Artificial intelligence has seen huge advances in recent years, with notable achievements like computers being able to compete with humans at the notoriously difficult to master ancient game of go, self-driving cars, and voice recognition in your pocket.
Even more recently, a new architecture emerged that led to spectacular results for generated images. The faces at this link are not real, they were created by a generative network - https://tinyurl.com/y6c6u5k2 .
In October 2018, the world-leading art auction house Christies sold the Portrait of Edmond Belamy for $432,500 - https://tinyurl.com/y5zjqjn3.
That portrait was not painted by a person, but a generative neural network trained using an adversarial machine learning architecture.
This talk - aimed at non-experts and newcomers - will introduce and explain how generative adversarial networks, or GANs, work.
After a quick review of how neural networks learn, we'll take gentle steps towards understanding how GANs work and building our own:
First we'll see the mechanics of learning using an almost trivially simple system of two learning nodes.
Introduce PyTorch basics - including the concept of computation graphs and automatic gradients.
We'll step up to using very small neural networks to learn to "fake" a short pattern.
After seeing the key concepts in action, we'll progress onto training a home-made GAN to learn to create convincing images.
Tutorial code will be provided as python notebooks so you can explore GANs yourself.
We'll also be learning just enough PyTorch basics to enable us to continue using it for other projects after the talk. No previous experience with PyTorch necessary.
The aim of the talk is to understand - at minimum - the key concepts behind neural networks and adversarial learning. We'll avoid unnecessary jargon and maths, and keep the content as simple as possible.
Artists and designers most welcome!
YOU MAY ALSO LIKE:
- Recursion For Beginners (SkillsCast recorded in April 2018)
- Martine Devos' Certified Scrum Product Owner (in London on 29th - 30th October 2019)
- Jenny Martin's BDD From Start to Finish - Successful Delivery through Continuous Collaboration (in London on 4th - 6th November 2019)
- P3X - People, Product & Process eXchange 2019 (in London on 31st October - 1st November 2019)
- Practical ML 2020 (in London on 2nd - 3rd July 2020)
- Knowledge Graph Applications in the Legal Profession & Fashion Industry (in London on 22nd October 2019)
- Business Agility and Continuous Compliance (in London on 22nd October 2019)
- Advances in Quantum Machine Learning (SkillsCast recorded in October 2019)
- Hiring a Personal Investigator for Your App (SkillsCast recorded in October 2019)