An Investigation of Why Overparameterization Exacerbates Spurious Correlations
Authors: Shiori Sagawa, Aditi Raghunathan, Pang Wei Koh, Percy Liang
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through simulations and experiments on two image datasets, we identify two key properties of the training data that drive this behavior: the proportions of majority versus minority groups, and the signal-to-noise ratio of the spurious correlations. |
| Researcher Affiliation | Academia | 1Stanford University. Correspondence to: Shiori Sagawa <ssagawa@cs.stanford.edu>, Aditi Raghunathan <aditir@stanford.edu>, Pang Wei Koh <pangwei@cs.stanford.edu>. |
| Pseudocode | No | The paper does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | Code is available at https://github. com/ssagawa/overparam_spur_corr. |
| Open Datasets | Yes | In the first task, the label is spuriously correlated with demographics: specifically, we use the Celeb A dataset (Liu et al., 2015)... In the second task, the label is spuriously correlated with image background. We use the Waterbirds dataset (based on datasets from Wah et al. (2011); Zhou et al. (2017) and modified by Sagawa et al. (2020)). |
| Dataset Splits | No | The paper discusses 'training error' and 'test error' but does not specify the explicit split percentages (e.g., 80/10/10) or methodology used for creating training, validation, and test splits for reproducibility. |
| Hardware Specification | No | The paper does not specify any hardware details such as GPU models, CPU types, or memory used for running the experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers (e.g., PyTorch 1.x, Python 3.x). |
| Experiment Setup | Yes | For Celeb A, we train a Res Net10 model (He et al., 2016), varying model size by increasing the network width from 1 to 96... For Waterbirds, we use logistic regression over random projections... We train an unregularized logistic regression model over the feature representation Re LU(Wx) Rm... We vary model size by increasing the number of projections m from 1 to 10,000. We train each model by minimizing the reweighted objective (Equation (3)). |