Semidefinite Programming versus Burer-Monteiro Factorization for Matrix Sensing
Authors: Baturalp Yalçın, Ziye Ma, Javad Lavaei, Somayeh Sojoudi
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The results are presented in Tables 1 and 2. We generated synthetic matrix completion problems with symmetric positive semidefinite ground truth matrices with dimensions n = 5, n = 10, and n = 20. An entry of the ground truth matrix is observed independently with probability p. 1000 different instances are constructed for each p and n. The BM factorization problem is solved with a gradient descent algorithm and SDP formulation is solved by using the CVX solver. We report the percentage of times each method successfully recovers the ground truth matrix and we take the average of the success rate for various problem dimensions n. |
| Researcher Affiliation | Academia | Baturalp Yalçın*1, Ziye Ma*2, Javad Lavaei1, Somayeh Sojoudi2 1 UC Berkeley, Industrial Engineering and Operations Research 2 UC Berkeley, Electrical Engineering and Computer Science |
| Pseudocode | No | The paper describes algorithms and methods but does not include any pseudocode blocks or figures explicitly labeled 'Algorithm' or 'Pseudocode'. |
| Open Source Code | No | The paper does not contain any statement about releasing source code or provide any links to a code repository. |
| Open Datasets | No | We generated synthetic matrix completion problems with symmetric positive semidefinite ground truth matrices with dimensions n = 5, n = 10, and n = 20. An entry of the ground truth matrix is observed independently with probability p. 1000 different instances are constructed for each p and n. |
| Dataset Splits | No | The paper describes generating synthetic data and running experiments but does not specify any training, validation, or test splits. It only mentions constructing 1000 instances. |
| Hardware Specification | No | The paper describes the computational methods (gradient descent, CVX solver) but does not mention any specific hardware used for running these experiments (e.g., GPU/CPU models, memory). |
| Software Dependencies | No | The paper mentions that 'SDP formulation is solved by using the CVX solver.' However, it does not provide a specific version number for the CVX solver or any other software dependencies. |
| Experiment Setup | No | The paper mentions that 'The BM factorization problem is solved with a gradient descent algorithm and SDP formulation is solved by using the CVX solver.' However, it does not provide specific hyperparameters for gradient descent (e.g., learning rate, batch size) or other detailed training configurations. |