Regularity Normalization: Neuroscience-Inspired Unsupervised Attention across Neural Network Layers
This work addresses input distribution challenges in neural networks for various domains, offering a novel normalization approach with broad applicability.
The paper tackles the problem of limited, imbalanced, and non-stationary input distributions in neural networks by proposing regularity normalization, an unsupervised attention mechanism inspired by neuronal adaptation. It outperforms existing normalization methods across tasks like image classification and reinforcement learning, demonstrating flexibility with data priors.
Inspired by the adaptation phenomenon of neuronal firing, we propose the regularity normalization (RN) as an unsupervised attention mechanism (UAM) which computes the statistical regularity in the implicit space of neural networks under the Minimum Description Length (MDL) principle. Treating the neural network optimization process as a partially observable model selection problem, the regularity normalization constrains the implicit space by a normalization factor, the universal code length. We compute this universal code incrementally across neural network layers and demonstrate the flexibility to include data priors such as top-down attention and other oracle information. Empirically, our approach outperforms existing normalization methods in tackling limited, imbalanced and non-stationary input distribution in image classification, classic control, procedurally-generated reinforcement learning, generative modeling, handwriting generation and question answering tasks with various neural network architectures. Lastly, the unsupervised attention mechanisms is a useful probing tool for neural networks by tracking the dependency and critical learning stages across layers and recurrent time steps of deep networks.