Alternating Randomized Block Coordinate Descent

Authors: Jelena Diakonikolas, Lorenzo Orecchia

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To illustrate the results, we solve the least squares problem on the Blog Feedback Data Set (Buza, 2014) obtained from UCI Machine Learning Repository (Lichman, 2013). The data set contains 280 attributes and 52,396 data points.
Researcher Affiliation Academia Jelena Diakonikolas 1 Lorenzo Orecchia 1 1Department of Computer Science, Boston University, Boston, MA, USA.
Pseudocode No The paper describes the algorithms (AR-BCD, AAR-BCD) using mathematical notation and textual descriptions of steps, but does not include a formally labeled 'Pseudocode' or 'Algorithm' block or figure.
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository for the described methodology.
Open Datasets Yes To illustrate the results, we solve the least squares problem on the Blog Feedback Data Set (Buza, 2014) obtained from UCI Machine Learning Repository (Lichman, 2013).
Dataset Splits No The paper mentions partitioning the data into blocks (e.g., 'group the first N/n coordinates...into the first block'), but it does not specify explicit training, validation, or test dataset splits with percentages or sample counts for model evaluation.
Hardware Specification No The paper does not provide any specific hardware details (e.g., CPU, GPU models, memory, or cloud instance types) used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes We partition the data into blocks as follows. We first sort the coordinates by their individual smoothness parameters. Then, we group the first N/n coordinates (from the sorted list of coordinates) into the first block, the second N/n coordinates into the second block, and so on. The chosen block sizes N/n are 5, 10, 20, 40, corresponding to n = {56, 28, 14, 7} coordinate blocks, respectively. In all the comparisons between the different methods, we define an epoch to be equal to n iterations. The sampling probabilities for RCDM and AR-BCD are proportional to the block smoothness parameters.