作者: Srinadh Bhojanapalli , Andreas Veit , Aditya Krishna Menon , Ankit Singh Rawat , Seungyeon Kim
DOI:
关键词: Contextual image classification 、 Machine learning 、 Computer science 、 Artificial intelligence 、 Artificial neural network 、 Reduction (complexity)
摘要: Standard training techniques for neural networks involve multiple sources of randomness, eg, initialization, mini-batch ordering and in some cases data augmentation. Given that …