SKILL: Similarity-aware Knowledge distILLation for Speech Self-Supervised Learning
This work addresses efficiency for speech processing applications, representing an incremental improvement over existing knowledge distillation techniques.
The paper tackles the problem of improving efficiency in speech self-supervised learning by introducing SKILL, a method that distills knowledge across groups of layers based on similarity measures, resulting in a distilled model that outperforms prior methods and achieves state-of-the-art results in the 30M parameters class across SUPERB tasks.
Self-supervised learning (SSL) has achieved remarkable success across various speech-processing tasks. To enhance its efficiency, previous works often leverage the use of compression techniques. A notable recent attempt is DPHuBERT, which applies joint knowledge distillation (KD) and structured pruning to learn a significantly smaller SSL model. In this paper, we contribute to this research domain by introducing SKILL, a novel method that conducts distillation across groups of layers instead of distilling individual arbitrarily selected layers within the teacher network. The identification of the layers to distill is achieved through a hierarchical clustering procedure applied to layer similarity measures. Extensive experiments demonstrate that our distilled version of WavLM Base+ not only outperforms DPHuBERT but also achieves state-of-the-art results in the 30M parameters model class across several SUPERB tasks.