Nonparanormal Information Estimation

Authors: Shashank Singh, Barnabás Póczos

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 7. Empirical Results
Researcher Affiliation Academia 1Carnegie Mellon University, Pittsburgh, USA. Correspondence to: Shashank Singh <sss1@andrew.cmu.edu>.
Pseudocode No The paper describes the methods verbally but does not include any formally structured pseudocode or algorithm blocks.
Open Source Code Yes All code can be found on Git Hub4. https://github.com/sss1/nonparanormal-information
Open Datasets No The paper generates synthetic data for its experiments, rather than using a pre-existing publicly available dataset. 'In each trial, a correlation matrix Σ was drawn by normalizing a random covariance matrix from a Wishart distribution, and data X1, ..., Xn i.i.d. N(0, Σ) drawn.'
Dataset Splits No The paper describes generating synthetic i.i.d. samples but does not specify train/validation/test splits as it's not a typical machine learning training setup.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions 'MATLAB source code is available', but does not specify the version of MATLAB or any other software dependencies with version numbers.
Experiment Setup Yes Except as specified otherwise, each experiment had the following basic structure: In each trial, a correlation matrix Σ was drawn by normalizing a random covariance matrix from a Wishart distribution, and data X1, ..., Xn i.i.d. N(0, Σ) drawn. All 5 estimators were computed from X1, ..., Xn and squared error from true mutual information (computed from Σ) was recorded. Unless specified otherwise, n = 100 and D = 25. For Iρ and Iτ, we used a regularization constant z = 10 3. For Ik NN, except as noted in Experiment 3, k = 2, based on recent analysis (Singh & P oczos, 2016b) suggesting that small values of k are best for estimation.