On transfer learning using a MAC model variant

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.

Paper accepted for Visually Grounded Interaction and Language (ViGIL) Workshop, NIPS 2018, Montreeal, Canada

References in corpus (7)

• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Deep Residual Learning for Image Recognition
• VQA: Visual Question Answering
• A simple neural network module for relational reasoning
• Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning
• Compositional Attention Networks for Machine Reasoning
• Learning to Reason: End-to-End Module Networks for Visual Question Answering

Cited by in corpus (4)

• REMIND Your Neural Network to Prevent Catastrophic Forgetting
• Leveraging Medical Visual Question Answering with Supporting Facts
• Think before you act: A simple baseline for compositional generalization
• Transfer Learning in Visual and Relational Reasoning