Multi-Objective Online Learning
Authors: Jiyan Jiang, Wenpeng Zhang, Shiji Zhou, Lihong Gu, Xiaodong Zeng, Wenwu Zhu
ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on several real-world datasets verify the effectiveness of the proposed algorithm. |
| Researcher Affiliation | Collaboration | Jiyan Jiang Tsinghua University Beijing, China scjjy95@outlook.com Wenpeng Zhang Ant Group Beijing, China zhangwenpeng0@gmail.com Shiji Zhou Tsinghua University Beijing, China zhoushiji00@gmail.com Lihong Gu, Xiaodong Zeng Ant Group Hangzhou, China {lihong.glh,xiaodong.zxd}@antgroup.com Wenwu Zhu Tsinghua University Beijing, China wwzhu@tsinghua.edu.cn |
| Pseudocode | Yes | Algorithm 1 Doubly Regularized Online Mirror Multiple Descent (DR-OMMD) |
| Open Source Code | No | The paper does not provide any explicit statements or links indicating that the source code for the described methodology is open-source or publicly available. |
| Open Datasets | Yes | We use two large-scale online benchmark datasets. (i) protein is a bioinformatics dataset for protein type classification (Wang, 2002), which has 17 thousand instances with 357 features. (ii) covtype is a biological dataset collected from a non-stationary environment for forest cover type prediction (Blackard & Dean, 1999), which has 50 thousand instances with 54 features. We use Multi MNIST (Sabour et al., 2017) |
| Dataset Splits | No | The paper does not provide specific numerical train/validation/test dataset splits. It states: 'In the online setting, samples arrive in a sequential manner, which is different from offline experiments where sample batches are randomly sampled from the training set.' |
| Hardware Specification | Yes | All runs are deployed on Xeon(R) E5-2699 @ 2.2GHz. |
| Software Dependencies | No | The paper mentions software components like 'Le Net' and 'Adam' but does not specify version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | The learning rates are decided by a grid search over {0.1, 0.2, . . . , 3.0}. For DR-OMMD, the parameter αt is simply set as 0.1. For linearization, we examine different weights (0.25, 0.75), (0.5, 0.5), and (0.75, 0.25). Learning rates in all methods are selected via grid search over {0.0001, 0.001, 0.01, 0.1}. For DR-OMMD, αt is set according to Theorem 1, and the initial weights are simply set as λ0 = (0.5, 0.5). |