Lun Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lun Wang has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Abhradeep Thakurta
Google research
1 shared publication - Arun Narayanan
Hypertrophic Cardiomyopathy Center, Division of Cardiovascular Medicine, University of Texas Medical Branch, Galveston (O.M.A., D.A.J., A.N.).
1 shared publication - Hongbin Liu
Duke University
1 shared publication Om Thakkar
Duke University 2 Google 1
1 shared publication
Latest Publications
- Open Access
Differentially Private Parameter-Efficient Fine-tuning for Large ASR Models
Large ASR models can inadvertently leak sensitive information, which can be mitigated by formal privacy measures like differential privacy (DP). However, traditional DP training is computationally expensive, and can hurt model performance. Our study explores DP parameter-efficient fine-tuning as a way to mitigate privacy risks with smaller computation and performance costs for ASR models. Through extensive experimentation and progressive optimization, we achieve 4.6%/8.1% word error rate on LibriSpeech clean/other test-sets, setting a new performance benchmark while maintaining (10, 3.52e-6)-DP in fine-tuning a large ASR model with over 600M parameters.
Open AccessDifferentially Private Parameter-Efficient Fine-tuning for Large ASR Models
Large ASR models can inadvertently leak sensitive information, which can be mitigated by formal privacy measures like differential privacy (DP). However, traditional DP training is computationally expensive, and can hurt model performance. Our study explores DP parameter-efficient fine-tuning as a way to mitigate privacy risks with smaller computation and performance costs for ASR models. Through extensive experimentation and progressive optimization, we achieve 4.6%/8.1% word error rate on LibriSpeech clean/other test-sets, setting a new performance benchmark while maintaining (10, 3.52e-6)-DP in fine-tuning a large ASR model with over 600M parameters.