The Power of Comparisons for Actively Learning Linear Classifiers

Authors: Max Hopkins, Daniel Kane, Shachar Lovett

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To confirm our theoretical findings, we have implemented a variant of Theorem 3.7 for finite samples. For a given sample size or dimension, the query complexity we present is averaged over 500 trials of the algorithm.
Researcher Affiliation Academia Max Hopkins Dept. of Computer Science and Engineering University of California, San Diego La Jolla, CA 92092 nmhopkin@eng.ucsd.edu Daniel Kane Dept. of Computer Science and Engineering University of California, San Diego La Jolla, CA 92092 dakane@eng.ucsd.edu Shachar Lovett Dept. of Computer Science and Engineering University of California, San Diego La Jolla, CA 92092 slovett@cs.ucsd.edu
Pseudocode Yes Algorithm 1: Perfect-Learning(N, Q, d, c)
Open Source Code No The paper states it implemented a variant of Theorem 3.7, but does not provide any concrete access to source code for its methodology, nor does it explicitly state that the code is being released.
Open Datasets No The paper states "our algorithm labels finite samples drawn from the uniform distribution over Bd." However, it does not provide concrete access information (link, DOI, formal citation) for a publicly available or open dataset.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes For a given sample size or dimension, the query complexity we present is averaged over 500 trials of the algorithm. We first note a few practical modifications. First, our algorithm labels finite samples drawn from the uniform distribution over Bd. Second, to match our methodology in lower bounding Label-Pool-RPU learning, we will draw our classifier uniformly from hyperplanes tangent to the unit ball. Finally, because the true inference dimension of the sample might be small, our algorithm guesses a low potential inference dimension to start, and doubles its guess on each iteration with low coverage.