From the book "Deep Learning and Convolutional Neural Networks for Medical Image Computing"
As we learned about the current state of research on deep learning, I was surprised to find that other investigators had used convolutional neural networks, one type of deep learning, in the past [22, 23]. But there seemed to be something different about the most recent crop of deep learning algorithms. They routinely used GPU processing to
accelerate training by as much as a factor of 40-fold. They also used multiple convolution layers and multiple data reduction layers.
40-fold mean "K-Fold Cross-Validation"?