Kernel Truncated Randomized Ridge Regression: Optimal Rates and Low Noise Acceleration

Authors: Kwang-Sung Jun, Ashok Cutkosky, Francesco Orabona

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In Section 5, we empirically evaluate our findings. Finally, Section 6 discusses open problems and future directions of research.
Researcher Affiliation Collaboration Kwang-Sung Jun The University of Arizona kjun@cs.arizona.edu Ashok Cutkosky Google Research ashok@cutkosky.com Francesco Orabona Boston University francesco@orabona.com
Pseudocode Yes Algorithm 1 KTR3: Kernel Truncated Randomized Ridge Regression
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets No The paper describes a synthetic dataset generation process (e.g., "We consider the uniform distribution ρX on X r0, 1s and define the target function to be f pxq Λ β 2 px, 0q for x P X. We define the observed response of x to be f pxq B where B is a uniform random variable r ϵ, ϵs") rather than using or providing access to a publicly available dataset.
Dataset Splits No The paper mentions drawing "n training points" and estimating excess risk by a "test set" but does not specify the splitting percentages, sample counts, or methodology for creating distinct training, validation, and test splits.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory, cloud instances) used for running the experiments.
Software Dependencies No The paper does not list any software or library names with specific version numbers.
Experiment Setup Yes For each n in fine-grained grid points in r102, 103s and λ in another fine-grained set of numbers, we draw n training points, compute fn by Algorithm 1, and estimate its excess risk by a test set. Finally, for each n we choose the λ that minimizes the average excess risk. We repeat the same 5 times. First, we set b 1/8 and β 7/16, and ϵ 0.1. [...] To verify our improved rate in the regime 2β b < 1, we also consider the case of β 1/6, and ϵ 0.1.