When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations

Rhys Compton, Lily Zhang, Aahlad Puli, Rajesh Ranganath
Proceedings of the 8th Machine Learning for Healthcare Conference, PMLR 219:110-127, 2023.

Abstract

In machine learning, incorporating more data is often seen as a reliable strategy for improving model performance; this work challenges that notion by demonstrating that the addition of external datasets in many cases can hurt the resulting model’s performance. In a large-scale empirical study across combinations of four different open-source chest x-ray datasets and 9 different labels, we demonstrate that in 43% of settings, a model trained on data from two hospitals has poorer worst group accuracy over both hospitals than a model trained on just a single hospital’s data. This surprising result occurs even though the added hospital makes the training distribution more similar to the test distribution. We explain that this phenomenon arises from the spurious correlation that emerges between the disease and hospital, due to hospital-specific image artifacts. We highlight the trade-off one encounters when training on multiple datasets, between the obvious benefit of additional data and insidious cost of the introduced spurious correlation. In some cases, balancing the dataset can remove the spurious correlation and improve performance, but it is not always an effective strategy. We contextualize our results within the literature on spurious correlations to help explain these outcomes. Our experiments underscore the importance of exercising caution when selecting training data for machine learning models, especially in settings where there is a risk of spurious correlations such as with medical imaging. The risks outlined highlight the need for careful data selection and model evaluation in future research and practice.

Cite this Paper

BibTeX
@InProceedings{pmlr-v219-compton23a, title = {When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations}, author = {Compton, Rhys and Zhang, Lily and Puli, Aahlad and Ranganath, Rajesh}, booktitle = {Proceedings of the 8th Machine Learning for Healthcare Conference}, pages = {110--127}, year = {2023}, editor = {Deshpande, Kaivalya and Fiterau, Madalina and Joshi, Shalmali and Lipton, Zachary and Ranganath, Rajesh and Urteaga, Iñigo and Yeung, Serene}, volume = {219}, series = {Proceedings of Machine Learning Research}, month = {11--12 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v219/compton23a/compton23a.pdf}, url = {https://proceedings.mlr.press/v219/compton23a.html}, abstract = {In machine learning, incorporating more data is often seen as a reliable strategy for improving model performance; this work challenges that notion by demonstrating that the addition of external datasets in many cases can hurt the resulting model’s performance. In a large-scale empirical study across combinations of four different open-source chest x-ray datasets and 9 different labels, we demonstrate that in 43% of settings, a model trained on data from two hospitals has poorer worst group accuracy over both hospitals than a model trained on just a single hospital’s data. This surprising result occurs even though the added hospital makes the training distribution more similar to the test distribution. We explain that this phenomenon arises from the spurious correlation that emerges between the disease and hospital, due to hospital-specific image artifacts. We highlight the trade-off one encounters when training on multiple datasets, between the obvious benefit of additional data and insidious cost of the introduced spurious correlation. In some cases, balancing the dataset can remove the spurious correlation and improve performance, but it is not always an effective strategy. We contextualize our results within the literature on spurious correlations to help explain these outcomes. Our experiments underscore the importance of exercising caution when selecting training data for machine learning models, especially in settings where there is a risk of spurious correlations such as with medical imaging. The risks outlined highlight the need for careful data selection and model evaluation in future research and practice.} }
Endnote
%0 Conference Paper %T When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations %A Rhys Compton %A Lily Zhang %A Aahlad Puli %A Rajesh Ranganath %B Proceedings of the 8th Machine Learning for Healthcare Conference %C Proceedings of Machine Learning Research %D 2023 %E Kaivalya Deshpande %E Madalina Fiterau %E Shalmali Joshi %E Zachary Lipton %E Rajesh Ranganath %E Iñigo Urteaga %E Serene Yeung %F pmlr-v219-compton23a %I PMLR %P 110--127 %U https://proceedings.mlr.press/v219/compton23a.html %V 219 %X In machine learning, incorporating more data is often seen as a reliable strategy for improving model performance; this work challenges that notion by demonstrating that the addition of external datasets in many cases can hurt the resulting model’s performance. In a large-scale empirical study across combinations of four different open-source chest x-ray datasets and 9 different labels, we demonstrate that in 43% of settings, a model trained on data from two hospitals has poorer worst group accuracy over both hospitals than a model trained on just a single hospital’s data. This surprising result occurs even though the added hospital makes the training distribution more similar to the test distribution. We explain that this phenomenon arises from the spurious correlation that emerges between the disease and hospital, due to hospital-specific image artifacts. We highlight the trade-off one encounters when training on multiple datasets, between the obvious benefit of additional data and insidious cost of the introduced spurious correlation. In some cases, balancing the dataset can remove the spurious correlation and improve performance, but it is not always an effective strategy. We contextualize our results within the literature on spurious correlations to help explain these outcomes. Our experiments underscore the importance of exercising caution when selecting training data for machine learning models, especially in settings where there is a risk of spurious correlations such as with medical imaging. The risks outlined highlight the need for careful data selection and model evaluation in future research and practice.
APA
Compton, R., Zhang, L., Puli, A. & Ranganath, R.. (2023). When More is Less: Incorporating Additional Datasets Can Hurt Performance By Introducing Spurious Correlations. Proceedings of the 8th Machine Learning for Healthcare Conference, in Proceedings of Machine Learning Research 219:110-127 Available from https://proceedings.mlr.press/v219/compton23a.html.

Related Material