Preventing Manifold Intrusion with Locality: Local Mixup

12 Jan 2022  ·  Raphael Baena, Lucas Drumetz, Vincent Gripon ·

Mixup is a data-dependent regularization technique that consists in linearly interpolating input samples and associated outputs. It has been shown to improve accuracy when used to train on standard machine learning datasets. However, authors have pointed out that Mixup can produce out-of-distribution virtual samples and even contradictions in the augmented training set, potentially resulting in adversarial effects. In this paper, we introduce Local Mixup in which distant input samples are weighted down when computing the loss. In constrained settings we demonstrate that Local Mixup can create a trade-off between bias and variance, with the extreme cases reducing to vanilla training and classical Mixup. Using standardized computer vision benchmarks , we also show that Local Mixup can improve test accuracy.

PDF Abstract


Introduced in the Paper:

Two Coiling Spirals

Used in the Paper:

Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Classification CIFAR-10 Local Mixup Resnet18 Percentage correct 95.97 # 115
Image Classification Fashion-MNIST Local Mixup DenseNet Percentage error 5.97 # 9
Image Classification SVHN Local Mixup LeNet Percentage error 8.20 # 45