bpp-phyl3  3.0.0
bpp::LogSumExp< R, T0, T1 > Class Template Reference

#include <Bpp/Phyl/Likelihood/DataFlow/DataFlowCWiseComputing.h>

+ Inheritance diagram for bpp::LogSumExp< R, T0, T1 >:
+ Collaboration diagram for bpp::LogSumExp< R, T0, T1 >:

Public Types

using Self = LogSumExp
 

Public Member Functions

 LogSumExp (NodeRefVec &&deps, const Dimension< T0 > &mDim)
 
std::string debugInfo () const override
 Node debug info (default = ""): user defined detailed info for DF graph debug. More...
 
bool compareAdditionalArguments (const Node_DF &other) const final
 Compare node-specific configuration to another. More...
 
NodeRef derive (Context &c, const Node_DF &node) final
 Returns a node computing d(this_node_expression)/d(node_expression). More...
 
NodeRef recreate (Context &c, NodeRefVec &&deps) final
 Recreate the node with different dependencies. More...
 
const R & targetValue ()
 Access value, recompute if needed. More...
 
const R & accessValueConst () const noexcept
 Raw value access (const). More...
 
ValueRef< R > deriveAsValue (Context &c, const Node_DF &node)
 
void modify (Callable &&modifier, bool makeValid)
 General case for modification of the T object. More...
 
bool isValid () const noexcept
 
std::size_t nbDependentNodes () const noexcept
 Number of dependent nodes (ie nodes that depend on this) More...
 
const std::vector< Node_DF * > & dependentNodes () const noexcept
 
std::size_t nbDependencies () const noexcept
 Number of dependencies (ie nodes we depend on) More...
 
const NodeRefVecdependencies () const noexcept
 
const NodeRefdependency (std::size_t i) const noexcept
 
virtual std::string description () const
 Node pretty name (default = type name). More...
 
virtual std::string color () const
 
virtual std::string shape () const
 
virtual bool hasNumericalProperty (NumericalProperty prop) const
 Test if the node has the given numerical property. More...
 
virtual std::size_t hashAdditionalArguments () const
 Return the hash of node-specific configuration. More...
 
void computeRecursively ()
 Compute this node value, recomputing dependencies (transitively) as needed. More...
 

Static Public Member Functions

static ValueRef< R > create (Context &c, NodeRefVec &&deps, const Dimension< T0 > &mDim)
 Build a new LogSumExp node with the given input matrix dimensions. More...
 

Protected Member Functions

R & accessValueMutable () noexcept
 
void invalidateRecursively () noexcept
 Invalidate (transitively) dependent nodes from this one. More...
 
void makeInvalid () noexcept
 
void makeValid () noexcept
 
void resetDependencies_ (NodeRefVec &&dependenciesArg)
 

Private Member Functions

void compute () final
 Computation implementation. More...
 
void registerNode (Node_DF *n)
 
void unregisterNode (const Node_DF *n)
 

Private Attributes

Dimension< T0 > mTargetDimension_
 
value_
 
NodeRefVec dependencyNodes_ {}
 
std::vector< Node_DF * > dependentNodes_ {}
 
bool isValid_ {false}
 

Detailed Description

template<typename R, typename T0, typename T1>
class bpp::LogSumExp< R, T0, T1 >

Definition at line 2028 of file DataFlowCWiseComputing.h.

Member Typedef Documentation

◆ Self

template<typename R , typename T0 , typename T1 >
using bpp::LogSumExp< R, T0, T1 >::Self = LogSumExp

Definition at line 2031 of file DataFlowCWiseComputing.h.

Constructor & Destructor Documentation

◆ LogSumExp()

template<typename R , typename T0 , typename T1 >
bpp::LogSumExp< R, T0, T1 >::LogSumExp ( NodeRefVec &&  deps,
const Dimension< T0 > &  mDim 
)
inline

Definition at line 2044 of file DataFlowCWiseComputing.h.

Member Function Documentation

◆ accessValueConst()

const R & bpp::Value< R >::accessValueConst ( ) const
inlinenoexceptinherited

Raw value access (const).

