Highest Rated Comments
debau3 karma
For example, we are seeing now in neuroevolution a new class of algorithms called "quality diveristy algorithms" (http://eplex.cs.ucf.edu/papers/pugh_gecco15.pdf) that focus on collecting a wide diversity of high quality solutions (something very natural for evolution), more like a repertoire than a single solution
Do you then use boosting to aggregate the outputs of the individual networks to improve the prediction accuracy? Doesn't drop-out learning do something very similar within a single network?
debau2 karma
With the rise of Deep Learning, for example Long Short Term Memory networks have been revived which have a very specific biologically-inspired topology. And they work great for time-series data. Do you know if NEAT creates or prefers certain topologies for certain tasks? If I feed it a lot of time-series data, will it come up with a LSTM-like topology?
debau72 karma
Where does the cell culture media come from?
View HistoryShare Link