Abstract

Task incremental learning aims to enable a system to maintain its performance on previously learned tasks while learning new tasks, solving the problem of catastrophic forgetting. One promising approach is to build an individual network or sub-network for future tasks. However, this leads to an ever-growing memory due to saving extra weights for new tasks and how to address this issue has remained an open problem in task incremental learning. In this paper, we introduce a novel Neural Weight Search technique that designs a fixed search space where the optimal combinations of frozen weights can be searched to build new models for novel tasks in an end-to-end manner, resulting in a scalable and controllable memory growth. Extensive experiments on two benchmarks, i.e., Split-CIFAR-100 and CUB-to-Sketches, show our method achieves state-of-the-art performance with respect to both average inference accuracy and total memory cost.

Neural Weight Search

Results

Takeaways

• • Neural Weight Search is out-of-the-box mechanism that can be easily integrated with modern deep learning methods.

  • NWS builds a new model by searching layer-wise pools for the optimal combinations of grouped weights (with repetition).

  • Inside a NWS-built model, there could be heavily redundant grouped weights in the same layer.

  • Given a number of models to build, NWS can offer a large memory reduction with competitive inference accuracy.

Potential Applications

• Network Sparsification

• Network Compression

• Mix-of-Experts

• Customized Network