1. Epic GamesDeveloper
  3. LearningAgentsTraining

LearningAgentsTraining

API > API/Plugins

Classes

  Type Name Description
Public struct UStruct BlueprintType category FLearningAgentsImitationTrainerSettings The configurable settings for a ULearningAgentsImitationTrainer.  
Public struct UStruct BlueprintType category FLearningAgentsImitationTrainerTrainingSettings The configurable settings for the training process.  
Public struct UStruct BlueprintType FLearningAgentsRecord A single recording of a series of observations and actions.  
Public struct UStruct BlueprintType category FLearningAgentsRecorderPathSettings The path settings for the recorder.  
Public struct UStruct BlueprintType category FLearningAgentsTrainerGameSettings The configurable game settings for a ULearningAgentsTrainer.  
Public struct UStruct BlueprintType category FLearningAgentsTrainerPathSettings The path settings for the trainer.  
Public struct UStruct BlueprintType category FLearningAgentsTrainerSettings The configurable settings for a ULearningAgentsTrainer.  
Public struct UStruct BlueprintType category FLearningAgentsTrainerTrainingSettings The configurable settings for the training process.  
Public class FLearningAgentsTrainingModule    
Public class UClass UConditionalCompletion A simple boolean completion.  
Public class UClass UConditionalReward A simple conditional reward that gives some constant reward value when a condition is true.  
Public class UClass UFloatReward A simple float reward.  
Public class UClass Abstract BlueprintType ULearningAgentsCompletion For functions in this file, we are favoring having more verbose names such as "AddConditionalCompletion" vs simply "Add" in order to keep it easy to find the correct function in blueprints.  
Public class UClass Blueprintable, BlueprintType ULearningAgentsImitationTrainer The ULearningAgentsImitationTrainer enable imitation learning, i.e. learning from human/AI demonstrations.  
Public class UClass Blueprintable, BlueprintType ULearningAgentsRecorder A component that can be used to create recordings of training data for imitation learning.  
Public class UClass BlueprintType ULearningAgentsRecording Data asset representing an array of records.  
Public class UClass Abstract BlueprintType ULearningAgentsReward For functions in this file, we are favoring having more verbose names such as "AddFloatReward" vs simply "Add" in order to keep it easy to find the correct function in blueprints.  
Public class UClass Abstract Blueprintable, BlueprintType ULearningAgentsTrainer The ULearningAgentsTrainer is the core class for reinforcement learning training.  
Public class UClass ULocalDirectionalVelocityReward A reward for maximizing velocity along a given local axis.  
Public class UClass UPlanarPositionDifferenceCompletion A completion for if two positions differ by some threshold in a plane, e.g. if the agent gets too far from a starting position.  
Public class UClass UPlanarPositionDifferencePenalty A penalty for being far from a goal position in a plane.  
Public class UClass UPlanarPositionSimilarityCompletion A completion for if two positions are near by some threshold in a plane, e.g. if the agent gets close to a position.  
Public class UClass UPositionArraySimilarityReward A reward for minimizing the distances of positions in the given arrays.  
Public class UClass UScalarVelocityReward A reward for maximizing speed.  
Public class UClass UTimeElapsedCompletion A completion for if a given amount of time has elapsed.  

Enums

  Type Name Description
Public enum UEnum BlueprintType category, meta ELearningAgentsCompletion Completion modes for episodes.  
Public enum UEnum BlueprintType category ELearningAgentsTrainerDevice Enumeration of the training devices.  

Ask questions and help your peers Developer Forums

Write your own tutorials or read those from others Learning Library

On this page