Selective Brain Damage: Measuring the Disparate Impact of Model Pruning

38 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Are we done with ImageNet? (2020, background): “Several studies have revisited common computer vision benchmarks [16, 17, 18, 19] and, regarding ImageNet [1] speciﬁcally, identiﬁed various sources of bias and noise [20, 21, 22].”
Datamodels: Predicting Predictions from Training Data (2022, background): “similar effect has been noted in the context of neural network pruning (Hooker et al., 2019).”
Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks (2021, methodology): “in the ImageNet dataset, 29 are well-known to contain errors [16, 31, 38], labeled data in test sets is often considered “correct” as 30 long as it is drawn from the same distribution as the train set — this is a fallacy — machine learning 31 test sets can, and do, contain pervasive errors and these”
Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth (2020, background): “Further study of accuracy and error sets have been conducted in (Hacohen & Weinshall, 2020) (error sets over training), and (Hooker et al., 2019) (error after pruning).”