1. Epic GamesDeveloper
  3. FNeuralNetworkPolicy

FNeuralNetworkPolicy

Neural-network based policy object.

API > API/Plugins > API/Plugins/Learning

References

   
Module Learning
Header /Engine/Plugins/Experimental/LearningAgents/Source/Learning/Public/LearningPolicy.h
Include #include "LearningPolicy.h"

Syntax

struct FNeuralNetworkPolicy

Remarks

Neural-network based policy object. Stores various settings and intermediate storage required to evaluate the given network for the provided number of instances.

Constructors

  Type Name Description
Public function  

FNeuralNetworkPolicy ( const int32 InMaxInstanceNum,
const int32 InObservationEncodedNum,
const int32 InActionEncodedNum,
const int32 InMemoryStateNum,
const TSharedPtr< FNeuralNetwork >& InNeuralNetwork,
const FNeuralNetworkInferenceSettings& InInferenceSettings
)

 Create a new Policy Object which takes the encoded observation state, and the memory state, and outputs the encoded action state.

Functions

  Type Name Description
Public function void

Evaluate ( TLearningArrayView< 2, float > OutputActionVectorsEncoded,
TLearningArrayView< 2, float > OutputMemoryState,
const TLearningArrayView< 2, const float > InputObservationVectorsEncoded,
const TLearningArrayView< 2, const float > InputMemoryState,
const FIndexSet Instances
)

 Evaluate the Policy
Public function Const const TSharedPtr< FNeuralNetwork > & Gets the NeuralNetwork associated with this Policy
Public function void

UpdateNeuralNetwork ( const TSharedPtr< FNeuralNetwork >& NewNeuralNetwork
)

 Sets the NeuralNetwork and re-creates the NeuralNetworkInference object

Ask questions and help your peers Developer Forums

Write your own tutorials or read those from others Learning Library

On this page