Closed-form Estimators for High-dimensional Generalized Linear Models

Authors: Eunho Yang, Aurelie C. Lozano, Pradeep K. Ravikumar

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

Reproducibility Variable Result LLM Response
Research Type Experimental We corroborate the surprising statistical and computational performance of our class of estimators via extensive simulations.
Researcher Affiliation Collaboration Eunho Yang IBM T.J. Watson Research Center eunhyang@us.ibm.com Aurelie C. Lozano IBM T.J. Watson Research Center aclozano@us.ibm.com Pradeep Ravikumar University of Texas at Austin pradeepr@cs.utexas.edu
Pseudocode No The paper does not contain any pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any statement or link indicating that open-source code for the methodology is available.
Open Datasets No The paper uses simulated data, which it describes how to generate, rather than a publicly available or open dataset for which access information is provided.
Dataset Splits Yes While our theorem specified an optimal setting of the regularization parameter λn and , this optimal setting depended on unknown model parameters. Thus, as is standard with high-dimensional regularized estimators, we set tuning parameters λn = c log p/n and = c0p log p/n by a holdoutvalidated fashion; finding a parameter that minimizes the 2 error on an independent validation set.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers.
Experiment Setup Yes We compare against standard 1 regularized MLE estimators with iteration bounds of 50, 100, and 500, denoted by 1 MLE1, 1 MLE2 and 1 MLE3 respectively. Noting that our theoretical results were not sensitive to the setting of in M(y), we simply report the results when = 10 4 across all experiments. we set tuning parameters λn = c log p/n and = c0p log p/n by a holdoutvalidated fashion; finding a parameter that minimizes the 2 error on an independent validation set.