Value is not guaranteed to be valid (no recomputation).

Definition at line 385 of file DataFlow.h.

◆ accessValueMutable()

R & bpp::Value< R >::accessValueMutable ( )
inlineprotectednoexceptinherited

Raw value access (mutable). Should only be used by subclasses to implement compute().

Definition at line 416 of file DataFlow.h.

◆ color()

◆ compareAdditionalArguments()

template<typename R , typename T0 , typename T1 >
bool bpp::LogSumExp< R, T0, T1 >::compareAdditionalArguments ( const Node_DF other) const
inlinefinalvirtual

Compare node-specific configuration to another.

Required by Context for merging. It must compare everything in the node configuration except its type and dependencies. See DataFlowNumeric for examples. The default returns false, so nodes are considered different and not merged.

Reimplemented from bpp::Node_DF.

Definition at line 2054 of file DataFlowCWiseComputing.h.

◆ compute()

template<typename R , typename T0 , typename T1 >
void bpp::LogSumExp< R, T0, T1 >::compute ( )
inlinefinalprivatevirtual

Computation implementation.

This functions is defined by derived classes. It should compute the new node value from dependency node values. When called, dependency node are guaranteed to have valid values.

This function is private to prevent use for invalid dependencies. Higher level functions like computeRecursively call it while ensuring dependency validity.

Compute has access to dependencies as a NodeRefVec (base Node_DF classes only). The recommended usage is to check dependency types at Node construction. Then use static_cast to access derived classes efficiently from the NodeRefVec. See DataFlowNumeric.h for examples.

Implements bpp::Node_DF.

Definition at line 2081 of file DataFlowCWiseComputing.h.

References bpp::Value< R >::accessValueMutable(), bpp::numeric::cwise(), bpp::Node_DF::dependency(), bpp::exp(), bpp::numeric::isinf(), and bpp::log().

◆ computeRecursively()

void bpp::Node_DF::computeRecursively ( )
inherited

Compute this node value, recomputing dependencies (transitively) as needed.

Not thread safe !

Definition at line 183 of file DataFlow.cpp.

References bpp::Node_DF::isValid().

Referenced by bpp::Value< T >::targetValue().

◆ create()

template<typename R , typename T0 , typename T1 >
static ValueRef<R> bpp::LogSumExp< R, T0, T1 >::create ( Context c,
NodeRefVec &&  deps,
const Dimension< T0 > &  mDim 
)
inlinestatic

Build a new LogSumExp node with the given input matrix dimensions.

Definition at line 2034 of file DataFlowCWiseComputing.h.

References bpp::checkDependenciesNotNull(), and bpp::checkDependencyVectorSize().

◆ debugInfo()

template<typename R , typename T0 , typename T1 >
std::string bpp::LogSumExp< R, T0, T1 >::debugInfo ( ) const
inlineoverridevirtual

Node debug info (default = ""): user defined detailed info for DF graph debug.

Reimplemented from bpp::Node_DF.

Definition at line 2047 of file DataFlowCWiseComputing.h.

References bpp::Value< R >::accessValueConst(), and bpp::numeric::debug().

◆ dependencies()

◆ dependency()

const NodeRef& bpp::Node_DF::dependency ( std::size_t  i) const
inlinenoexceptinherited

Definition at line 185 of file DataFlow.h.

References bpp::Node_DF::dependencyNodes_.

