Joint quantile regression in vector-valued RKHSs

Authors: Maxime Sangnier, Olivier Fercoq, Florence d'Alché-Buc

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments on benchmark and real datasets highlight the enhancements of our approach regarding the prediction error, the crossing occurrences and the training time.
Researcher Affiliation Academia Maxime Sangnier Olivier Fercoq Florence d Alch e-Buc LTCI, CNRS, T el ecom Paris Tech Universit e Paris-Saclay 75013, Paris, France
Pseudocode Yes Algorithm 1 Primal-Dual Coordinate Descent.
Open Source Code No The paper mentions using CVXOPT, a third-party tool, but does not provide any link or explicit statement about releasing the source code for their own methodology or implementation. The footnotes point to data sources, not code.
Open Datasets Yes To present an honorable comparison of these four methods, we did not choose datasets for the benefit of our method but considered the ones used in [26]. These 20 datasets (whose names are indicated in Table 1) come from the UCI repository and three R packages: quantreg, alr3 and MASS. ... Data are available at www.census.gov/census2000/PUMS5.html and www.nber.org/data/ vital-statistics-natality-data.html.
Dataset Splits Yes Results are given in Table 1 thanks to the mean and the standard deviation of the test losses recorded on 20 random splits train-test with ratio 0.7-0.3.
Hardware Specification No The paper mentions 'CPU time' in Table 2, but it does not specify any details about the CPU model, GPU, or any other hardware components used for the experiments.
Software Dependencies Yes [2] M.S. Anderson, J. Dahl, and L. Vandenberghe. CVXOPT: A Python package for convex optimization, version 1.1.5., 2012.
Experiment Setup Yes Quantile levels of interest are τ = (0.1, 0.3, 0.5, 0.7, 0.9). ... The parameter C is chosen by cross-validation (minimizing the pinball loss) inside a logarithmic grid (10 5, 10 4, . . . , 105) for all methods and datasets. For our approach (JQR), the parameter γ is chosen in the same grid as C with extra candidates 0 and + . ... Parameters for the models are: (C, γ) = (102, 10 2).