Marmote Core
The project aims at realizing the prototype of a software environment dedicated to modeling with Markov chains.
|
Markov Chain class. More...
#include <markovChain.h>
Public Member Functions | |
markovChain (int sz, timeType t) | |
Simple constructor for the Markov chain from the size. More... | |
markovChain (transitionStructure *tr) | |
Constructor for the Markov chain using a transition structure. More... | |
markovChain (string format, string param[], int nbreParam, string modelName, bool isAbstract) | |
Constructor for Markov chains from files in various formats. In the abstract form, the object just stores the name(s) of the files that define the mode. In the non-abstract (concrete) form, the chain is instantiated in the memory with a concrete transition structure. Only the ERS, PSI and Xborne formats are supported at this time for concrete chains. More... | |
virtual | ~markovChain () |
Standard destructor. The generator and the initial distrib are destroyed. More... | |
int | stateSpaceSize () |
Read accessor to get the number of states in the state space of the Markov chain. More... | |
transitionStructure * | generator () |
Read accessor to get the value of _generator which is a transitionStructure. More... | |
void | setInitDistribution (discreteDistribution *d) |
Write accessor to set the value of _initDistribution which is a discreteDistribution. More... | |
void | setGenerator (transitionStructure *tr) |
Write accessor to set the value of _generator which is a transitionStructure. More... | |
RInside * | Rmotor () |
Read accessor to the embedded R engine, a static variable named _Rmotor. When accessed for the first time, an instance of RInside is created. More... | |
Distribution * | read () |
Method(s) for deserializing Distribution from Xborne file (.pi). More... | |
void | setFormat (string format) |
Utility to set the value of _format. More... | |
void | setModelName (string modelName) |
Utility to set the value of _modelName. More... | |
void | setAbstractNbre (int abstractNbre) |
Utility to set the value of _abstractNbre. More... | |
void | setAbstract (string abstract[]) |
Utility to set the value of the table containing names related to the model: file names, extensions etc. More... | |
int | abstractNbre () |
Utility to get _abstractNbre. More... | |
string | modelName () |
Utility to get _modelName. More... | |
string | format () |
Utility to get _format. More... | |
void | abstract () |
Utility to display the value of the table _abstract[]. More... | |
virtual simulationResult * | simulateChain (double tMax, bool Stats, bool Traj, bool withIncrements, bool Print) |
Simulates the evolution of a Markov Chain using the PSI program. This is a front-end function to both discrete-time and the continuous-time simulators. More... | |
virtual simulationResult * | simulateChainDT (int tMax, bool stats, bool traj, bool trace) |
Simulates the evolution of a discrete-time Markov Chain. More... | |
virtual simulationResult * | simulateChainCT (double tMax, bool stats, bool traj, bool withIncrements, bool trace) |
Simulates the evolution of a continuous-time Markov Chain. More... | |
simulationResult * | simulatePSI (int tMax, bool stats, bool Traj, bool Print) |
Simulates the evolution of a Markov Chain. This is a front-end function to the discrete-time simulator. For continuous time, the PSI simulator does not allow to control the simulation horizon. The effect of "Print" is to be handled in the specific simulation function, not here. More... | |
Distribution * | stationaryDistribution_iterative (string method, int tmax, double precision, string initDistribType, discreteDistribution *initDistrib, bool progress) |
Multi-purpose entry point for iterative methods for approximating the stationary distribution of discrete-time Markov chains. Offers the maximal flexibility with respect to parameters, as well as the possibility to set defaults. More... | |
Distribution * | stationaryDistribution (bool progress) |
Entry point for methods Computing the stationary distribution of chains using the iterative method. More... | |
virtual Distribution * | stationaryDistributionCT (bool progress) |
Computing the stationary distribution of a CTMC using uniformization and the iterative method. More... | |
virtual Distribution * | stationaryDistributionCT_embedding (int tMax, double epsilon, discreteDistribution *iDis, bool progress) |
Computing the stationary distribution of a CTMC using embedding and the iterative method. More... | |
virtual Distribution * | stationaryDistribution_power (int tMax, double epsilon, discreteDistribution *iDis, bool progress) |
Computing the stationary distribution of a DTMC using the standard iterative power method. More... | |
void | NCDProperty (double epsilon) |
Entry point for NCD command of XBORNE (Near Complete Decomposibility) More... | |
void | BandIMSUB (std::string modelName="modelName") |
Entry point for BandIMSUB command of XBORNE. More... | |
void | Vincent () |
Entry point for Vincent command of XBORNE. More... | |
void | RowVincent () |
Entry point for RowVincent command of XBORNE. More... | |
void | Absorbing () |
