Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Multi-task learning with summary statistics
Authors: Parker Knight, Rui Duan
NeurIPS 2023 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate our theoretical findings and the performance of the method through extensive simulations. |
| Researcher Affiliation | Academia | Parker Knight Department of Biostatistics Harvard University Boston, MA EMAIL Rui Duan Department of Biostatistics Harvard University Boston, MA EMAIL |
| Pseudocode | No | The paper describes the methods through mathematical formulations and textual explanations but does not include any pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code, further implementation details, and additional simulations which explore the use of our adaptive tuning procedure are also available in the supplement. |
| Open Datasets | Yes | We use a multi-site data obtained from the electronic Medical Records and Genomics (e MERGE) network [28], which includes individual-level genotype data from multiple research sites in the United States. |
| Dataset Splits | No | We split the data (with sample sizes n1 = 3813, n2 = 546, n3 = 2666, n4 = 1435, n5 = 525) at each task into a training and test set (with a test set data size of 100 for each task) and evaluate the performance of our method using the prediction MSE on the test set. |
| Hardware Specification | No | The paper does not specify the hardware used for its experiments, such as CPU or GPU models, or cloud computing resources. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., programming languages, libraries, frameworks) used for implementation or experiments. |
| Experiment Setup | Yes | We generate synthetic Gaussian data with nmin = 100, p = 100, nmin = Οnmin for Ο {0.5, 1, 2, 5, 10}, and Οq = 0 for each q. The number of tasks was fixed at 8. Furthermore, we generate a row-sparse B matrix with 10 nonzero rows and a B with rank 2 for the sparse and low-rank multi-task estimators, respectively. |