Machine Learning and Automated Programming in a Computational Creativity context

Abstract
Software writing software is a major research direction for researchers in the field of Computational Creativity, where we study how to write systems that can take on some of the creative responsibility in arts and science projects. In the talk, I will motivate some of the many issues in the field, and how automating programming addresses some of these issues. Machine learning implementations clearly undertake automated programming – this is most obvious when evolutionary programming or inductive logic programming are employed, but is true in the general case. In the talk, I will present unsupervised machine learning directly as an automated programming task, and describe how I have modified the HR automated theory formation system (http://ccg.doc.gold.ac.uk/research/hr/) to perform automated programming. I will describe the application of this for Computational Creativity projects, such as automated poetry generation, puzzle design and mathematical discovery, and also present a Machine Learning context, touching on overlaps with deep learning.

Bio
Simon Colton is a Professor of Computational Creativity in the Department of Computing at Goldsmiths, University of London, and also an EPSRC leadership fellow. His research interests have covered many areas of Artificial Intelligence, including machine learning, constraint solving, automated theorem proving, evolutionary approaches and philosophical issues, all within a context of Computational Creativity, where the aim is to engineer autonomously creative software. He has published more than 150 papers, and his research has been recognised with national and international awards. He leads the Computational Creativity Research Group at Goldsmiths (http://ccg.doc.gold.ac.uk), which comprises 12 people investigating issues surrounding creative systems with applications to pure mathematics, graphic design, creative language, ideation, the visual arts and videogame design. Prof. Colton is best known for his work on creative systems such as The Painting Fool (www.thepaintingfool.com), for which the aim is to build an automated painter which is one day taken seriously as a creative artist in its own right.