A Dataset and Architecture for Visual Reasoning with a Working Memory

TL;DR

This paper introduces a new visual reasoning dataset called COG, inspired by cognitive science, and proposes a deep learning architecture that performs well on this and other VQA datasets, with interpretability and zero-shot generalization.

Contribution

The paper presents a configurable visual reasoning dataset (COG) and a deep learning model that generalizes and is interpretable, addressing challenges in visual reasoning and memory.

Findings

Model performs competitively on CLEVR and COG datasets.

Network can zero-shot generalize to new tasks.

Preliminary analysis shows human-interpretable reasoning.

Abstract

A vexing problem in artificial intelligence is reasoning about events that occur in complex, changing visual stimuli such as in video analysis or game play. Inspired by a rich tradition of visual reasoning and memory in cognitive psychology and neuroscience, we developed an artificial, configurable visual question and answer dataset (COG) to parallel experiments in humans and animals. COG is much simpler than the general problem of video analysis, yet it addresses many of the problems relating to visual and logical reasoning and memory -- problems that remain challenging for modern deep learning architectures. We additionally propose a deep learning architecture that performs competitively on other diagnostic VQA datasets (i.e. CLEVR) as well as easy settings of the COG dataset. However, several settings of COG result in datasets that are progressively more challenging to learn. After training, the network can zero-shot generalize to many new tasks. Preliminary analyses of the network architectures trained on COG demonstrate that the network accomplishes the task in a manner interpretable to humans.