Querying across time to interactively evolve animations

Compositional Pattern Producing Networks (CPPNs) are a generative encoding that has been used to evolve a variety of novel artifacts, such as 2D images, 3D shapes, audio timbres, soft robots, and neural networks. This paper takes systems that generate static 2D images and 3D shapes with CPPNs and introduces a time input, allowing each CPPN to produce a different set of results for each slice of time. Displaying the results in sequence creates smooth animations that can be interactively evolved to suit users' personal aesthetic preferences. A human subject study involving 40 individuals was conducted to demonstrate that people find the dynamic animations more complex than static outputs, and find interactive evolution of animations more enjoyable than evolution of static outputs. The novel idea of indirectly generating artifacts as a function of time could also be useful in other domains.

[1]  Hod Lipson,et al.  Unshackling evolution: evolving soft robots with multiple materials and a powerful generative encoding , 2013, GECCO '13.

[2]  Sebastian Risi,et al.  Creative Generation of 3D Objects with Deep Learning and Innovation Engines , 2016, ICCC.

[3]  Sebastian Risi,et al.  Interactively Evolving Compositional Sound Synthesis Networks , 2015, GECCO.

[4]  Kenneth O. Stanley,et al.  Compositional Pattern Producing Networks : A Novel Abstraction of Development , 2007 .

[5]  Kenneth O. Stanley,et al.  A Hypercube-Based Encoding for Evolving Large-Scale Neural Networks , 2009, Artificial Life.

[6]  Sebastian Risi,et al.  A unified approach to evolving plasticity and neural geometry , 2012, The 2012 International Joint Conference on Neural Networks (IJCNN).

[7]  Risto Miikkulainen,et al.  Automatic feature selection in neuroevolution , 2005, GECCO '05.

[8]  Kenneth O. Stanley,et al.  Picbreeder: A Case Study in Collaborative Evolutionary Exploration of Design Space , 2011, Evolutionary Computation.

[9]  Kenneth O. Stanley,et al.  The Emergence of Canalization and Evolvability in an Open-Ended, Interactive Evolutionary System , 2017, Artificial Life.

[10]  Christina Backes,et al.  EDISON-WMW: Exact Dynamic Programing Solution of the Wilcoxon–Mann–Whitney Test , 2016, Genom. Proteom. Bioinform..

[11]  William E. Lorensen,et al.  Marching cubes: A high resolution 3D surface construction algorithm , 1987, SIGGRAPH.

[12]  Risto Miikkulainen,et al.  Evolving Neural Networks through Augmenting Topologies , 2002, Evolutionary Computation.

[13]  Hod Lipson,et al.  Evolving three-dimensional objects with a generative encoding inspired by developmental biology , 2011, ECAL.