Change get/set state/random seed to load/save

[GoogleCodeExporter]

Make them function the way that they did in the competition.  Document the
change.  Implement this function in the skeleton agent/environment.

Original issue reported on code.google.com by brian.ta...@gmail.com on 20 Oct 2008 at 9:41

[GoogleCodeExporter]

Original comment by brian.ta...@gmail.com on 29 Oct 2008 at 4:30

• Changed state: Started

[GoogleCodeExporter]

Done with r922.

Original comment by brian.ta...@gmail.com on 29 Oct 2008 at 5:23

• Changed state: Fixed

[GoogleCodeExporter]

I've re-opened this.

I'm afraid it's not going to work.  The vision for save/load was that it would 
allow
us to easily build models, but that's not true.  Save/Load are called by the
experiment, not the agent.  However, the experiment can't decide what actions 
to take.

So, without a complex messaging system or a custom agent, you can't learn one of
these models using this approach.

I got to the point of trying to write an example and I just couldn't see how it 
would
look. 

This has me very worried.  I'd like to hash this out today.

Original comment by brian.ta...@gmail.com on 3 Nov 2008 at 2:45

• Changed state: GatheringData
• Added labels: Priority-Critical
• Removed labels: Priority-Medium

[GoogleCodeExporter]

Original comment by brian.ta...@gmail.com on 3 Nov 2008 at 2:46

[GoogleCodeExporter]

We have decided to remove this functionality entirely from RL-Glue.

Original comment by brian.ta...@gmail.com on 4 Nov 2008 at 5:00

• Changed state: WontFix
• Added labels: Type-Task
• Removed labels: Type-Defect