DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion

Deep network architectures struggle to continually learn new tasks without forgetting the previous tasks. A recent trend indicates that dynamic architectures based on an expansion of the parameters can reduce catastrophic forgetting efficiently in continual learning. However, existing approaches often require a task identifier at test-time, need complex tuning to balance the growing number of parameters, and barely share any information across tasks. As a result, they struggle to scale to a large number of tasks without significant overhead. In this paper, we propose a transformer architecture based on a dedicated encoder/decoder framework. Critically, the encoder and decoder are shared among all tasks. Through a dynamic expansion of special tokens, we specialize each forward of our decoder network on a task distribution. Our strategy scales to a large number of tasks while having negligible memory and time overheads due to strict control of the parameters expansion. Moreover, this efficient strategy doesn't need any hyperparameter tuning to control the network's expansion. Our model reaches excellent results on CIFAR100 and state-of-the-art performances on the large-scale ImageNet100 and ImageNet1000 while having less parameters than concurrent dynamic frameworks.

PDF Abstract CVPR 2022 PDF CVPR 2022 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Incremental Learning ImageNet100 - 10 steps DyTox Average Incremental Accuracy 77.15 # 4
Final Accuracy 69.10 # 1
Average Incremental Accuracy Top-5 92.04 # 3
Final Accuracy Top-5 87.98 # 2
# M Params 11.01 # 1
Incremental Learning ImageNet - 10 steps DyTox Average Incremental Accuracy 71.29 # 1
Final Accuracy 63.34 # 1
Average Incremental Accuracy Top-5 88.59 # 1
Final Accuracy Top-5 84.49 # 1
# M Params 11.36 # 1

Methods


No methods listed for this paper. Add relevant methods here