Distribution Density, Tails, and Outliers in Machine Learning: Metrics and Applications
This work addresses the need for metrics to assess data quality and outliers in machine learning, offering tools for dataset analysis and model improvement, though it is incremental as it builds on existing outlier detection concepts.
The paper tackles the problem of quantifying how well-represented examples are in datasets, evaluating five methods on MNIST, Fashion-MNIST, CIFAR-10, and ImageNet, finding high correlation among them and applications such as identifying prototypical examples, memorized training data, and improving curriculum learning and adversarial robustness.
We develop techniques to quantify the degree to which a given (training or testing) example is an outlier in the underlying distribution. We evaluate five methods to score examples in a dataset by how well-represented the examples are, for different plausible definitions of "well-represented", and apply these to four common datasets: MNIST, Fashion-MNIST, CIFAR-10, and ImageNet. Despite being independent approaches, we find all five are highly correlated, suggesting that the notion of being well-represented can be quantified. Among other uses, we find these methods can be combined to identify (a) prototypical examples (that match human expectations); (b) memorized training examples; and, (c) uncommon submodes of the dataset. Further, we show how we can utilize our metrics to determine an improved ordering for curriculum learning, and impact adversarial robustness. We release all metric values on training and test sets we studied.