REVIEW 6 cited by
Limitations of the NTK for Understanding Generalization in Deep Learning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The ``Neural Tangent Kernel'' (NTK) (Jacot et al 2018), and its empirical variants have been proposed as a proxy to capture certain behaviors of real neural networks. In this work, we study NTKs through the lens of scaling laws, and demonstrate that they fall short of explaining important aspects of neural network generalization. In particular, we demonstrate realistic settings where finite-width neural networks have significantly better data scaling exponents as compared to their corresponding empirical and infinite NTKs at initialization. This reveals a more fundamental difference between the real networks and NTKs, beyond just a few percentage points of test accuracy. Further, we show that even if the empirical NTK is allowed to be pre-trained on a constant number of samples, the kernel scaling does not catch up to the neural network scaling. Finally, we show that the empirical NTK continues to evolve throughout most of the training, in contrast with prior work which suggests that it stabilizes after a few epochs of training. Altogether, our work establishes concrete limitations of the NTK approach in understanding generalization of real networks on natural datasets.
Forward citations
Cited by 6 Pith papers
-
A Defense of the Quadratic Model
Local Taylor-expanded quadratic models reproduce a 150M-parameter LLM's validation loss for up to 10% of training late in the run, and LLM pretraining operates within a factor of 2 of a stochastic or deterministic edg...
-
The Cost of Discretization in Functional Linear Regression: Minimax Rates and Adaptation
Matching minimax prediction rates for discretely observed functional linear regression are n^{-ν/(ν+1)}+(nm)^{-ν/κ} under independent design, and those two terms plus m^{-ν}+m^{-4α} under common design.
-
Adaptive kernel predictors from feature-learning infinite limits of neural networks
Feature-learning infinite-width neural networks are kernel machines with data-dependent kernels, defined by a min-max saddle point (Bayesian/Langevin) or a DMFT fixed point (gradient flow with weight decay).
-
A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning
In diagonal linear networks, fine-tuning generalization is governed by a tunable per-dimension penalty whose sparsity and pretraining dependence define four regimes and a trade-off between feature reuse and new-featur...
-
Harnessing Optimization Dynamics for Curvature-Informed Model Merging
Optimization Trajectory Aware merging uses Adam second moments as a curvature proxy, first pruning task-vector edits with Fast Fisher Grafting, then reweighting survivors with a compressed curvature preconditioner.
-
Feature learning is decoupled from generalization in high capacity neural networks
Current feature learning measures quantify the magnitude of representation change, which the authors argue is decoupled from the generalization benefit that neural networks show over their neural tangent kernel.
Discussion (0). Continue with ORCID to comment.