I love this work by Otoro / David Ha.<p>For come context, this uses a technique called CPPN’s (compositional pattern producing networks), where the inputs are (x, y, sqrt(x^2 + y^2)) and then there are several neural net layers with different activation functions and randomly initialized weights. It actually resembles the optical physics of light reflecting through glass (which explains why it sometimes looks like light reflecting through a gem).<p>Otoro wrote some great blog posts on this:<p><a href="https://blog.otoro.net/2016/03/25/generating-abstract-patterns-with-tensorflow/" rel="nofollow">https://blog.otoro.net/2016/03/25/generating-abstract-patter...</a><p><a href="https://blog.otoro.net/2015/07/31/neurogram/" rel="nofollow">https://blog.otoro.net/2015/07/31/neurogram/</a><p><a href="https://github.com/hardmaru/cppn-tensorflow" rel="nofollow">https://github.com/hardmaru/cppn-tensorflow</a>
At the risk of sounding like an idiot. What even is this? Is there a purpose? It looks just like random pixel generation to me. There is no explanation anywhere I can find on a "purpose" or "motivation".