Entry point for Absorbing command of XBORNE: computes the absorbing states of a DTMC. More... | |
void | ProdFundSW (std::string modelName="modelName") |
Entry point for ProdFundSW command of XBORNE. More... | |
void | RowSum (std::string modelName="modelName") |
Entry point for RowSum command of XBORNE. More... | |
Distribution * | stationaryDistributionGthLD () |
Entry point for methods computing stationary distributions using the GTH method for solving the linear system. More... | |
Distribution * | stationaryDistributionSOR () |
Entry point for methods computing stationary distributions using the SOR method for solving the linear system. More... | |
Distribution * | transientDistributionR (int fromState, double t) |
Method for computing the transient distribution of a Markov chain. It wraps the 'solve.uc' method of the R package of L. Cerda-Alabern. More... | |
Distribution * | transientDistributionDT (int fromState, int t) |
Method for computing the transient distribution of a DTMC. Uses the general method evaluateMeasure. More... | |
Distribution * | stationaryDistributionR () |
Entry point for methods computing stationary distributions using the R package 'markovchain' for solving the linear system. More... | |
simulationResult * | simulateChainR (double tMax, bool stats, bool traj, bool trace) |
Simulates the evolution of a discrete-time Markov Chain. Wraps the 'rmarkovchain' method of the R package 'markovchain'. More... | |
double | transitionProbability (int stateFrom, int stateTo) |
Method to get the transition probabilities from some intial to some destination state. It wraps the 'transitionProbability' method of the R package 'markovchain'. More... | |
std::vector< int > | absorbingStates () |
Method that returns the list of absorbing states of the markovchain object. It wraps the 'absorbingStates()' method of the R package 'markovchain'. More... | |
std::vector< std::vector< int > > | recurrentClasses () |
Method that returns the recurrent classes of the markovchain object as a list of lists of states. It wraps the 'recurrentClasses()' method of the R package 'markovchain'. More... | |
std::vector< std::vector< int > > | communicatingClasses () |
Method that returns the communicating classes of the markovchain object as a list of lists of states. It wraps the 'communicatingClasses()' method of the R package 'markovchain'. More... | |
bool | isirreducible () |
Method verifing whether a Markov chain is irreducible. It wraps the "is.irreducible()' method of the R package 'markovchain'. More... | |
bool | isaccessible (int stateFrom, int stateTo) |
Method verifying if two states communicate in a Markov chain. Wraps the 'is.accessible()' method from the R package 'markovchain'. More... | |
simulationResult * | stationaryDistributionSample (int nbSamples) |
Methods to sample from the stationary distribution using backwards coupling. The result is returned in a simulationResult object, but the interpretation is different: it does not represent trajectories. Here, "state" entries are the measured states, and "time" entries are the measured coupling time. More... | |
Distribution * | hittingTimeDistribution (int iState, bool *hitSetIndicator) |
Entry point for methods computing the distribution of the hitting time (first entry times) from some state to some set of states. More... | |
int * | simulateHittingTime (int iState, bool *hittingSet, int nbSamples, int tMax) |
Obtain samples of hitting times through Monte Carlo simulation. A trajectory is simulated until it hits the target set, or its length attains a maximum, whichever comes first. Samples with the maximum are returned although they do not represent a proper hitting time. It is the responsibility of the calling party to ignore these values. More... | |
double * | averageHittingTime (bool *hitSetIndicator) |
Entry point for methods computing average hitting times (first entry times) from every state to some set of states. More... | |
double * | averageHittingTimeDT (bool *hitSetIndicator) |
Computing the average hitting times in a discrete-time Markov chain. Uses a direct Gauss-Seidel matrix inversion. More... | |
double * | averageHittingTimeDT_iterative (bool *hitSetIndicator) |
Computing the average hitting times in a discrete-time Markov chain. Uses an iterative approximate computation. More... | |
virtual markovChain * | copy () |
copy utility More... | |
virtual markovChain * | uniformize () |
Uniformize Markov Chain, by uniformizing the generator. If the chain is already discrete time, a copy is returned. More... | |
virtual markovChain * | embed () |
Construct discrete-time Markov Chain obtained at transition times. If the chain is already discrete time, a copy is returned. More... | |
void | setSizeType (const string path) |
Function to find out the size and the type of a Markov chain described in the MARCA format. Both quantities are directly set in the method, which returns nothing. Adapted from the method HBF::read_marka of Psi/Unix/v1.0. More... | |
int | charVectorElt2State (SEXP elt, std::string function) |
Utility function to convert the element of a Rcpp::CharacterVector to a state number. More... | |
