Optimal Kernel Quantile Learning with Random Features

Authors: Caixing Wang, Xingdong Feng

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To validate our theoretical findings, simulated experiments and a real data application are conducted. and 5. Numerical Experiments and F. Additional numerical Experiments
Researcher Affiliation Academia 1School of Statistics and Management, Shanghai University of Finance and Economics.
Pseudocode No The paper describes methods like the ADMM algorithm but does not include explicit pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statements about releasing code or links to source code repositories.
Open Datasets Yes In this study, we consider the UK used car prices dataset from Kaggle (https://www.kaggle.com/datasets/ kukuroo3/used-car-price-dataset-competition-format).
Dataset Splits Yes We generate training data with size Ntr = 1000, and testing data with size Nte = 10000. The regularization parameter λ is selected via a grid search based on a validation set with 1000 samples... and In our experimental setup, we randomly choose Ntr = 5000 samples as the training data and assume they are randomly distributed, Nva = 1000 samples as the validation data, and the rest as the testing data.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions algorithms like ADMM and the use of the Gaussian kernel, but does not specify any software versions (e.g., Python, PyTorch, TensorFlow versions, or other libraries with their specific version numbers).
Experiment Setup Yes The regularization parameter λ is selected via a grid search based on a validation set with 1000 samples, where the grid is set as {100.5s : s = 20, 19, . . . , 2}. The number of random features is selected according to Theorem 3.13. and In all the scenarios, we employ the standard Gaussian kernel K(x, x ) = exp( x x 2/2). As suggested by Rahimi & Recht (2007) and Rudi & Rosasco (2017), the corresponding random features are taken as ϕ(x, ω) = 2 cos(ωT x + b), where ω N(0, I) and b U(0, 2π).