search

cosmoharrigan / rl-glue

Automatically exported from code.google.com/p/rl-glue
0 stars 0 forks source link

unfreeze #38

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
We should have agent_unfreeze() and RL_unfreeze().
The idea that unfreeze lets the agent adapt again.
The rationale is that you may want to check the performance of the agent
multiple times during a longer training experiment.

Original issue reported on code.google.com by Csaba.Szepesvari on 31 Oct 2007 at 1:31

GoogleCodeExporter commented 9 years ago 
Sorry, I did not know how to change the type and priority.

Original comment by Csaba.Szepesvari on 31 Oct 2007 at 1:32

GoogleCodeExporter commented 9 years ago 
This has come up.  The offical word from the top is that RL_freeze() is 
intended to
be a freeform function that you should use as you see fit.  Perhaps the first 
call
could freeze, and the second could unfreeze().

As we realized that people had all sorts of fancy ideas that they wanted to push
through the RL_glue interface, we invented the new agent/env_message functions 
to
fill this need.  Unofficially (and it should become official), RL_freeze() is
deprecated in favor of user-managed messages.

Original comment by brian.ta...@gmail.com on 29 Jan 2008 at 6:25