Deep learning has enough randomness. The random weight initialization, the random choice of samples (as in name Stochastic Gradient Descent), the random flip/crop/noise in data augmentation, and randomness injected by dropout.