Near Optimal Frequent Directions for Sketching Dense and Sparse Matrices

Authors: Zengfeng Huang

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

Reproducibility Variable Result LLM Response
Research Type Theoretical In this paper, we almost settle the time complexity of this problem. In particular, we provide new space-optimal algorithms with faster running times. Moreover, we also show that the running times of our algorithms are near-optimal unless the state-of-the-art running time of matrix multiplication can be improved significantly. We provide new space-optimal algorithms with improved running time. We also prove that our running times cannot be significantly improved unless the state-of-the-art matrix multiplication algorithms can. Thus, we almost settle the time complexity of this problem.
Researcher Affiliation Academia Zengfeng Huang 1 School of Data Science, Fudan University, China. Correspondence to: Zengfeng Huang <huangzf@fudan.edu.cn>.
Pseudocode Yes Algorithm 1 Dense Shrink, Algorithm 2 Dense Shrink R, Algorithm 3 FDShrink, Algorithm 4 FFDdense, Algorithm 5 FFDsparse
Open Source Code No The paper does not contain any explicit statements or links indicating that source code for the described methodology is publicly available.
Open Datasets No This paper is theoretical and focuses on algorithm design and complexity analysis. It does not use specific publicly available datasets for empirical evaluation. The input is described as a large matrix A R^n d, but no concrete dataset name or access information is provided.
Dataset Splits No The paper is theoretical and does not describe empirical experiments on datasets; therefore, it does not provide information about training/validation/test dataset splits.
Hardware Specification No The paper is theoretical and focuses on algorithmic complexity, not empirical execution. Therefore, no hardware specifications for experiments are mentioned.
Software Dependencies No The paper describes theoretical algorithms and does not mention any specific software dependencies or version numbers needed to replicate the work.
Experiment Setup No The paper is theoretical and focuses on algorithm design and analysis, rather than empirical experiments. As such, it does not provide details about experimental setup, hyperparameters, or training settings.