Continual Sequence Generation with Adaptive Compositional Modules
By Yanzhe Zhang, Xuezhi Wang, Diyi Yang. 2022
55 citations recorded by the source corpus.
Continual learning is essential for real-world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks. Existing work on continual sequence generation either always reuses existing parameters to learn new tasks, which is vulnerable to catastrophic forgetting on dissimilar tasks, or blindly adds new parameters for every new task, which could prevent knowledge sharing between similar tasks. To get the best of both worlds, in this work, we propose continual sequence generation with adaptive compositional modules to adaptively add modules in transformer architectures and compose both old and new modules for new tasks. We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules. Experiment results on various sequences of generation tasks show that our framework can adaptively add modules or reuse modules based on task similarity, outperforming state-of-the-art baselines in terms of both performance and parameter efficiency. We make our code public at https://github.com/GT-SALT/Adaptive-Compositional-Modules.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Generation, Meta & Continual Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Hang Su
- Jun Zhu
- Liyuan Wang
- Xingxing Zhang
- Leticia Pinto-Alva
- Mohammad Rostami
- Tejas Srinivasan
- Georgios Chochlakis
In Computer Vision and Pattern Recognition
- Hang Su
- Jun Zhu
- Liyuan Wang
- Xingxing Zhang
- Leticia Pinto-Alva
- Mohammad Rostami
- Tejas Srinivasan
- Georgios Chochlakis
In Meta & Continual Learning
- Xingxing Zhang
- Liyuan Wang
- Hang Su
- Jun Zhu
- Mohammad Rostami
- Simone Scardapane
- Valerio Marsocci
- Liyuan Wang
Citations in the corpus
- A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, background): “MBPA++ [442], EMAR [465], DnR [396], ARPER [298], Total Recall [253]), generative replay (PAGeR [415], LAMOL [395], ACM [499], NER [434]), parameter allocation (TPEM [131]), modular network (ProgModel [377]), meta-learning (MetaMBPA++ [442], MeLL [422], CML [455]), etc.”
- CLiMB: A Continual Learning Benchmark for Vision-and-Language Tasks (2022, background): “Within unimodal multi-task and continual learning, Hypernetworks [Mahabadi et al., 2021] and compositional Adapter modules [Zhang et al., 2022] have been shown to facilitate knowledge transfer by generating Adapter parameters from shared task information.”
- Continual Learning of Large Language Models: A Comprehensive Survey (2024, unknown): “Overall Performance (OP) [106, 286, 291] calculates the average performance up until the current training stage, measuring the overall ability of a model balancing the performance of each task.”