LGSep 30, 2022

Task Formulation Matters When Learning Continually: A Case Study in Visual Question Answering

arXiv:2210.00044v27 citationsh-index: 38
Originality Synthesis-oriented
AI Analysis

This work addresses the challenge of applying continual learning to Vision+Language tasks, providing insights for researchers in multimodal AI, though it is incremental as it focuses on analyzing existing settings rather than proposing new methods.

The study investigates how different task formulations affect continual learning performance in Visual Question Answering, finding that performance and sensitivity to task order depend on output distribution shifts and highlighting the importance of stabilizing visual representations in deeper layers.

Continual learning aims to train a model incrementally on a sequence of tasks without forgetting previous knowledge. Although continual learning has been widely studied in computer vision, its application to Vision+Language tasks is not that straightforward, as settings can be parameterized in multiple ways according to their input modalities. In this paper, we present a detailed study of how different settings affect performance for Visual Question Answering. We first propose three plausible task formulations and demonstrate their impact on the performance of continual learning algorithms. We break down several factors of task similarity, showing that performance and sensitivity to task order highly depend on the shift of the output distribution. We also investigate the potential of pretrained models and compare the robustness of transformer models with different visual embeddings. Finally, we provide an analysis interpreting model representations and their impact on forgetting. Our results highlight the importance of stabilizing visual representations in deeper layers.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes