Little problem on the size of input layer #12
Description
In this paper, the author proposed that for mnist dataset, the input layer size is 3636 . For CIFAR100, it is 9494. As we all know that each imageis 28281 in mnist and 32323 in CIFAR10. How can we understand this setting? Do you have some ideas?