Characterizing signal propagation to close the performance gap in unnormalized ResNets
arXiv:2101.08692
Abstract
Batch Normalization is a key component in almost all state-of-the-art image classifiers, but it also introduces practical challenges: it breaks the independence between training examples within a batch, can incur compute and memory overhead, and often results in unexpected bugs. Building on recent theoretical analyses of deep ResNets at initialization, we propose a simple set of analysis tools to characterize signal propagation on the forward pass, and leverage these tools to design highly performant ResNets without activation normalization layers. Crucial to our success is an adapted version of the recently proposed Weight Standardization. Our analysis tools show how this technique preserves the signal in networks with ReLU or Swish activation functions by ensuring that the per-channel activation means do not grow with depth. Across a range of FLOP budgets, our networks attain performance competitive with the state-of-the-art EfficientNets on ImageNet.
Published as a conference paper at ICLR 2021
References in corpus (6)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Normalization Techniques in Training DNNs: Methodology, Analysis and Application
- On the Generalization Benefit of Noise in Stochastic Gradient Descent
- Deep Residual Networks and Weight Initialization
- Mean Shift Rejection: Training Deep Neural Networks Without Minibatch Statistics or Normalization
Cited by in corpus (5)
- SPIDE: A Purely Spike-based Method for Training Feedback Spiking Neural Networks
- Making EfficientNet More Efficient: Exploring Batch-Independent Normalization, Group Convolutions and Reduced Resolution Training
- A Loss Curvature Perspective on Training Instability in Deep Learning
- Parameter Prediction for Unseen Deep Architectures
- "BNN - BN = ?": Training Binary Neural Networks without Batch Normalization