DPVI: A Dynamic-Weight Particle-Based Variational Inference Framework

Authors: Chao Zhang, Zhijian Li, Xin Du, Hui Qian

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our algorithms on various synthetic and realworld tasks. The empirical results demonstrate the superiority of our dynamic weight strategy over the fixed weight strategy, and our DPVI algorithms constantly outperform their fixedweight counterparts in all the tasks.
Researcher Affiliation Collaboration Chao Zhang1,2 , Zhijian Li3 , Xin Du3 and Hui Qian1,2 1College of Computer Science and Technology, Zhejiang University 2Alibaba-Zhejiang University Joint Research Institute of Frontier Technologies 3Information Science and Electronic Engineering, Zhejiang University
Pseudocode Yes Algorithm 1 Dynamic-weight Particle-based Variational Inference (DPVI) Framework
Open Source Code No No explicit statement or link to open-source code for the methodology was found.
Open Datasets No The paper mentions using 'LIDAR', 'UCI', and 'LIBSVM' datasets, but does not provide specific links, DOIs, or formal citations with author/year for accessing these datasets as used in the paper. UCI and LIBSVM are large repositories, not specific datasets with concrete access info.
Dataset Splits No The paper does not provide specific percentages, sample counts, or explicit methodologies for training, validation, or test splits. It refers to 'batchsize' for BNN but not data splits.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory specifications) used for running the experiments are provided.
Software Dependencies No The paper mentions the 'POT library' but does not specify its version number or any other software dependencies with version information.
Experiment Setup Yes In this task, we set the particle number to M = 128 for all the algorithms. For all the datasets, we set the batchsize as 128.