edu.stanford.rsl.jpop
Interface FunctionAssembler
-
- All Known Implementing Classes:
- AdditiveFunctionAssembler
public interface FunctionAssemblerThis interface provides methods to assemble the information after the block-wise processing of the evaluate, gradient, and hessian.
-
-
Method Summary
All Methods Instance Methods Abstract Methods Modifier and Type Method and Description doubleassembleEvaluationBlocks(double[] blockResults)This method is provided with an array of block results and assembles them into a single function evaluation result.double[]assembleGradientBlocks(double[][] blockResults)This method is provided with an array of gradients which were obtained by the different processing blocks.double[][]assembleHessianBlocks(double[][][] blockResults)This method is provided with an array of hessians as computed fron the parallel procesing blocks.
-
-
-
Method Detail
-
assembleEvaluationBlocks
double assembleEvaluationBlocks(double[] blockResults)
This method is provided with an array of block results and assembles them into a single function evaluation result.- Parameters:
blockResults- the results of the processing blocks.- Returns:
- the evaluation result
-
assembleGradientBlocks
double[] assembleGradientBlocks(double[][] blockResults)
This method is provided with an array of gradients which were obtained by the different processing blocks. The methods assembles the data into a single gradient vector. The first index of the gradient array is supposed to refer to the block number.- Parameters:
blockResults- the block results- Returns:
- the assembled gradient vector
-
assembleHessianBlocks
double[][] assembleHessianBlocks(double[][][] blockResults)
This method is provided with an array of hessians as computed fron the parallel procesing blocks. The first index of the array should refer to the block index.- Parameters:
blockResults- the block results.- Returns:
- the assembled Hessian.
-
-
DMelt 3.0 © DataMelt by jWork.ORG