Referenced by bpp::BackwardHmmLikelihood_DF::BackwardHmmLikelihood_DF(), bpp::ForwardHmmLikelihood_DF::build(), bpp::ValueFromConfiguredParameter::color(), bpp::CWiseFill< R, T >::compute(), bpp::CWiseApply< R, T, F >::compute(), bpp::CWiseAdd< R, std::tuple< T0, T1 > >::compute(), bpp::CWiseAdd< R, T >::compute(), bpp::CWiseMul< R, std::tuple< T0, T1 > >::compute(), bpp::CWiseDiv< R, std::tuple< T0, T1 > >::compute(), bpp::SumOfLogarithms< F >::compute(), bpp::CombineDeltaShifted< T >::compute(), bpp::CWiseSub< R, std::tuple< T0, T1 > >::compute(), bpp::CWiseMean< R, ReductionOf< T >, ReductionOf< P > >::compute(), bpp::CWiseMean< R, ReductionOf< T >, P >::compute(), bpp::CWiseNegate< T >::compute(), bpp::CWiseInverse< T >::compute(), bpp::CWiseLog< T >::compute(), bpp::CWiseExp< T >::compute(), bpp::CWiseConstantPow< T >::compute(), bpp::ScalarProduct< R, T0, T1 >::compute(), bpp::LogSumExp< R, T0, T1 >::compute(), bpp::MatrixProduct< R, T0, T1 >::compute(), bpp::ShiftDelta< T >::compute(), bpp::Convert< R, F >::compute(), bpp::Identity< R >::compute(), bpp::ProbabilitiesFromDiscreteDistribution::compute(), bpp::ProbabilityFromDiscreteDistribution::compute(), bpp::CategoryFromDiscreteDistribution::compute(), bpp::FrequenciesFromFrequencySet::compute(), bpp::EquilibriumFrequenciesFromModel::compute(), bpp::TransitionMatrixFromModel::compute(), bpp::TransitionFunctionFromModel::compute(), bpp::ProbabilitiesFromMixedModel::compute(), bpp::ProbabilityFromMixedModel::compute(), bpp::ShiftParameter::compute(), bpp::ValueFromConfiguredParameter::compute(), bpp::FrequenciesFromSimplex::compute(), bpp::EquilibriumFrequenciesFromTransitionMatrix::compute(), bpp::TransitionMatrixFromTransitionMatrix::compute(), bpp::CWisePattern< R >::compute(), bpp::CWiseMatching< R, ReductionOf< T > >::compute(), bpp::CWiseCompound< R, ReductionOf< T > >::compute(), bpp::ConfiguredParameter::compute(), bpp::CombineDeltaShifted< T >::create(), bpp::TransitionMatrixFromModel::debugInfo(), bpp::TransitionFunctionFromModel::debugInfo(), bpp::ValueFromConfiguredParameter::debugInfo(), bpp::CondLikelihood::debugInfo(), bpp::CWiseFill< R, T >::derive(), bpp::CWisePattern< R >::derive(), bpp::CWiseMatching< R, ReductionOf< T > >::derive(), bpp::CWiseCompound< R, ReductionOf< T > >::derive(), bpp::CWiseApply< R, T, F >::derive(), bpp::CWiseAdd< R, std::tuple< T0, T1 > >::derive(), bpp::CWiseSub< R, std::tuple< T0, T1 > >::derive(), bpp::CWiseAdd< R, ReductionOf< T > >::derive(), bpp::CWiseAdd< R, T >::derive(), bpp::CWiseMean< R, ReductionOf< T >, ReductionOf< P > >::derive(), bpp::CWiseMean< R, ReductionOf< T >, P >::derive(), bpp::CWiseMul< R, std::tuple< T0, T1 > >::derive(), bpp::CWiseMul< R, ReductionOf< T > >::derive(), bpp::CWiseDiv< R, std::tuple< T0, T1 > >::derive(), bpp::CWiseNegate< T >::derive(), bpp::CWiseInverse< T >::derive(), bpp::CWiseLog< T >::derive(), bpp::CWiseExp< T >::derive(), bpp::CWiseConstantPow< T >::derive(), bpp::ScalarProduct< R, T0, T1 >::derive(), bpp::SumOfLogarithms< F >::derive(), bpp::LogSumExp< R, T0, T1 >::derive(), bpp::MatrixProduct< R, T0, T1 >::derive(), bpp::ShiftDelta< T >::derive(), bpp::CombineDeltaShifted< T >::derive(), bpp::Convert< R, F >::derive(), bpp::Identity< R >::derive(), bpp::ProbabilitiesFromDiscreteDistribution::derive(), bpp::ProbabilityFromDiscreteDistribution::derive(), bpp::CategoryFromDiscreteDistribution::derive(), bpp::FrequenciesFromFrequencySet::derive(), bpp::EquilibriumFrequenciesFromModel::derive(), bpp::TransitionMatrixFromModel::derive(), bpp::TransitionFunctionFromModel::derive(), bpp::ProbabilitiesFromMixedModel::derive(), bpp::ProbabilityFromMixedModel::derive(), bpp::ShiftParameter::derive(), bpp::ValueFromConfiguredParameter::derive(), bpp::FrequenciesFromSimplex::derive(), bpp::EquilibriumFrequenciesFromTransitionMatrix::derive(), bpp::TransitionMatrixFromTransitionMatrix::derive(), bpp::BackwardHmmLikelihood_DF::derive(), bpp::ConfiguredParameter::derive(), bpp::ValueFromConfiguredParameter::description(), bpp::ConfiguredParameter::getValue(), bpp::ConfiguredParameter::recreate(), bpp::ConfiguredParameter::setValue(), and bpp::writeDotEdge().

