Incremental Embedding Learning via Zero-Shot Translation
arXiv:2012.15497
Abstract
Modern deep learning methods have achieved great success in machine learning and computer vision fields by learning a set of pre-defined datasets. Howerver, these methods perform unsatisfactorily when applied into real-world situations. The reason of this phenomenon is that learning new tasks leads the trained model quickly forget the knowledge of old tasks, which is referred to as catastrophic forgetting. Current state-of-the-art incremental learning methods tackle catastrophic forgetting problem in traditional classification networks and ignore the problem existing in embedding networks, which are the basic networks for image retrieval, face recognition, zero-shot learning, etc. Different from traditional incremental classification networks, the semantic gap between the embedding spaces of two adjacent tasks is the main challenge for embedding networks under incremental learning setting. Thus, we propose a novel class-incremental method for embedding network, named as zero-shot translation class-incremental method (ZSTCI), which leverages zero-shot translation to estimate and compensate the semantic gap without any exemplars. Then, we try to learn a unified representation for two adjacent tasks in sequential learning process, which captures the relationships of previous classes and current classes precisely. In addition, ZSTCI can easily be combined with existing regularization-based incremental learning methods to further improve performance of embedding networks. We conduct extensive experiments on CUB-200-2011 and CIFAR100, and the experiment results prove the effectiveness of our method. The code of our method has been released.
arXiv comment9 page
References in corpus (11)
- Adam: A Method for Stochastic Optimization
- Overcoming catastrophic forgetting in neural networks
- Continual Learning with Deep Generative Replay
- Deep Generative Dual Memory Network for Continual Learning
- Continual Learning in Generative Adversarial Nets
- Parameter-Free Spatial Attention Network for Person Re-Identification
- Rotate your Networks: Better Weight Consolidation and Less Catastrophic Forgetting
- Multi-modal Cycle-consistent Generalized Zero-Shot Learning
- What Can Be Transferred: Unsupervised Domain Adaptation for Endoscopic Lesions Segmentation
- Semantic Drift Compensation for Class-Incremental Learning
- Few-Shot Class-Incremental Learning