Generalization and Robustness of the Tilted Empirical Risk
Authors: Gholamali Aminian, Amir R. Asadi, Tian Li, Ahmad Beirami, Gesine Reinert, Samuel N. Cohen
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically corroborate our findings in simple experimental setups where we evaluate our bounds to select the value of tilt in a data-driven manner. |
| Researcher Affiliation | Collaboration | 1The Alan Turing Institute, London, UK 2Statistical Laboratory, University of Cambridge, Cambridge, UK 3 Computer Science Department, University of Chicago, USA 4Google Deep Mind, USA 5Department of Statistics, University of Oxford, Oxford, UK 6Mathematical Institute, University of Oxford, Oxford, UK. |
| Pseudocode | No | The paper describes methods using mathematical formulations and textual explanations but does not include any explicitly labeled pseudocode blocks or algorithms. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code, nor does it include links to repositories for the methodology described. |
| Open Datasets | No | In the experiment details, the paper states, "For logistic regression, we consider 500 samples from 2d Gaussian distributions... We add outlier samples ... from Gaussian distribution or Pareto distribution." and "For true data, we consider X ~ N(1, 0.5) and y = x + 0.5 + N(0, 0.5)." This indicates that the datasets were generated rather than being specific public datasets with access information provided. |
| Dataset Splits | No | The paper mentions "The training dataset consists of 1,000 samples" and discusses adding outliers to this training dataset. However, it does not specify explicit training, validation, or test splits (e.g., percentages or counts for each split) for reproducing experiments. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as CPU, GPU models, or computing environment specifications used for running the experiments. |
| Software Dependencies | No | The paper describes experimental parameters like "learning rate and 10000 iterations" and specifies loss functions. However, it does not list any specific software (e.g., libraries, frameworks, or programming languages) along with their version numbers. |
| Experiment Setup | Yes | In Appendix G, 'Experiment Details', the paper states: 'For logistic regression, we consider 0.01 as learning rate and 10000 iterations.' It also details the loss function 'ℓ(h, z) = log(1 + exp( yh T x))', the number of samples (n = 1,000), and the parameters for generating Gaussian and Pareto outlier data for both logistic and linear regression scenarios. |