Optimal Learning from Verified Training Data
Authors: Nicholas Bishop, Long Tran-Thanh, Enrico Gerding
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We also provide empirical results on two real-world datasets, the medical personal costs dataset and the red wine dataset, showcasing the performance of our algorithm relative to algorithms which are optimal under adversarial assumptions, outperforming the state of the art. |
| Researcher Affiliation | Academia | Nicholas Bishop University of Southampton, UK nb8g13@soton.ac.uk Enrico Gerding University of Southampton, UK eg@ecs.soton.ac.uk Long Tran-Thanh University of Warwick, UK long.tran-thanh@warwick.ac.uk |
| Pseudocode | Yes | Algorithm 1 Bisection search for q Input: data matrix X, learner s labels y, data provider s labels z, tolerance ϵ Initialize q1 = 0 Initialize q2 = y T y repeat q = (q1 + q2)/2 if F(q) 0 then q1 = q else q2 = q end if until q2 q1 ϵ return q2 |
| Open Source Code | No | The paper does not provide a specific link or explicit statement about releasing the source code for the described methodology. |
| Open Datasets | Yes | We also provide empirical results on two real-world datasets, the medical personal costs dataset [9] and the red wine dataset [11]. [9] Miri Choi. Medical Cost Personal Datasets | Kaggle, 2018. URL https://www.kaggle.com/mirichoi0218/insurance/metadata. [11] Paulo Cortez, António Cerdeira, Fernando Almeida, Telmo Matos, and José Reis. Modeling wine preferences by data mining from physicochemical properties. Decision Support Systems, 47(4):547 553, 2009. |
| Dataset Splits | Yes | In order to evaluate Algorithm 1, we perform 10-fold cross validation and compare its performance to ridge regression for γ [1 10 5, 1]. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory specifications). |
| Software Dependencies | Yes | For ridge regression, and the SDPs in Algorithm 1, we use the SDPT3 solver [23] to find global solutions. [23] K. C. Toh, M. J. Todd, and R. H. Tütüncü. Sdpt3 a matlab software package for semidefinite programming, version 1.3. Optimization Methods and Software, 11(1-4):545 581, 1999. |
| Experiment Setup | Yes | For each value of γ, we compute the regularisation parameter for ridge regression via grid search on 8 logarithmically spaced points in the interval [1 10 5, 1000] during cross validation. The same error tolerances are used for both Algorithm 1 and the interior point method we use to solve noncovex problem reformulation proposed by Brückner and Scheffer [4]. |