◆ dependentNodes()

const std::vector<Node_DF*>& bpp::Node_DF::dependentNodes ( ) const
inlinenoexceptinherited

Definition at line 175 of file DataFlow.h.

References bpp::Node_DF::dependentNodes_.

◆ derive()

template<typename R , typename T0 , typename T1 >
NodeRef bpp::LogSumExp< R, T0, T1 >::derive ( Context c,
const Node_DF node 
)
inlinefinalvirtual

Returns a node computing d(this_node_expression)/d(node_expression).

The expression represented by 'node' is considered as a variable. Event if 'node' is a constant value node, d(node)/d(node) == 1. The derivative of a matrix is the matrix of the derivatives. Derivation is undefined by default, and this function will throw an exception. Implementations will usually recursively derive sub-expressions and combine them.

Reimplemented from bpp::Node_DF.

Definition at line 2059 of file DataFlowCWiseComputing.h.

References bpp::ConstantOne< T >::create(), bpp::ScalarProduct< R, T0, T1 >::create(), bpp::CWiseApply< R, T, F >::create(), bpp::CWiseExp< T >::create(), and bpp::Node_DF::dependency().

◆ deriveAsValue()

ValueRef<R > bpp::Value< R >::deriveAsValue ( Context c,
const Node_DF node 
)
inlineinherited

Derive and cast result as Value<T> (most nodes derive to the same value type).

Definition at line 389 of file DataFlow.h.

◆ description()

◆ hashAdditionalArguments()

std::size_t bpp::Node_DF::hashAdditionalArguments ( ) const
virtualinherited

Return the hash of node-specific configuration.

Compute a hash from additional arguments of the node. The hashed values must the same as the ones compared by compareAdditionalArguments. The default returns 0, which is ok if there are no additional arguments.

Reimplemented in bpp::Sequence_DF, bpp::ConfiguredParameter, bpp::ConfiguredModel, bpp::NumericConstant< T >, bpp::ShiftParameter, bpp::ConstantOne< T >, bpp::ConstantZero< T >, bpp::CombineDeltaShifted< T >, bpp::ShiftDelta< T >, bpp::CWiseConstantPow< T >, bpp::ConfiguredTransitionMatrix, bpp::ConfiguredSimplex, bpp::ConfiguredFrequencySet, and bpp::ConfiguredDistribution.

Definition at line 171 of file DataFlow.cpp.

◆ hasNumericalProperty()

bool bpp::Node_DF::hasNumericalProperty ( NumericalProperty  prop) const
virtualinherited

Test if the node has the given numerical property.

This is an optional indication only, used for optimisations. If unsure, leave it to always false (the default implementation). This should be non recursive, to ensure a constant time check.

Reimplemented in bpp::NumericConstant< T >, bpp::ConstantOne< T >, and bpp::ConstantZero< T >.

Definition at line 168 of file DataFlow.cpp.

