neural tangent kernel
A mathematical construct used to understand the training dynamics of neural networks in the infinite width limit, helping researchers analyze the behavior of gradient descent in deep learning.
- A Closer Look at NTK Alignment: Linking Phase Transitions in Deep Image Regression
- Better NTK Conditioning: A Free Lunch from (ReLU) Nonlinear Activation in Wide Neural Networks
- Certifying Deep Network Risks and Individual Predictions with PAC-Bayes Loss via Localized Priors
- Just One Layer Norm Guarantees Stable Extrapolation
- Learning Provably Improves the Convergence of Gradient Descent
- Learning to Add, Multiply, and Execute Algorithmic Instructions Exactly with Neural Networks
- Linearization Explains Fine-Tuning in Large Language Models
- NTKMTL: Mitigating Task Imbalance in Multi-Task Learning from Neural Tangent Kernel Perspective
- Shortcut Features as Top Eigenfunctions of NTK: A Linear Neural Network Case and More
- Uncertainty Quantification with the Empirical Neural Tangent Kernel