Gradient-Free Methods for Nonconvex Nonsmooth Stochastic Compositional Optimization

Authors: Zhuanghua Liu, Luo Luo, Bryan Kian Hsiang Low

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

Reproducibility Variable Result LLM Response
Research Type Experimental Furthermore, we conduct numerical experiments to demonstrate the effectiveness of the proposed methods.
Researcher Affiliation Academia Zhuanghua Liu Department of Computer Science, National University of Singapore CNRS@CREATE LTD, 1 Create Way, #08-01 CREATE Tower, Singapore 138602 liuzhuanghua9@gmail.com Luo Luo School of Data Science, Fudan University Shanghai Key Laboratory for Contemporary Applied Mathematics luoluo@fudan.edu.cn Bryan Kian Hsiang Low Department of Computer Science, National University of Singapore lowkh@comp.nus.edu.sg
Pseudocode Yes Algorithm 1: GFCOM(x0, η, T, bf, bg) ... Algorithm 2: GFCOM+(x0, η, T, bf, b f, bg, b g, m) ... Algorithm 3: WS-GFCOM(x0, η0, T0, bg,0, η, T, bf, bg, b f, b g, m)
Open Source Code No We are clearing the code with internal compliance and will release it upon approval.
Open Datasets Yes We compare all the methods on 6 different portfolio datasets formed on Size and Operating Profitability2. 2http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
Dataset Splits No The paper does not explicitly provide details about training, validation, or test dataset splits, percentages, or methodologies for splitting.
Hardware Specification No The paper does not mention any specific hardware (e.g., GPU/CPU models, memory, or cloud instance types) used for running the experiments.
Software Dependencies No The paper does not list any specific software dependencies with version numbers.
Experiment Setup Yes We set δ = 0.1 for the GFCOM and GFCOM+ methods. ... For all algorithms, we tune the stepsize among {1 10 5, 3 10 5, . . . , 1 10 3, 3 10 3}. ... We choose the mini-batch size bf = bg = 1000. In addition, we set b f = 100, b g = 1000 and m = bf/b f = 10 for the GFCOM+ algorithm.