Hands-On Machine Learning for Algorithmic Trading

Deep Convolutional GAN (DCGAN)

Deep Convolutional GAN (DCGAN) was motivated by the successful application of CNN to supervised learning for grid-like data. The architecture pioneered the use of GANs for unsupervised learning by developing a feature extractor based on adversarial training. It's also easier to train and generates higher-quality images. It is now considered a baseline implementation with numerous open source examples available (see references on GitHub: https://github.com/PacktPublishing/Hands-On-Machine-Learning-for-Algorithmic-Trading).

The DCGAN network takes uniformly-distributed random numbers as input and outputs a color image with a resolution of 64 x 64 pixels. As the input changes incrementally, so do the generated ...