Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Gaussian Mixture Models with Rare Events

Authors: Xuetong Li, Jing Zhou, Hansheng Wang

JMLR 2024 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The finite sample performance of the proposed method is illustrated by both simulation studies and a real-world dataset of Swedish traffic signs. Extensive simulation studies are presented in this paper to demonstrate the numerical convergence properties of these two algorithms.
Researcher Affiliation Academia Xuetong Li EMAIL Guanghua School of Management Peking University Beijing, China Jing Zhou EMAIL Center for Applied Statistics, School of Statistics Renmin University of China Beijing, China Hansheng Wang EMAIL Guanghua School of Management Peking University Beijing, China
Pseudocode No The paper describes iterative algorithms for EM and MEM using mathematical formulas and prose in sections 2.2 and 3.2, but does not present them in structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statement about the release of source code for the methodology described, nor does it include a link to a code repository.
Open Datasets Yes The dataset used here is the Swedish Traffic Signs (STS) dataset (Larsson and Felsberg, 2011; Larsson et al., 2011), which can be publicly obtained from https://www.cvl.isy.liu.se/research/datasets/traffic-signs-dataset/.
Dataset Splits Yes In the first part, the labels of the samples are treated as if they were missing. The second part is treated as if the labels of the samples were observed. Thereafter, the MEM algorithm can be applied to the training data. For a comprehensive evaluation, six different labeled percentages (i.e., m/(N + m) 100%) are evaluated. They are, respectively, 0%, 5%, 25%, 50%, 75% and 100%. and we randomly partition all labeled data into two parts. The first part contains about 80% of the whole data for training. The remaining 20% part is used for testing.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used to run the experiments, only mentioning the use of a pretrained VGG16 model.
Software Dependencies No The paper mentions applying a pretrained VGG16 model but does not specify any software frameworks, libraries, or their version numbers used in the implementation or for running experiments.
Experiment Setup Yes To evaluate the rare events effect, a total of five different response probabilities are considered. They are 50%, 20%, 10%, 1%, 0.1%, respectively. ... six different percentages of labeled data (i.e., m/(N + m) 100%) are evaluated. They are 0%, 1%, 5%, 10%, 25%, 50%, respectively. ... The experiment is randomly replicated for D = 500 times. and For a comprehensive evaluation, six different labeled percentages (i.e., m/(N + m) 100%) are evaluated. They are, respectively, 0%, 5%, 25%, 50%, 75% and 100%. For a reliable evaluation, the experiment is randomly replicated for D = 20 times.