Spectral Methods for Supervised Topic Models

Authors: Yining Wang, Jun Zhu

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Thorough experiments on a diverse range of synthetic and real-world datasets verify the theory and demonstrate the practical effectiveness of the algorithm.
Researcher Affiliation Academia Machine Learning Department, Carnegie Mellon University, yiningwa@cs.cmu.edu Dept. of Comp. Sci. & Tech.; Tsinghua National TNList Lab; State Key Lab of Intell. Tech. & Sys., Tsinghua University, dcszj@mail.tsinghua.edu.cn
Pseudocode Yes Algorithm 1 spectral parameter recovery algorithm for s LDA.
Open Source Code No The paper states, 'Both Gibbs sampling for the s LDA model in Fig. 1 (b) and the proposed spectral recovery algorithm are implemented in C++,' but it does not provide any link or explicit statement about the availability of the source code.
Open Datasets Yes For real-world data, we use the large-scale dataset built on Amazon movie reviews [16] to demonstrate the practical effectiveness of our algorithm.
Dataset Splits Yes Error bars indicate the standard deviation of 5-fold cross-validation.
Hardware Specification No The paper does not specify any hardware details such as GPU/CPU models or memory used for running the experiments.
Software Dependencies No The paper states that the algorithms are 'implemented in C++' but does not provide specific version numbers for C++ compilers or any other libraries/software dependencies.
Experiment Setup Yes For our spectral algorithm, the hyperparameters L and T are set to 100, which is sufficiently large for all settings in our experiments. The hyper-parameters used in our Gibbs sampling implementation are the same with the ones used to generate the datasets.