Referenced by bpp::ScalarProduct< R, T0, T1 >::create(), bpp::CWisePattern< R >::create(), bpp::CWiseAdd< R, T >::create(), bpp::CWiseMean< R, ReductionOf< T >, ReductionOf< P > >::create(), bpp::CWiseDiv< R, std::tuple< T0, T1 > >::create(), bpp::CWiseNegate< T >::create(), bpp::CWiseInverse< T >::create(), bpp::CWiseLog< T >::create(), bpp::CWiseExp< T >::create(), bpp::CWiseConstantPow< T >::create(), and bpp::CombineDeltaShifted< T >::create().

◆ invalidateRecursively()

void bpp::Node_DF::invalidateRecursively ( )
protectednoexceptinherited

Invalidate (transitively) dependent nodes from this one.

Not thread safe !

Definition at line 219 of file DataFlow.cpp.

References bpp::Node_DF::isValid().

Referenced by bpp::Value< T >::modify(), and bpp::Node_DF::resetDependencies_().

◆ isValid()

bool bpp::Node_DF::isValid ( ) const
inlinenoexceptinherited

◆ makeInvalid()

void bpp::Node_DF::makeInvalid ( )
inlineprotectednoexceptinherited

Definition at line 270 of file DataFlow.h.

References bpp::Node_DF::isValid_.

Referenced by bpp::Node_DF::resetDependencies_().

◆ makeValid()

void bpp::Node_DF::makeValid ( )
inlineprotectednoexceptinherited

◆ modify()

void bpp::Value< R >::modify ( Callable &&  modifier,
bool  makeValid 
)
inlineinherited

General case for modification of the T object.

Takes a callable object (lambda, function pointer) that performs the modification. It must take a single T& as argument, which will refer to the T object to modify. The callable is called exactly once.

Parameters
modifiercallable to modify this object
makeValidboolean if this object is valid after modification (which means no call to this->compute()

Definition at line 405 of file DataFlow.h.

◆ nbDependencies()

◆ nbDependentNodes()

std::size_t bpp::Node_DF::nbDependentNodes ( ) const
inlinenoexceptinherited

Number of dependent nodes (ie nodes that depend on this)

Definition at line 172 of file DataFlow.h.

References bpp::Node_DF::dependentNodes_.

◆ recreate()

template<typename R , typename T0 , typename T1 >
NodeRef bpp::LogSumExp< R, T0, T1 >::recreate ( Context c,
NodeRefVec &&  deps 
)
inlinefinalvirtual

Recreate the node with different dependencies.

Reimplemented from bpp::Node_DF.

Definition at line 2075 of file DataFlowCWiseComputing.h.

References bpp::CWiseApply< R, T, F >::create().

◆ registerNode()

void bpp::Node_DF::registerNode ( Node_DF n)
privateinherited

Definition at line 240 of file DataFlow.cpp.

References bpp::Node_DF::dependentNodes_.

◆ resetDependencies_()

void bpp::Node_DF::resetDependencies_ ( NodeRefVec &&  dependenciesArg)
inlineprotectedinherited

◆ shape()

◆ targetValue()

const R & bpp::Value< R >::targetValue ( )
inlineinherited

Access value, recompute if needed.

Recompute the value if it is not up to date. Then access it as const. Recomputation is single threaded and not thread safe.

Definition at line 374 of file DataFlow.h.

◆ unregisterNode()

void bpp::Node_DF::unregisterNode ( const Node_DF n)
privateinherited

Definition at line 245 of file DataFlow.cpp.

References bpp::Node_DF::dependentNodes_.

Member Data Documentation

◆ dependencyNodes_

◆ dependentNodes_

std::vector<Node_DF*> bpp::Node_DF::dependentNodes_ {}
privateinherited

◆ isValid_

bool bpp::Node_DF::isValid_ {false}
privateinherited

◆ mTargetDimension_

template<typename R , typename T0 , typename T1 >
Dimension<T0> bpp::LogSumExp< R, T0, T1 >::mTargetDimension_
private

Definition at line 2100 of file DataFlowCWiseComputing.h.

◆ value_

R bpp::Value< R >::value_
privateinherited

Definition at line 419 of file DataFlow.h.


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