As the implementation complexity of deep learning increases, ‘Pythonic’ and easy-to-use PyTorch are attracting attention. The author, an executive of PyTorch Korea, wrote this book for the purpose of lowering the barriers to entry into deep learning with his experience of studying and lecturing deep learning for many years. From the installation of PyTorch to CNN, RNN, and even the latest research results such as style transfer, auto encoder and GAN. With the mind of a developer who has suffered the same difficulty, the concept principle and implementation are well balanced so readers can easily understand.

Features

Install anaconda + CUDA + cuDNN (just use Colab to feel comfortable)

• Understand the loss function and gradient descent while looking at the linear regression analysis.

• The basics of artificial neural networks such as chain rule, propagation, and back propagation

• Learn CNN with friendly pictures and explore VGGNet, GoogLeNet, ResNet

• From the principle of RNN to LSTM, GRU, embedding, word2vec

• How to improve learning performance by solving overfitting and underfitting, dropout, formalization, initialization, and normalization.

• Style transfer, transition learning, L-BFGS

• Auto encoder and Semantic Segmentation

• GAN and Friends (DCGAN, SRGAN, Pix2Pix, CycleGAN, DiscoGAN)