Debiasing Conditional Stochastic Optimization

Authors: Lie He, Shiva Kasiviswanathan

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we demonstrate the numerical performance of our proposed algorithms. We focus on the application of invariant logistic regression here. In Appendix F, we discuss performance of our proposed algorithms on other common CSO/FCCO applications.
Researcher Affiliation Collaboration Lie He EPFL lie.he@epfl.ch Shiva Prasad Kasiviswanathan Amazon kasivisw@gmail.com
Pseudocode Yes Algorithm 1 E-Nested VR; Algorithm 2 E-BSGD; Algorithm 3 E-BSpider Boost
Open Source Code No The paper does not contain an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets No The paper describes generating synthetic datasets for experiments (e.g., "We generate a synthetic dataset with d = 10 dimensions" in Section 5.1 and "We generate the data similar to [31]" in Section 5.2) but does not mention using or providing access to any publicly available or open datasets.
Dataset Splits No The paper describes the data generation process for its experiments but does not provide specific training, validation, or test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory specifications, or cloud instance types) used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes In each pair of experiments, the samples used per iteration are fixed same, i.e.: 1) BSGD uses m = 2 and E-BSGD uses m = 1; 2) For BSpider Boost and E-BSpider Boost, we use cycle length of 10, small batch and large batch in Spider to be 10 and 100 respectively, and we choose inner batch sizes m = 2 for BSpider Boost and m = 1 for E-BSpider Boost; 3) For Nested VR and E-Nested VR, we fix the outer batch size to 10 and 5 respectively, and choose fix the inner Spider Cycle to be 10 with large batch 100 and small batch 10.