Leon Freese

Student Type: MEng

Link: https://www.linkedin.com/in/leon-freese-359784281/

Publications

2025

Freese, L. ., & Theunissen, M. W. (2025). Data Pruning: Redundant, Problematic, and Interdependent Samples. Artificial Intelligence Research. SACAIR 2025. Communications in Computer and Information Science, vol 2784. http://doi.org/https://doi.org/10.1007/978-3-032-11733-5_12

Abstract BibTex Entry BibTeX Download

The performance of deep learning models is affected by not only data quantity but also data quality. Data pruning is a process by which practitioners can reduce the size of a dataset by only keeping the most important training data points, thereby achieving similar test set performance. We empirically investigate two popular data pruning methods under noisy and noiseless conditions and show that these methods fail in the presence of significant label noise. We highlight that the success of data pruning is distinctly affected by three factors: redundancy in the dataset, the presence of problematic samples, and interdependence between samples. We perform a detailed investigation on commonly used benchmark classification datasets and neural network architectures. We find that our observations are consistent across data distributions and training protocols.

@article{529,
  author = {Leon Freese and Marthinus Theunissen},
  title = {Data Pruning: Redundant, Problematic, and Interdependent Samples},
  abstract = {The performance of deep learning models is affected by not only data quantity but also data quality. Data pruning is a process by which practitioners can reduce the size of a dataset by only keeping the most important training data points, thereby achieving similar test set performance. We empirically investigate two popular data pruning methods under noisy and noiseless conditions and show that these methods fail in the presence of significant label noise. We highlight that the success of data pruning is distinctly affected by three factors: redundancy in the dataset, the presence of problematic samples, and interdependence between samples. We perform a detailed investigation on commonly used benchmark classification datasets and neural network architectures. We find that our observations are consistent across data distributions and training protocols.},
  year = {2025},
  journal = {Artificial Intelligence Research. SACAIR 2025. Communications in Computer and Information Science},
  volume = {vol 2784},
  month = {25 November 2025},
  doi = {https://doi.org/10.1007/978-3-032-11733-5_12},
}