RanDumb: a simple approach that questions the efficacy of continual representation learning

Conference item

We propose RanDumb to examine the efficacy of continual representation learning. RanDumb embeds raw pixels using a fixed random transform which approximates an RBF-Kernel, initialized before seeing any data, and learns a simple linear classifier on top. We present a surprising and consistent finding: RanDumb significantly outperforms the continually learned representations using deep networks across numerous continual learning benchmarks, demonstrating the poor performance of representation learning in these scenarios. RanDumb stores no exemplars and performs a single pass over the data, processing one sample at a time. It complements GDumb [39], operating in a lowexemplar regime where GDumb has especially poor performance. We reach the same consistent conclusions when RanDumb is extended to scenarios with pretrained models replacing the random transform with pretrained feature extractor. Our investigation is both surprising and alarming as it questions our understanding of how to effectively design and train models that require efficient continual representation learning, and necessitates a principled reinvestigation of the widely explored problem formulation itself. Our code is available here.

Authors: Prabhu, A; Sinha, S; Kumaraguru, P; Torr, PHS; Sener, O

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Neural Information Processing Systems
• Host title: Advances in Neural Information Processing Systems 37 (NeurIPS 2024) Main Conference Track
• Volume: 37
• Pages: 37988-38006
• Publication date: 2025-02-01
• Acceptance date: 2024-02-26
• Event title: 38th Annual Conference on Neural Information Processing Systems (NeurIPS 2024)
• Event location: Vancouver, BC, Canada
• Event website: https://neurips.cc/Conferences/2024
• Event start date: 2024-12-10
• Event end date: 2024-12-15
• EISBN: 9798331314385
• Language: English