Line data    Source code

       1             : //* This file is part of the MOOSE framework
       2             : //* https://mooseframework.inl.gov
       3             : //*
       4             : //* All rights reserved, see COPYRIGHT for full restrictions
       5             : //* https://github.com/idaholab/moose/blob/master/COPYRIGHT
       6             : //*
       7             : //* Licensed under LGPL 2.1, please see LICENSE for details
       8             : //* https://www.gnu.org/licenses/lgpl-2.1.html
       9             : 
      10             : #pragma once
      11             : 
      12             : #include "ActiveLearningReporterBase.h"
      13             : #include "ActiveLearningGaussianProcess.h"
      14             : #include "GaussianProcessSurrogate.h"
      15             : #include "SurrogateModelInterface.h"
      16             : 
      17             : class ActiveLearningGPDecision : public ActiveLearningReporterTempl<Real>,
      18             :                                  public SurrogateModelInterface
      19             : {
      20             : public:
      21             :   static InputParameters validParams();
      22             :   ActiveLearningGPDecision(const InputParameters & parameters);
      23             : 
      24             :   /// Access the number of training samples
      25          16 :   const int & getTrainingSamples() const { return _n_train; }
      26             : 
      27             : protected:
      28             :   /**
      29             :    * This is where most of the computations happen:
      30             :    *   - Data is accumulated for training
      31             :    *   - GP models are trained
      32             :    *   - Decision is made whether more data is needed for GP training
      33             :    */
      34             :   virtual void preNeedSample() override;
      35             : 
      36             :   /**
      37             :    * Based on the computations in preNeedSample, the decision to get more data is passed and results
      38             :    * from the GP fills @param val
      39             :    *
      40             :    * @param row Input parameters to the model
      41             :    * @param local_ind Current processor row index
      42             :    * @param global_ind All processors row index
      43             :    * @param val Output predicted by either the LF model + GP correction or the HF model
      44             :    * @return bool Whether a full order model evaluation is required
      45             :    */
      46             :   virtual bool needSample(const std::vector<Real> & row,
      47             :                           dof_id_type local_ind,
      48             :                           dof_id_type global_ind,
      49             :                           Real & val) override;
      50             : 
      51             :   /**
      52             :    * Make decisions whether to call the full model or not based on
      53             :    * GP prediction and uncertainty.
      54             :    *
      55             :    * @return bool Whether a full order model evaluation is required
      56             :    */
      57             :   virtual bool facilitateDecision();
      58             : 
      59             :   /**
      60             :    * This sets up data for re-training the GP.
      61             :    *
      62             :    * @param inputs Matrix of inputs for the current step
      63             :    * @param outputs Vector of outputs for the current step
      64             :    */
      65             :   virtual void setupData(const std::vector<std::vector<Real>> & inputs,
      66             :                          const std::vector<Real> & outputs);
      67             : 
      68             :   /**
      69             :    * This method evaluates the active learning acquisition function and returns bool
      70             :    * that indicates whether the GP model failed.
      71             :    *
      72             :    * @param gp_mean Mean of the gaussian process model
      73             :    * @param gp_mean Standard deviation of the gaussian process model
      74             :    * @return bool If the GP model failed
      75             :    */
      76             :   bool learningFunction(const Real & gp_mean, const Real & gp_std) const;
      77             : 
      78             :   /// The learning function for active learning
      79             :   const MooseEnum & _learning_function;
      80             :   /// The learning function threshold
      81             :   const Real & _learning_function_threshold;
      82             :   /// The learning function parameter
      83             :   const Real & _learning_function_parameter;
      84             : 
      85             :   /// Store all the input vectors used for training
      86             :   std::vector<std::vector<Real>> _inputs_batch;
      87             :   /// Store all the outputs used for training
      88             :   std::vector<Real> _outputs_batch;
      89             : 
      90             :   /// The active learning GP trainer that permits re-training
      91             :   const ActiveLearningGaussianProcess & _al_gp;
      92             :   /// The GP evaluator object that permits re-evaluations
      93             :   const SurrogateModel & _gp_eval;
      94             : 
      95             :   /// Flag samples when the GP fails
      96             :   std::vector<bool> & _flag_sample;
      97             : 
      98             :   /// Number of initial training points for GP
      99             :   const int _n_train;
     100             : 
     101             :   /// Storage for the input vectors to be transferred to the output file
     102             :   std::vector<std::vector<Real>> & _inputs;
     103             : 
     104             :   /// Broadcast the GP mean prediciton to JSON
     105             :   std::vector<Real> & _gp_mean;
     106             :   /// Broadcast the GP standard deviation to JSON
     107             :   std::vector<Real> & _gp_std;
     108             : 
     109             :   /// GP pass/fail decision
     110             :   bool _decision;
     111             : 
     112             :   /// Reference to global input data requested from base class
     113             :   const std::vector<std::vector<Real>> & _inputs_global;
     114             :   /// Reference to global output data requested from base class
     115             :   const std::vector<Real> & _outputs_global;
     116             : };