Precise High-Dimensional Asymptotics for Quantifying Heterogeneous Transfers
Authors: Fan Yang, Hongyang R. Zhang, Sen Wu, Christopher Re, Weijie J. Su
JMLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Simulations validate the accuracy of the high-dimensional asymptotics for finite dimensions. Simulations are provided to show the accuracy of our estimates in finite dimensions. |
| Researcher Affiliation | Academia | Fan Yang EMAIL Yau Mathematical Sciences Center, Tsinghua University Beijing 100084, China Hongyang R. Zhang EMAIL Khoury College of Computer Sciences, Northeastern University Boston, MA 02115, US Sen Wu EMAIL Department of Computer Science, Stanford University Stanford, CA 94305, US Christopher Ré EMAIL Department of Computer Science, Stanford University Stanford, CA 94305, US Weijie J. Su EMAIL Department of Statistics and Data Science, University of Pennsylvania Philadelphia, PA 19104, US |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. It describes methodologies and proofs using mathematical notation and natural language. |
| Open Source Code | Yes | The experiment code for reproducing these simulation results can be found at https://github.com/Virtuoso-Research/Transfer_learning_random_matrix_simulations. |
| Open Datasets | No | The paper describes generating synthetic data for its simulations (e.g., 'We sample the covariates X from a p-dimensional isotropic Gaussian', 'generate covariate-shifted features and different linear models'). It does not use or provide access information for any established public datasets. |
| Dataset Splits | No | The paper uses simulated data and defines sample sizes for source and target tasks (e.g., 'n1 samples', 'n2 samples'). However, it does not describe traditional train/test/validation splits for empirical data evaluation. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments. |
| Software Dependencies | No | The paper does not specify any software libraries, tools, or programming languages with their version numbers that are needed to replicate the experiments. |
| Experiment Setup | Yes | In Figure 1, the text states: 'In this simulation, we set p = 100, n2 = 300, and σ2 = 1/4.' Figure 2b mentions: 'This simulation fixes p = 100, n2 = 300 and varies n1, λ. Both simulations use σ = 1/2.' Figure 3a states: 'For this simulation, we set p = 50, n1 = n2 = 100, and σ = 1/2.' Figure 4a mentions: 'We fix model shift µ = 0.1 while varying covariate shift λ for each curve.' Figure 5b states: 'Figure 5b fixes r = 1 and varies µ, n. The results under different values of µ also match the conditions in Proposition 14.' |