Crowdsourcing via Tensor Augmentation and Completion

Authors: Yao Zhou, Jingrui He

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on 6 real data sets demonstrate the superior performance of the proposed approach over state-of-the-art techniques.
Researcher Affiliation Academia Yao Zhou, Jingrui He Arizona State University, Tempe, Arizona yzhou174@asu.edu, jingrui.he@asu.edu
Pseudocode Yes Algorithm 1: PG-TAC
Open Source Code No The paper does not provide an explicit statement or link to its own open-source code. It mentions verifying baselines using an 'open source implementation provided by [Zhou et al., 2015]' but this is not for their described methodology.
Open Datasets Yes We also evaluate our methods with various other methods on six real world crowdsourcing data sets. [...] Temp data set [Snow et al., 2008]), RTE data set [Snow et al., 2008] and Spam data set [Zhou et al., 2015]. The multi-class labeling data sets include Dog data set [Zhou et al., 2012], Web data set [Zhou et al., 2012] and Age data set [Han et al., 2015].
Dataset Splits No For each possible parameter pair on the searching grid, a subset of worker labels is randomly chosen from current data set without replacement. In practice, we empirically choose 90 percent of worker labels as a subset, run our methods, and evaluate the performance. Then we repeat the same procedure ten times for each possible parameter pair on the grid.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using 'open source implementation provided by [Zhou et al., 2015]' for verifying baselines, but it does not specify version numbers for any software dependencies relevant to its own methodology.
Experiment Setup Yes Eventually we can apply the grid search on two regularization parameters δl and γ, and the procedure is described as follows: all data sets we used are publicly available online and they all come with ground truth labels.