C ++ and CUDA C Deep Learning Algorithm Vol.2

by Timothy Masters

C ++ and CUDA C Deep Learning Algorithm Vol.2This is the second volume in a three-volume "Deep Learning Neural Network Implementation with C ++ and CUDA C" series. The autocoder, which is based on the encoder and decoder structure and restores the input information, is becoming more and more important due to the development of the GAN model. First of all, I would like to have a good understanding of the contents of the first volume because I will mention the contents covered in the first volume.I first examine the concept of the Mollet wavelet at the basic signal processing level and the concept of the Fourier transform when it is extended to image processing. see. We then explain the concept of auto-encoding in a complex domain. We extend these contents to the neural network in the complex number domain and calculate the activation function, the gradient, and the SoftMax layer. We start with a single-thread-based implementation, and finally with a multi-thread-based implementation. In Chapter 3, the contents implemented in each layer in Chapter 2 are implemented in CUDA C to utilize GPGPU. Finally, chapter 4 introduces the manual of the DEEP program provided by the author, and helps the user confirm the execution result of each function in advance and confirm easily.

Paperback

Published August 11, 2017