ReHLine: Regularized Composite ReLU-ReHU Loss Minimization with Linear Computation and Linear Convergence

Authors: Ben Dai, Yixuan Qiu

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

Reproducibility Variable Result LLM Response
Research Type Experimental The algorithm is implemented with both Python and R interfaces, and its performance is benchmarked on various tasks and datasets. Our experimental results demonstrate that Re HLine significantly surpasses generic optimization solvers in terms of computational efficiency on large-scale datasets.
Researcher Affiliation Academia Ben Dai Department of Statistics The Chinese University of Hong Kong bendai@cuhk.edu.hk Yixuan Qiu , School of Statistics and Management Shanghai University of Finance and Economics qiuyixuan@sufe.edu.cn
Pseudocode Yes Algorithm 1: The Re HLine algorithm that solves (4).
Open Source Code Yes The source code, project page, accompanying software, and the Python/R interface can be accessed through the link: https://github.com/softmin/Re HLine.
Open Datasets Yes Specifically, we focus on four classification datasets and five regression datasets sourced from Open ML (https://www.openml.org/)...
Dataset Splits No No specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning was found.
Hardware Specification No No specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments were found.
Software Dependencies No The algorithm is implemented with both Python and R interfaces, and its performance is benchmarked on various tasks and datasets. To achieve a fair comparison, we use a well-organized toolset and framework, the BENCHOPT library [29], to implement optimization benchmarks for all the SOTA solvers.
Experiment Setup Yes Moreover, we examine the performance of Elastic QR by considering the model defined in (A.2) with λ1 = λ2 = 1, Ridge Huber of (A.4) with λ1 = 0, λ2 = 1, and SVM of (A.1) and s SVM of (A.5) with C = 1.