Local Context Sparse Coding

Authors: Seungyeon Kim, Joonseok Lee, Guy Lebanon, Haesun Park

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 ExperimentsWe examine in this section using features generated by LCSC in classification. We used a standard classifier, support vector machine1, with different sets of features. Specifically, we used ν-SVM whose ν value was selected from 10 candidate values using cross-validation. Our classification task was the standard 20 newsgroup2 classification data with the official train-test split and standard preprocessing
Researcher Affiliation Collaboration College of Computing, Georgia Institute of Technology, Atlanta, GA, USA Amazon, Seattle, WA, USA
Pseudocode Yes Algorithm 1 Greedy coordinate descent for β and z
Open Source Code No The paper does not provide an explicit statement or link to the open-source code for the Local Context Sparse Coding (LCSC) methodology described.
Open Datasets Yes Our classification task was the standard 20 newsgroup2 classification data with the official train-test split and standard preprocessing: lowercasing, stripping non-english characters, tokenizing sentences and words, Porter stemming, and removing rare features and stop words. (Footnote 2 points to http://qwone.com/~jason/20Newsgroups/)
Dataset Splits Yes Specifically, we used ν-SVM whose ν value was selected from 10 candidate values using cross-validation. Our classification task was the standard 20 newsgroup2 classification data with the official train-test split and standard preprocessing
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory, or cloud instance types).
Software Dependencies No The paper mentions using 'support vector machine' (referencing LibSVM), 'Fast LDA', and 'STC' but does not provide specific version numbers for these software dependencies or any other ancillary software.
Experiment Setup Yes Specifically, we used ν-SVM whose ν value was selected from 10 candidate values using cross-validation. Our classification task was the standard 20 newsgroup2 classification data with the official train-test split and standard preprocessing: lowercasing, stripping non-english characters, tokenizing sentences and words, Porter stemming, and removing rare features and stop words. ... The bandwidth of LCSC was fixed to h=1, which covers about 7 words ( 3h). We tried a set of candidates for the remaining parameters and chose the best performing one (for example, λ = {10 4, 10 2, 10 1, 0.5, 1} for STC). ... Figure 5 shows test set classification accuracies of LCSC with various bandwidths while other parameters are fixed (K=4000, ρ=10 4, λ=10 2).