The Statistical Cost of Robust Kernel Hyperparameter Turning

Authors: Raphael Meyer, Christopher Musco

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical This paper studies the statistical complexity of kernel hyperparameter tuning in the setting of active regression under adversarial noise. We provide finite-sample guarantees for the problem, characterizing how increasing the complexity of the kernel class increases the complexity of learning kernel hyperparameters.
Researcher Affiliation Academia Raphael A. Meyer Tandon School of Engineering New York University Brooklyn, NY 11201 ram900@nyu.edu Christopher Musco Tandon School of Engineering New York University Brooklyn, NY 11201 cmusco@nyu.edu
Pseudocode Yes Algorithm 1 Time Point Sampling input: Kernel functions kµ1, . . . kµQ, non-negative function p(t) on [0, T] with known integral P = R T 0 p(t)dt, number of samples n. output: Times t1, . . . , tn [0, T], weights v1, . . . , vn, PSD matrices Kµ1, . . . , KµQ Cn n. Algorithm 2 Computing the Interpolant input: Time points t1, . . . , tn [0, T], weights v1, . . . , vn, PSD matrix Kµ Cn n, regularization parameter ε > 0. ouput: Reconstructed function y, represented implicitly
Open Source Code No The paper does not provide any concrete access to source code for the methodology described.
Open Datasets No The paper is theoretical and focuses on statistical complexity and sample guarantees. It does not use or specify any publicly available datasets for empirical evaluation.
Dataset Splits No The paper is theoretical and does not discuss training, validation, or test splits, as it does not involve empirical experiments with datasets.
Hardware Specification No The paper is theoretical and does not describe any hardware used for running experiments.
Software Dependencies No The paper is theoretical and does not mention any specific software dependencies with version numbers needed to replicate experiments.
Experiment Setup No The paper is theoretical and provides algorithms and proofs, but does not detail a concrete experimental setup with hyperparameters for an empirical evaluation.