Posted: 3/4/2015 2:13 PM
Post subject: Overview of deep learning
Quote
gbagcn
Joined: 5/30/2005
Posts: 98
http://arxiv.org/abs/1404.7828 This is the only paper I have seen where citations make up over half its length (not even Wikipedia is this bad). Since these techniques have already been used to play atari games well maybe they will be used to make tas's one day. They would need to be an expert in both fields though and I am not sure if there is anyone like that.
Posted: 3/31/2015 11:23 PM
Quote
£e_Nécroyeur
Skilled player (1672)
Joined: 7/1/2013
Posts: 448
Arcade TASer of 2017 Arcade TAS of 2016 Exotic platforms TASer of 2014 Exotic platforms TAS of 2014
Interesting stuff, to be certain! However, the methods used in the Atari 2600 experiments are far more complex than would be necessary for making a TAS. The computers used in the experiment played the games directly from video input, with no access to the emulator state. A computer with access to the emulator state (e.g. LuaBot) can discover solutions much more quickly and with less processing power.