Asynchronous Proximal Stochastic Gradient Algorithm for Composition Optimization Problems
Authors: Pengfei Wang, Risheng Liu, Nenggan Zheng, Zhefeng Gong1633-1640
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we evaluate our algorithm Async-Prox SCVR on two representative composition optimization problems including value function evaluation in reinforcement learning and sparse mean-variance optimization problem. Experimental results show that the algorithm achieves significant speedups and is much faster than existing compared methods. |
| Researcher Affiliation | Academia | Pengfei Wang,1,2 Risheng Liu,3 Nenggan Zheng,1 Zhefeng Gong4 1Qiushi Academy for Advanced Studies, Zhejiang University, Hangzhou, Zhejiang, China 2College of Computer Science and Technology, Zhejiang University, Hangzhou, Zhejiang, China 3International School of Information Science & Engineering, Dalian University of Technology, Liaoning, China 4Department of Neurobiology, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China |
| Pseudocode | Yes | Algorithm 1: Async-Prox SCVR |
| Open Source Code | No | The paper does not contain any statements about making source code publicly available, nor does it provide any links to code repositories. |
| Open Datasets | No | The paper describes generating data for experiments: "our experiments are conducted on random MDPs." and "The reward vectors are generated through Gaussian distributions with zero mean and positive semi-definite co-variance matrix Σ LT L, where L P Rdˆm and m ă d. Each element of L is drawn from the normal distribution." It does not refer to or provide access information for a publicly available or open dataset. |
| Dataset Splits | No | The paper describes how data for experiments was generated ("random MDPs", "Gaussian distributions") but does not specify explicit training, validation, or test dataset splits or cross-validation details. |
| Hardware Specification | Yes | We conduct all the experiments on an identical server which has 16 Intel(R) Xeon(R) E5-2650 CPUs and 64GB memory. |
| Software Dependencies | No | The paper mentions: "Furthermore, all of the methods are coded in C++ using the standard thread library." However, it does not specify version numbers for C++ compilers, the thread library, or any other critical software dependencies required for reproducibility. |
| Experiment Setup | Yes | For a fair comparison, we use the best tuned learning rates for compared methods, specifically, we select them from t100, 10 1, 10 2, 10 3, 10 4, 10 5u. For variance reduction based algorithms, we set the inner loop size m kn1, in which k is best tuned from t0.5, 1, 2u. The mini-batch sizes of A, B and I are chosen casually as in (Huo, Gu, and Huang 2018). |