Implicit Regularization in Over-Parameterized Support Vector Machine

Authors: Yang Sui, Xin HE, Yang Bai

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Additionally, we verify our theoretical findings through a variety of numerical experiments and compare the proposed method with explicit regularization. Our results illustrate the advantages of employing implicit regularization via gradient descent in conjunction with over-parameterization in sparse SVMs.4 Numerical Study
Researcher Affiliation Academia Yang Sui Xin He Yang Bai School of Statistics and Management Shanghai University of Finance and Economics suiyang1027@stu.sufe.edu.cn;he.xin17@mail.shufe.edu.cn; statbyang@mail.shufe.edu.cn
Pseudocode Yes Algorithm 1: Gradient Descent Algorithm for High-Dimensional Sparse SVM.
Open Source Code No The paper does not contain any statement about making the source code publicly available or providing a link to a code repository.
Open Datasets No In our simulations, unless otherwise specified, we follow a default setup. We generate 3n independent observations, divided equally for training, validation, and testing. The true parameters β is set to m1S with a constant m. Each entry of x is sampled as i.i.d. zero-mean Gaussian random variable, and the labels y are determined by a binomial distribution with probability p = 1/(1 + exp(x T β )).
Dataset Splits Yes We generate 3n independent observations, divided equally for training, validation, and testing.
Hardware Specification No The paper does not specify any hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not specify software dependencies with version numbers, such as 'Python 3.8' or 'PyTorch 1.9'.
Experiment Setup Yes Default parameters are: true signal strength m = 10, number of signals s = 4, sample size n = 200, dimension p = 400, step size η = 0.5, smoothness parameter γ = 10 4, and initialization size α = 10 8.