virtual void | write (FILE *out, bool withReward) |
Method for writing Markov chains in a file with the ERS format. More... | |
virtual void | write (string format, string modelName) |
Method for writing Markov chains in files with various formats. The ERS and R formats are supported at this time. More... | |
void | write (string format) |
Method for writing Markov chains in files with various formats. This version uses the given model name. More... | |
std::string | toString (std::string format) |
String serialization method for a Markov chain. More... | |
Protected Types | |
typedef void * | RInside |
recast nonexisting RInside type to void More... | |
typedef void * | SEXP |
recast nonexisting SEXP type to void More... | |
Protected Attributes | |
timeType | _type |
time type: discrete or continuous More... | |
int | _stateSpaceSize |
size of the state space (should be a pointer on the state space itself) More... | |
transitionStructure * | _generator |
transition structure of the chain More... | |
discreteDistribution * | _initDistribution |
initial distribution of the process More... | |
bool | _debug |
internal debugging indicator More... | |
bool | _isAbstract |
true if the object is "abstract", i.e. a pointer to some files More... | |
int | _abstractNbre |
number of abstraction parameters More... | |
string * | _abstract |
table of abstraction parameters More... | |
string | _format |
format/language of the model More... | |
string | _modelName |
name of the model More... | |
Markov Chain class.
|
protected |
recast nonexisting RInside type to void
|
protected |
recast nonexisting SEXP type to void
markovChain::markovChain | ( | int | sz, |
timeType | t | ||
) |
Simple constructor for the Markov chain from the size.
sz | the size of the state space (may be infinite) |
t | the type of chain: CONTINUOUS or DISCRETE |
markovChain::markovChain | ( | transitionStructure * | tr | ) |
Constructor for the Markov chain using a transition structure.
tr | the transition structure |
markovChain::markovChain | ( | string | format, |
string | param[], | ||
int | nbreParam, | ||
string | modelName, | ||
bool | isAbstract | ||
) |
Constructor for Markov chains from files in various formats. In the abstract form, the object just stores the name(s) of the files that define the mode. In the non-abstract (concrete) form, the chain is instantiated in the memory with a concrete transition structure. Only the ERS, PSI and Xborne formats are supported at this time for concrete chains.
format | the format or language in which the model is specified |
param[] | is the list of parameters |
nbreParam | the size of param |
modelName | the name of the model, usually the prefix for various files |
isAbstract | specifies if the chain is abstract or not |
|
virtual |
Standard destructor. The generator and the initial distrib are destroyed.
void markovChain::Absorbing | ( | ) |
Entry point for Absorbing command of XBORNE: computes the absorbing states of a DTMC.
std::vector< int > markovChain::absorbingStates | ( | ) |
Method that returns the list of absorbing states of the markovchain object. It wraps the 'absorbingStates()' method of the R package 'markovchain'.
|
inline |
Utility to display the value of the table _abstract[].
|
inline |
Utility to get _abstractNbre.
double * markovChain::averageHittingTime | ( | bool * | hitSetIndicator | ) |
Entry point for methods computing average hitting times (first entry times) from every state to some set of states.
hitSetIndicator | a boolean array where states in the hitting set are marked with true |
double * markovChain::averageHittingTimeDT | ( | bool * | hitSetIndicator | ) |
Computing the average hitting times in a discrete-time Markov chain. Uses a direct Gauss-Seidel matrix inversion.
hitSetIndicator | a boolean array where states in the hitting set are marked with true |
double * markovChain::averageHittingTimeDT_iterative | ( | bool * | hitSetIndicator | ) |
Computing the average hitting times in a discrete-time Markov chain. Uses an iterative approximate computation.
hitSetIndicator | a boolean array where states in the hitting set are marked with true |
void markovChain::BandIMSUB | ( | std::string | modelName = "modelName" | ) |
Entry point for BandIMSUB command of XBORNE.
int markovChain::charVectorElt2State | ( | SEXP | elt, |
std::string | function | ||
) |
Utility function to convert the element of a Rcpp::CharacterVector to a state number.
elt | the element to be converted |
function | the name of the function where the call occurs, for tracing |
std::vector< std::vector< int > > markovChain::communicatingClasses | ( | ) |
Method that returns the communicating classes of the markovchain object as a list of lists of states. It wraps the 'communicatingClasses()' method of the R package 'markovchain'.
|
virtual |
copy utility
Copy a Markov Chain, including the generator and the initial distribution.
|
virtual |
Construct discrete-time Markov Chain obtained at transition times. If the chain is already discrete time, a copy is returned.
|
inline |
Utility to get _format.
|
inline |
Read accessor to get the value of _generator which is a transitionStructure.
