AI for Art: Recipes for Art Generation with Machine Learning

1. Auflage 2020
ISBN: 978-1-4842-4997-0
Verlag: APRESS

Buch, Englisch, 180 Seiten, Format (B × H): 155 mm x 235 mm

The ability of a machine to learn how to create art makes for an intriguing partnership and perhaps a clear test of artificial intelligence. This book shows you how to create artificial intelligence capable of generating art—such as music and pictures—that can perform style transfers.

You'll review the math behind generative ML models from classical Gaussian Mixture to the modern day GANs. Generative models are not something new in the ML world, but the recent creation of new types of Neural Networks such as GAN allows AI to produce something that can be truly called art.

Rather than an academic and detached approach, this book offers concrete recipes with math explanations so that readers can learn by directly interacting with these models. Detailed code examples expand on each concept to create ML models on popular frameworks such as PyTorch and TensorFlow+Keras. The focus is on the practical aspects of music and picture generation with artificial intelligence—providing useful tips and best practices obtained from experience. Beyond GAN, different examples of how art can be generated with other types of models, even with simple statistical models, are presented.

With AI for Art you'll learn about generative Machine Learning models on several levels and the math behind each model.

What You'll Learn
- Understand the difference between generative and discriminative models, including Mixture models, Hidden Markov models, Bayesian networks, and more.

- Work with both PyTorch and TensorFlow+Keras

- Create music, pictures, and text with AI

Who This Book Is For
Programmers with a knowledge of Python 3 and Deep Learning libraries, such as Keras and PyTorc. Libraries will be covered briefly for newer programmers less familiar with the topic.

Autoren/Hrsg.

Osipenko, Alexander

Fachgebiete

• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz

Inhaltsverzeichnis

1. Introduction and history of generative models
Provide an historical overview and set some basics definitions

2. Types of generative models

Provide explanation for each type of model

· Differences between generative and discriminative models

· Mixture models

· Hidden Markov model

· Bayesian network

· Latent Dirichlet allocation

· Boltzmann machine

· Autoencoders

· Generative adversarial network

3. Keras and PyTorch crash cource

Provide an introduction for two popular deep learning libraries

4. Music generation

Showcase different examples of how music can be generated with ML

· Statistical models

· Recurrent Neural Networks and LSTM

· Recurrent GAN

5. Picture generation and style transfer

Showcase different examples of how pictures can be generated with ML

· Convolutional Autoencoders

· Convolutional GAN

6. Text generation

Showcase different examples of how text can be generated with ML

· Markov models vs Kanye West

· Seq-2-Seq and Chat Bots

Über Autor(innen)

Alexander Osipenko has an experienced background in computational chemistry, data analysis, machine learning, and data science. He has also worked in statistical analysis for science projects. Currently he serves at the Machine Learning Expert managing the machine learning branch at Eagle6, focusing on real-time data analysis and anomaly detection in industrial IoT networks.

To solve anomaly detection problems, Andrew uses different types of generative models from Gaussian mixture, Bayesian networks to Deep Autoencoders. He also has a passion for music, and can play several instruments. So in his free time, he uses generative models to create music and pictures and to perform style transfers.