Projection-Free Variance Reduction Methods for Stochastic Constrained Multi-Level Compositional Optimization
Authors: Wei Jiang, Sifan Yang, Wenhao Yang, Yibo Wang, Yuanyu Wan, Lijun Zhang
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, numerical experiments on different tasks demonstrate the effectiveness of our methods. and In this section, we evaluate the effectiveness of our proposed methods through numerical experiments on three different problems. |
| Researcher Affiliation | Academia | 1National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China 2School of Artificial Intelligence, Nanjing University, Nanjing, China 3School of Software Technology, Zhejiang University, Ningbo, China. |
| Pseudocode | Yes | Algorithm 1 PMVR Algorithm and Algorithm 2 PMVR-v2 and Algorithm 3 Stage-wise PMVR and Algorithm 4 Stage-wise PMVR-v2 |
| Open Source Code | No | The paper does not contain any statement about making its source code publicly available or providing a link to a repository. |
| Open Datasets | Yes | For experimental validation, we utilize real-world datasets Industry-10 and Industry-12 from the Kenneth R. French Data Library^2. ^2https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ |
| Dataset Splits | No | The paper mentions using 'real-world datasets Industry-10 and Industry-12' but does not specify any training, validation, or test dataset splits or cross-validation setup. |
| Hardware Specification | Yes | All experiments are conducted on a personal laptop. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers. |
| Experiment Setup | Yes | For our algorithm, we select the momentum parameter α from the set {0.01, 0.03, 0.05, 0.1, 0.3} and search the parameter N for PMVR-v2 from the range {10, 50, 100}. For the other methods, we adopt the hyper-parameters recommended in their original papers or perform a grid search to select the best ones. The learning rate is fine-tuned within the range of {0.001, 0.005, 0.01, 0.05, 0.1}. |