GEGELATI
|
Interface for creating a Learning Environment in adversarial mode. More...
#include <adversarialLearningEnvironment.h>
Public Member Functions | |
virtual std::shared_ptr< Learn::AdversarialEvaluationResult > | getScores () const =0 |
Computes scores of each root and returns them. More... | |
double | getScore () const override |
Simply returns the first score of the results, allowing compatibility with non adversarial learning agents. More... | |
![]() | |
LearningEnvironment ()=delete | |
Delete the default constructor of a LearningEnvironment. | |
virtual | ~LearningEnvironment ()=default |
Default virtual destructor. | |
LearningEnvironment (uint64_t nbAct) | |
Constructor for LearningEnviroment. More... | |
virtual LearningEnvironment * | clone () const |
Get a copy of the LearningEnvironment. More... | |
virtual bool | isCopyable () const |
Can the LearningEnvironment be copy constructed to evaluate several LearningAgent in parallel. More... | |
uint64_t | getNbActions () const |
Get the number of actions available for this LearningEnvironment. More... | |
virtual void | doAction (uint64_t actionID) |
Execute an action on the LearningEnvironment. More... | |
virtual void | reset (size_t seed=0, LearningMode mode=LearningMode::TRAINING)=0 |
Reset the LearningEnvironment. More... | |
virtual std::vector< std::reference_wrapper< const Data::DataHandler > > | getDataSources ()=0 |
Get the data sources for this LearningEnvironment. More... | |
virtual double | getScore () const =0 |
Returns the current score of the Environment. More... | |
virtual bool | isTerminal () const =0 |
Method for checking if the LearningEnvironment has reached a terminal state. More... | |
Additional Inherited Members | |
![]() | |
LearningEnvironment (const LearningEnvironment &other)=default | |
Make the default copy constructor protected. | |
![]() | |
const uint64_t | nbActions |
Interface for creating a Learning Environment in adversarial mode.
The main difference with the base Learning Environment is that several roots will participate in a single simulation. It is the user responsibility to correctly adapt this class to the problem. In fact, there should be 2 overrides that will be different from the basic Learning Environment :
|
inlineoverridevirtual |
Simply returns the first score of the results, allowing compatibility with non adversarial learning agents.
Inherited from LearningEnvironment
Implements Learn::LearningEnvironment.
|
pure virtual |
Computes scores of each root and returns them.