Kernel Alignment Risk Estimator: Risk Prediction from Training Data
Authors: Arthur Jacot, Berfin Simsek, Francesco Spadaro, Clement Hongler, Franck Gabriel
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We numerically investigate our findings on the Higgs and MNIST datasets for various classical kernels: the KARE gives an excellent approximation of the risk, thus supporting our universality assumption. |
| Researcher Affiliation | Academia | Arthur Jacot Ecole Polytechnique Fédérale de Lausanne arthur.jacot@epfl.ch Berfin Sim sek Ecole Polytechnique Fédérale de Lausanne berfin.simsek@epfl.ch Francesco Spadaro Ecole Polytechnique Fédérale de Lausanne francesco.spadaro@epfl.ch Clément Hongler Ecole Polytechnique Fédérale de Lausanne clement.hongler@epfl.ch Franck Gabriel Ecole Polytechnique Fédérale de Lausanne franck.gabriel@epfl.ch |
| Pseudocode | No | The paper provides mathematical derivations, theorems, and proofs but does not include any pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about open-sourcing the code for the described methodology, nor does it include links to a code repository. |
| Open Datasets | Yes | We numerically investigate our findings on the Higgs and MNIST datasets for various classical kernels: the KARE gives an excellent approximation of the risk, thus supporting our universality assumption. ... (a) MNIST, ℓ= d ... (c) Higgs, ℓ= d |
| Dataset Splits | No | The paper mentions using N=2000 for MNIST and N=1000 for Higgs (and N=200 for another MNIST experiment) as total data points, but it does not specify how these datasets were split into training, validation, or test sets using percentages, counts, or references to predefined splits. |
| Hardware Specification | No | The paper does not contain any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud resources) used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., Python, PyTorch, specific libraries or solvers) that would be needed to replicate the experiments. |
| Experiment Setup | Yes | We numerically investigate our findings on the Higgs and MNIST datasets for various classical kernels: the KARE gives an excellent approximation of the risk, thus supporting our universality assumption. ... (a) MNIST, ℓ= d ... (b) MNIST, λ = 10 5 ... (c) Higgs, ℓ= d ... (d) Higgs, λ = 10 4 ... We calculate the risk (i.e. test error) of ˆf ϵ λ on MNIST with the RBF Kernel for various values of ℓand λ on N = 200 data points (same setup as Fig. 1). |