The Power of Preconditioning in Overparameterized Low-Rank Matrix Sensing

Authors: Xingyu Xu, Yandi Shen, Yuejie Chi, Cong Ma

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we conduct numerical experiments to demonstrate the efficacy of Scaled GD(λ) for solving overparameterized low-rank matrix sensing.
Researcher Affiliation Academia 1Carnegie Mellon University, Pittsburgh, United States 2University of Chicago, Chicago, United States.
Pseudocode No The paper provides mathematical update rules (e.g., equations 6, 7, 8) but does not include any clearly labeled "Pseudocode" or "Algorithm" blocks.
Open Source Code No The paper does not contain any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes We set the ground truth matrix X = U Σ Rn r where U Rn r is a random orthogonal matrix and Σ Rr r is a diagonal matrix whose condition number is set to be κ. We set n = 150 and r = 3, and use random Gaussian measurements with m = 10nr .
Dataset Splits No No specific dataset splits (e.g., percentages, sample counts, or explicit references to standard splits) for training, validation, or testing are mentioned. The paper describes generating synthetic data for its experiments.
Hardware Specification No No specific hardware details (e.g., CPU, GPU models, memory, or cloud instances) are provided for running the experiments.
Software Dependencies No No specific software components with version numbers (e.g., Python 3.8, PyTorch 1.9) are mentioned.
Experiment Setup Yes We set n = 150 and r = 3, and use random Gaussian measurements with m = 10nr . The learning rate of GD has been fine-tuned to achieve fastest convergence for each κ, while that of Scaled GD(λ) is fixed to 0.3. The initialization scale α in each case has been fine-tuned so that the final accuracy is 10 9.