KFNN: K-Free Nearest Neighbor For Crowdsourcing

Authors: Wenjun Zhang, Liangxiao Jiang, Chaoqun Li

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimental results demonstrate that KFNN significantly outperforms all the other state-of-the-art algorithms and exhibits greater robustness in various crowdsourcing scenarios.
Researcher Affiliation Academia Wenjun Zhang School of Computer Science China University of Geosciences Wuhan 430074, China wjzhang@cug.edu.cn Liangxiao Jiang School of Computer Science China University of Geosciences Wuhan 430074, China ljiang@cug.edu.cn Chaoqun Li School of Mathematics and Physics China University of Geosciences Wuhan 430074, China chqli@cug.edu.cn
Pseudocode Yes Algorithm 1 The learning process of KFNN
Open Source Code Yes Our codes and datasets are available at https://github.com/jiangliangxiao/KFNN.
Open Datasets Yes To evaluate the effectiveness of KFNN, we construct extensive experiments on the whole 34 simulated and two real-world crowdsourced datasets published on the Crowd Environment and its Knowledge Analysis (CEKA) [35] platform. ... The Income dataset... The Leaves dataset...
Dataset Splits No Moreover, label integration does not divide the crowdsourced dataset into training, validation and test sets, which means that KFNN has to determine K i immediately when inferring ˆyi, rather than with a validation phase.
Hardware Specification Yes All experiments are independently repeated ten times on a Windows 10 machine with an AMD Athlon(tm) X4 860K Quad Core Processor @ 3.70 GHz and 16 GB of RAM, and we report the average results of ten experiments.
Software Dependencies No The paper mentions software by name (e.g., "CEKA platform", "implementations provided by their authors", "Waikato Environment and Knowledge Analysis (WEKA)") but does not provide specific version numbers for these software components.
Experiment Setup Yes All parameters of the comparison algorithms are set to the recommended values in the corresponding published papers. In our KFNN, α and β are set to 0.1 and 1 by default.