matthew and Dimitrov, Dimitar and Djordjevic, Pavle and Gehr, Timon and Vechev, Martin},
booktitle = {Proceedings of the 35th International Conference on Machine Learning},
pages = {3569--3577},
year = {2018},
editor = {Dy, Jennifer and Krause, Andreas},
volume = {80},
series = {Proceedings of Machine Learning Research},
month = {10--15 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v80/mirman18a/mirman18a.pdf},
url = {https://proceedings.mlr.press/v80/mirman18a.html},
abstract = {We investigate the effectiveness of trace-based supervision methods for training existing neural abstract machines. To define the class of neural machines amenable to trace-based supervision, we introduce the concept of a differential neural computational machine (dNCM) and show that several existing architectures (NTMs, NRAMs) can be described as dNCMs. We performed a detailed experimental evaluation with NTM and NRAM machines, showing that additional supervision on the interpretable portions of these architectures leads to better convergence and generalization capabilities of the learning phase than standard training, in both noise-free and noisy scenarios.}
}