added mdp maze task. |
|
More...
|
over 15 years ago
|
argh |
|
More...
|
over 15 years ago
|
changed pgpe paras |
|
More...
|
over 15 years ago
|
Parameter write access limitation for fast networks (cf. issue #23). |
|
More...
|
over 15 years ago
|
Explanatory Exception added for improper use of BackPropTrainer. |
|
More...
|
over 15 years ago
|
Some changes to the documentation. |
|
More...
|
over 15 years ago
|
Added testing for minimize/maximize consistency, and fixed a few related bugs in the same time. |
|
More...
|
over 15 years ago
|
Documented most of the utility functions. |
|
More...
|
over 15 years ago
|
Added script for thorough testing of optimization algorithms. Some consistency fixes. |
|
More...
|
over 15 years ago
|
Streamlined parameter passing for PGPE. Fixed the cart_pgpe example. |
|
More...
|
over 15 years ago
|
Moved all files from the rl/tasks folder to their respective subfolders in rl/environments. |
|
More...
|
over 15 years ago
|
Git now ignores the temporary files. |
|
More...
|
over 15 years ago
|
Fixing some sphinx documentation issues. |
|
More...
|
over 15 years ago
|
Neuro-fitted Q-learning example added. |
|
More...
|
over 15 years ago
|
Main Learner class split into its functional components (left-over from restructuring). |
|
More...
|
over 15 years ago
|
Consistently renamed SPLA into PGPE |
|
More...
|
over 15 years ago
|
Renamed all python files to lower-case filenames. |
|
More...
|
over 15 years ago
|
Import path fixes (due to restructuring). |
|
More...
|
over 15 years ago
|
Merged and fixed reinforcement learning and blackbox optimization restructuring stuff by Tom Schaul and Thomas Rueckstiess. |
|
More...
|
over 15 years ago
|
Removed arac installation completely from setup.py |
|
More...
|
over 15 years ago
|
removed unnecessary backward pass from ActionValueNetwork |
|
More...
|
over 15 years ago
|
moved call of newEpisode() for explorers up to the learner base class. |
|
More...
|
over 15 years ago
|
added discrete SDE exploration. |
|
More...
|
over 15 years ago
|
added another task for pole balancing. |
|
More...
|
over 15 years ago
|
Minimize/maximize consistency fixes. All black-box optimizers not maximize by default. Some other minor bugfixes. |
|
More...
|
over 15 years ago
|
bug fix for learning agent when using no learner. |
|
More...
|
over 15 years ago
|
Tables inherit from ParameterContainer now, storing their values as self.params. Things like table.mutate() is now possible. |
|
More...
|
over 15 years ago
|
added one-of-n coding tools. |
|
More...
|
over 15 years ago
|
return sensors always as array (not list). |
|
More...
|
over 15 years ago
|
added DiscreteBalanceTask. |
|
More...
|
over 15 years ago
|