Neural Architecture Search with Reinforcement Learning

Re:Neural Architecture Search with Reinforcement Learning by gallerdude on Thursday January 19, 2017 @04:38PM (#13439708)

I think this is the way that Neural Networks achieve some modicum of generality - chaining them together.

Let's say you have a robot that you want to grab a can of beer off the counter. You say "grab that beer" and point to it. The first neural network interprets the speech and visual input. A second neural network chooses the proper neural nets to continue the task based on the information interpreted from the first net - it picks one for walking and one for grabbing.

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by randcraw on Thursday January 19, 2017 @05:00PM (#13439848)

Of course compound task federation like this is not exactly new. Blackboard systems did the very same thing using expert systems back in the 1970s-80s.

The problems then remain today: complex and compound tasks don't decompose neatly into well-defined constituent subtasks. Sub-task recombination rapidly devolves into a rat's nest of selecting from competing subtask components that don't have clearly distinct semantics, that are not free of contextual dependencies, nor do they plug-in-and-play independently.

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by ced on Thursday January 19, 2017 @06:02PM (#13440284)

Do you have any interesting material on this? Where's a good in-depth analysis of blackboard systems' pros and cons?

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by randcraw on Thursday January 19, 2017 @06:53PM (#13440558)

I'd start here: https://en.wikipedia.org/wiki/Blackboard_system

I'm not up-to-date, but I haven't heard of new work on blackboards or federated rule-based systems for maybe 20 years now, after expert systems grew increasingly probabilistic (and procrustean), and BBSes based on binary rules showed little sign of escaping the classic brittleness of RBSes.

The wiki article mentions 'Bayesian Blackboard' systems. Maybe they had greater success?

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by joshmarlow on Thursday January 19, 2017 @07:38PM (#13440819)

BBS systems also interest me quite a bit. Check out https://en.wikipedia.org/wiki/LIDA_(cognitive_architecture). It's based on a Global Workspace model (conceptually similar to a Black Board System). You can find papers on LIDA online (including one or two from the references on the wiki page). If you find anything particularly interesting, feel free to share!

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by westoncb on Thursday January 19, 2017 @05:29PM (#13440071)

So maybe now we need an NN that figures out the best way of wiring up various pre-existing NNs?

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by halflings on Thursday January 19, 2017 @07:19PM (#13440706)

Isn't that exactly what the paper is about?

[ Parent | on HN ]

Re:Neural Architecture Search with Reinforcement Learning by westoncb on Thursday January 19, 2017 @07:23PM (#13440729)

It's sort of related. In the paper they are still learning parameters for a single new network, as opposed to training a new network to be good at linking up other networks.

[ Parent | on HN ]