Evolutionary Algorithms (EA) are a sub-field of Artificial Intelligence that solve optimisation problems by mimicking biological evolution through mutation, reproduction, and selection of the fittest solutions. EA are able to optimise non-linear and non-parametric problems and can converge to global optima through smart experiment design.
Connectionist models are generally trained using back-propagation. Neuro-Evolution (NE) combines the power of EA and Artificial Neural Networks by evolving network topologies and connection weights. The power of NE grows yearly as computation power increases with advances in distributed-, and GPU computing.
The processing power needed to run large evolutionary experiments can be expensive and time consuming. The parallelism properties of EA make it possible to leverage distributed computing and to that end, Zipfian explores ways in creating crowd computing platforms to aide in large scale experimentation.
A collective of scientists and engineers researching the potential of using evolutionary and machine learning algorithms to find novel problem solutions. We aim to conduct and publish research, aide in the development of open source tools for experimentation, and create awareness of evolutionary algorithms within the machine learning community. Zipfian Science is a non-profit group and brings together talented young minds with the freshest ideas to have fun while solving challenging problems. From a myriad of different backgrounds, reading the latest research and writing Python code daily, we have no end to creative ideas.