Dynamic Acoustic Model Architecture Optimization in Training for ASR
This work addresses the challenge of optimizing acoustic model architectures for ASR, offering a computationally efficient alternative to handcrafted rules or intensive neural architecture search, though it appears incremental as it builds on existing training methods.
The paper tackles the problem of complex architecture design in automatic speech recognition by introducing DMAO, a framework that dynamically reallocates parameters during training using a grow-and-drop strategy, resulting in up to 6% relative improvement in word error rate across multiple datasets and architectures with negligible training overhead.
Architecture design is inherently complex. Existing approaches rely on either handcrafted rules, which demand extensive empirical expertise, or automated methods like neural architecture search, which are computationally intensive. In this paper, we introduce DMAO, an architecture optimization framework that employs a grow-and-drop strategy to automatically reallocate parameters during training. This reallocation shifts resources from less-utilized areas to those parts of the model where they are most beneficial. Notably, DMAO only introduces negligible training overhead at a given model complexity. We evaluate DMAO through experiments with CTC on LibriSpeech, TED-LIUM-v2 and Switchboard datasets. The results show that, using the same amount of training resources, our proposed DMAO consistently improves WER by up to 6% relatively across various architectures, model sizes, and datasets. Furthermore, we analyze the pattern of parameter redistribution and uncover insightful findings.