We introduce a variant of the MAC model (Hudson and Manning, ICLR 2018) with a simplified set of equations that achieves comparable accuracy, while training faster. We evaluate both models on CLEVR and CoGenT, and show that, transfer learning with fine-tuning results in a 15 point increase in accuracy, matching the state of the art. Finally, in contrast, we demonstrate that improper fine-tuning can actually reduce a model's accuracy as well.
@article{arxiv.1811.06529,
title = {On transfer learning using a MAC model variant},
author = {Vincent Marois and T. S. Jayram and Vincent Albouy and Tomasz Kornuta and Younes Bouhadjar and Ahmet S. Ozcan},
journal= {arXiv preprint arXiv:1811.06529},
year = {2018}
}
Comments
Paper accepted for Visually Grounded Interaction and Language (ViGIL) Workshop, NIPS 2018, Montreeal, Canada