On transfer learning using a MAC model variant

Ahmet S. Ozcan; Tomasz Kornuta; T.S. Jayram; Vincent Albouy; Vincent Marois; Younes Bouhadjar

arxiv: 1811.06529 · v2 · pith:IKLTCO5R · submitted 2018-11-15 · cs.CV · cs.LG

On transfer learning using a MAC model variant

Vincent Marois , T.S. Jayram , Vincent Albouy , Tomasz Kornuta , Younes Bouhadjar , Ahmet S. Ozcan This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:IKLTCO5R record.json open to challenge →

classification cs.CV cs.LG

keywords accuracymodelfine-tuninglearningtransfervariantachievesactually

0 comments

read the original abstract

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.

This paper has not been read by Pith yet.

On transfer learning using a MAC model variant

discussion (0)