Fig. 1
From: Multi-sample \(\zeta \)-mixup: richer, more realistic synthetic samples from a p-series interpolant
Overview of mixup  (b) and \(\zeta \)-mixup  (a, c, d). The original and synthesized samples are denoted by \(\circ \) and \(\bigtriangleup \) respectively, and line segments indicate which original samples were used to create the new ones. The line thicknesses denote the relative weights assigned to original samples. Observe how \(\zeta \)-mixup  can mix any number of samples (e.g., 3 in (a), 4 or 8 in (c), and 4 in (d)), and that \(\zeta \)-mixup ’s formulation allows the generated samples to be close to the original distribution while still incorporating rich information from several samples. d Illustrates a toy dataset with 3 classes, wherein a mini-batch of 4 elements is sampled, then the data and the labels are mixed using a set of weights generated with an example value of the hyperparameter \(\gamma \), and finally this synthesized data is used to train a classification model