exatn::numerics::ContractionSeqOptimizerHeuro Class Reference

Inheritance diagram for exatn::numerics::ContractionSeqOptimizerHeuro:

Public Member Functions
void	resetNumWalkers (unsigned int num_walkers)
virtual double	determineContractionSequence (const TensorNetwork &network, std::list< ContrTriple > &contr_seq, std::function< unsigned int()> intermediate_num_generator) override

Static Public Member Functions
static std::unique_ptr< ContractionSeqOptimizer >	createNew ()

Protected Attributes
unsigned int	num_walkers_

Member Function Documentation

determineContractionSequence()

double exatn::numerics::ContractionSeqOptimizerHeuro::determineContractionSequence ( const TensorNetwork &  network, std::list< ContrTriple > &  contr_seq, std::function< unsigned int()>  intermediate_num_generator  )

overridevirtual

Determines the pseudo-optimal tensor contraction sequence required for evaluating a given tensor network. The unique intermediate tensor id's are generated by the provided intermediate number generator (each invocation returns a new tensor id). The latter can be conveniently passed as a lambda closure. The returned double value is an estimate of the total flop count associated with the determined contraction sequence. The tensor network must have at least two input tensors in order to get a single contraction. No contraction sequence is generated for tensor networks with a single input tensor.

Implements exatn::numerics::ContractionSeqOptimizer.

---

The documentation for this class was generated from the following files:

• /Users/aqw/exatn/src/numerics/contraction_seq_optimizer_heuro.hpp
• /Users/aqw/exatn/src/numerics/contraction_seq_optimizer_heuro.cpp