On the Variance of the Adaptive Learning Rate and Beyond
Abstrak
The learning rate warmup heuristic achieves remarkable success in stabilizing training, accelerating convergence and improving generalization for adaptive stochastic optimization algorithms like RMSprop and Adam. Here, we study its mechanism in details. Pursuing the theory behind warmup, we identify a problem of the adaptive learning rate (i.e., it has problematically large variance in the early stage), suggest warmup works as a variance reduction technique, and provide both empirical and theoretical evidence to verify our hypothesis. We further propose RAdam, a new variant of Adam, by introducing a term to rectify the variance of the adaptive learning rate. Extensive experimental results on image classification, language modeling, and neural machine translation verify our intuition and demonstrate the effectiveness and robustness of our proposed method. All implementations are available at: this https URL.
Topik & Kata Kunci
Penulis (7)
Liyuan Liu
Haoming Jiang
Pengcheng He
Weizhu Chen
Xiaodong Liu
Jianfeng Gao
Jiawei Han
Akses Cepat
PDF tidak tersedia langsung
Cek di sumber asli →- Tahun Terbit
- 2019
- Bahasa
- en
- Total Sitasi
- 2182×
- Sumber Database
- Semantic Scholar
- Akses
- Open Access ✓