mixup: Beyond Empirical Risk Minimization (Paper Explained)

Neural Networks often draw hard boundaries in high-dimensional space, which makes them very brittle. Mixup is a technique that linearly interpolate...