Class Normalization for (Continual)? Generalized Zero-Shot Learning
This work addresses normalization issues in ZSL, a domain-specific problem for machine learning researchers, with incremental improvements in method and broader impact through new benchmarks for continual ZSL.
The paper tackles the under-explored use of normalization in zero-shot learning (ZSL) by proposing Class Normalization, which improves training stability and performance, achieving state-of-the-art results on 4 standard datasets with ~50 times faster training speed.
Normalization techniques have proved to be a crucial ingredient of successful training in a traditional supervised learning regime. However, in the zero-shot learning (ZSL) world, these ideas have received only marginal attention. This work studies normalization in ZSL scenario from both theoretical and practical perspectives. First, we give a theoretical explanation to two popular tricks used in zero-shot learning: normalize+scale and attributes normalization and show that they help training by preserving variance during a forward pass. Next, we demonstrate that they are insufficient to normalize a deep ZSL model and propose Class Normalization (CN): a normalization scheme, which alleviates this issue both provably and in practice. Third, we show that ZSL models typically have more irregular loss surface compared to traditional classifiers and that the proposed method partially remedies this problem. Then, we test our approach on 4 standard ZSL datasets and outperform sophisticated modern SotA with a simple MLP optimized without any bells and whistles and having ~50 times faster training speed. Finally, we generalize ZSL to a broader problem -- continual ZSL, and introduce some principled metrics and rigorous baselines for this new setup. The project page is located at https://universome.github.io/class-norm.