{%hackmd SybccZ6XD %}
###### tags: `paper`

# CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features

Previous method
> regional dropout remove informative pixels on training images by overlaying a patch of either black pixels or random noise

Why previous is useless?
> Information loss

Algorithm
> 
> Sample box coordinates B = ($r_x, r_y, r_w, r_h$)
> Cropped ratio: $\frac{r_wr_h}{WH} = 1 - \lambda$
> 

Why this method is useful
> **Diverse training samples** reduce overfitting
>