[kthejoker2]

Cool article, lots to digest, one thing caught my eye:

"To the best of our knowledge, the only prior work using unsupervised learning as a pre-processor for neuroevolution is (cite)."

Just amazing how much low-hanging fruit there still is in the space.

[giuse]

Author here. The idea is low-hanging indeed, several friends (including @togelius!) commented "I always wanted to do that -- eventually". Realization is another matter. Have a look at the mess necessary to make it work: we had to discard UL initialization for online learning, accept that the encoding would grow in size, adapt the network sensibly to these changes, and tweak the ES to account for the extra weights.