A Precise Performance Analysis of Support Vector Regression

Authors: Houssem Sifaou, Abla Kammoun, Mohamed-Slim Alouini

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Particularly, we show numerically that the double descent behavior appears only when the H-SVR or S-SVR parameters are not properly tuned. Such a behavior reminds the recent findings in (Nakkiran et al., 2020b) that suggest that unregularized models often suffer from the sample-wise double descent phenomenon, while optimally tuned models usually present a monotonic risk with respect to the number of samples.
Researcher Affiliation Academia 1Computer, Electrical, and Mathematical Sciences & Engineering Division, King Abdullah University of Science and Technology (KAUST), Thuwal, Saudi Arabia. Correspondence to: Houssem Sifaou <houssem.sifaou@kaust.edu.sa>.
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository for the methodology described.
Open Datasets No The paper describes generating synthetic data ('standard normal distribution') and does not provide concrete access information for a public dataset.
Dataset Splits No The paper does not mention using a validation set, nor does it specify any train/validation/test splits or cross-validation methodology.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running its experiments.
Software Dependencies No The paper does not provide any specific software dependencies with version numbers.
Experiment Setup Yes In a first experiment, we investigate the behavior of the test risk of H-SVR as a function of the number of samples for different values of the signal power β2 = β 2. Particularly, for each β {0.5, 1, 2}, we fix the noise variance σ2 and ϵ and plot the test risk and cosine similarity over the range [0, δ ] over which the H-SVR is feasible. Fig. 2 represents the theoretical results along with their empirical averages obtained for p = 200 and n = δp .