Training a neural network's tangent kernel with the KARE risk estimate produces a kernel predictor that matches or beats the network itself and its after-training kernel on several benchmarks.
For datasets with fewer than 100 observations, we standardize the data using RobustScaler() from sklearn
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Training NTK to Generalize with KARE
Training a neural network's tangent kernel with the KARE risk estimate produces a kernel predictor that matches or beats the network itself and its after-training kernel on several benchmarks.