Quasi-Monte Carlo Features for Kernel Approximation

Authors: Zhen Huang, Jiajin Sun, Yian Huang

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

Reproducibility Variable Result LLM Response
Research Type Experimental In practice, the QMC kernel approximation approach is easily implementable and shows superior performance, as supported by the empirical evidence provided in the paper.
Researcher Affiliation Academia Zhen Huang 1 Jiajin Sun 1 Yian Huang 1 1Department of Statistics, Columbia University, New York, NY 10027, USA.
Pseudocode No The paper describes methods and theoretical derivations but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statements about releasing source code or provide links to a code repository for the described methodology.
Open Datasets Yes We consider two choices of kernels: (i) the min kernel K(x, x ) = Qd i=1 min(xi, x i), and (ii) the Gaussian kernel K(x, x ) = exp( 1 2σ2 x x 2 2)... Cadata (Pace & Barry, 1997): In this data set (n = 20640, d = 6)... Cod-rna (Uzilov et al., 2006): This benchmark dataset (n = 59535 (train) + 271617 (test), d = 8)...
Dataset Splits No The paper mentions 'training samples' and 'test data points' with specific sizes for synthetic data, and 'random train-test split, allocating 25% of the data to the test set' for Cadata, and explicit train/test sizes for Cod-rna, but does not explicitly describe a separate validation split.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments, such as CPU/GPU models, memory, or specific cloud instance types. It only mentions general terms like 'training samples' and 'data points'.
Software Dependencies Yes Halton sequence implemented in the Sci Py package in Python (Virtanen et al., 2020) is used.
Experiment Setup Yes The training and test data are generated from Y = f(X) + ε, where f is the regression function, X Unif[0, 1]d, and ε N(0, 1). We consider two choices of kernels: (i) the min kernel K(x, x ) = Qd i=1 min(xi, x i), and (ii) the Gaussian kernel K(x, x ) = exp( 1 2σ2 x x 2 2), with the bandwidth σ set as the median of X X (computed numerically)... The kernel ridge regularization parameter is set as λ = 0.25n 1 2r+1 .