Finite-Sample Maximum Likelihood Estimation of Location
Authors: Shivam Gupta, Jasper Lee, Eric Price, Paul Valiant
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we give experimental evidence supporting our proposed algorithmic theory. Our goals are to demonstrate that 1) r-smoothing is a beneficial pre-processing to the MLE, that r-smoothed Fisher information does capture the algorithmic performance in location estimation and 2) r-smoothed MLE can outperform the standard MLE, as well as standard mean estimation algorithms which do not leverage information about the distribution shape. |
| Researcher Affiliation | Academia | Shivam Gupta The University of Texas at Austin shivamgupta@utexas.edu Jasper C.H. Lee University of Wisconsin Madison jasper.lee@wisc.edu Eric Price The University of Texas at Austin ecprice@cs.utexas.edu Paul Valiant Purdue University pvaliant@gmail.com |
| Pseudocode | Yes | Algorithm 1 Local MLE for known parametric model; Algorithm 2 Global MLE for known parametric model |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository for its methodology. |
| Open Datasets | No | The paper uses a 'Gaussian-spiked Laplace model for experiments' which appears to be a synthetic model generated by the authors, not a publicly available dataset with concrete access information. |
| Dataset Splits | No | The paper does not specify training, validation, or test dataset splits. It mentions using 'n samples' but no partitioning details for these subsets. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments. |
| Software Dependencies | No | The paper does not list any specific software dependencies or their version numbers (e.g., programming languages, libraries, frameworks). |
| Experiment Setup | Yes | We use the Gaussian-spiked Laplace model for experiments, with a Laplace distribution of density proportional to e |x|, and a Gaussian of mass 0.001 and width roughly 0.002 (the discretization granularity) added at x = 4. The x-axis varies the number of samples n from 50 to 5000, and the y-axis varies the smoothing parameter r from 0.001 to 1 in log scale. |