Distribution* markovChain::hittingTimeDistribution | ( | int | iState, |
bool * | hitSetIndicator | ||
) |
Entry point for methods computing the distribution of the hitting time (first entry times) from some state to some set of states.
iState | index of the initial state |
hitSetIndicator | a boolean array where states in the hitting set are marked with true |
bool markovChain::isaccessible | ( | int | stateFrom, |
int | stateTo | ||
) |
Method verifying if two states communicate in a Markov chain. Wraps the 'is.accessible()' method from the R package 'markovchain'.
stateFrom | index of the origin state |
stateTo | index of the destination state |
bool markovChain::isirreducible | ( | ) |
Method verifing whether a Markov chain is irreducible. It wraps the "is.irreducible()' method of the R package 'markovchain'.
|
inline |
Utility to get _modelName.
void markovChain::NCDProperty | ( | double | epsilon | ) |
Entry point for NCD command of XBORNE (Near Complete Decomposibility)
epsilon | threshold for detecting decomposability |
void markovChain::ProdFundSW | ( | std::string | modelName = "modelName" | ) |
Entry point for ProdFundSW command of XBORNE.
Distribution * markovChain::read | ( | ) |
Method(s) for deserializing Distribution from Xborne file (.pi).
std::vector< std::vector< int > > markovChain::recurrentClasses | ( | ) |
Method that returns the recurrent classes of the markovchain object as a list of lists of states. It wraps the 'recurrentClasses()' method of the R package 'markovchain'.
RInside * markovChain::Rmotor | ( | ) |
Read accessor to the embedded R engine, a static variable named _Rmotor. When accessed for the first time, an instance of RInside is created.
void markovChain::RowSum | ( | std::string | modelName = "modelName" | ) |
Entry point for RowSum command of XBORNE.
void markovChain::RowVincent | ( | ) |
Entry point for RowVincent command of XBORNE.
void markovChain::setAbstract | ( | string | abstract[] | ) |
Utility to set the value of the table containing names related to the model: file names, extensions etc.
abstract | table with the strings to be copied to _abstract |
|
inline |
Utility to set the value of _abstractNbre.
abstractNbre | the number of abstract parameters to be set |
|
inline |
Utility to set the value of _format.
format |
|
inline |
Write accessor to set the value of _generator which is a transitionStructure.
tr | the transition structure to be set |
|
inline |
Write accessor to set the value of _initDistribution which is a discreteDistribution.
d | the distribution to be set |
|
inline |
Utility to set the value of _modelName.
modelName |
void markovChain::setSizeType | ( | const string | path | ) |
Function to find out the size and the type of a Markov chain described in the MARCA format. Both quantities are directly set in the method, which returns nothing. Adapted from the method HBF::read_marka of Psi/Unix/v1.0.
|
virtual |
Simulates the evolution of a Markov Chain using the PSI program. This is a front-end function to both discrete-time and the continuous-time simulators.
tMax | time until which the Markov chain is simulated |
Stats | indicates whether occupancy statistics are collected and returned |
Traj | indicates whether a trajectory is returned |
withIncrements | indicates whether time increments should be produced |
indicates whether values should be printed along the way |
Reimplemented in homogeneous1DBirthDeath, and felsenstein81.
|
virtual |
Simulates the evolution of a continuous-time Markov Chain.
tMax | time until which the Markov chain is simulated |
stats | indicates whether occupancy statistics are collected and returned |
traj | indicates whether a trajectory is returned |
withIncrements | indicates whether time increments should be printed |
trace | indicates whether the trajectory should be printed along the way (on stdout) |
|
virtual |
Simulates the evolution of a discrete-time Markov Chain.
tMax | time until which the Markov chain is simulated |
stats | indicates whether occupancy statistics are collected and returned |
traj | indicates whether a trajectory is returned |
trace | indicates whether the trajectory should be printed along the way (on stdout) |
simulationResult * markovChain::simulateChainR | ( | double | tMax, |
bool | stats, | ||
bool | traj, | ||
bool | trace | ||
) |
Simulates the evolution of a discrete-time Markov Chain. Wraps the 'rmarkovchain' method of the R package 'markovchain'.
tMax | time until which the Markov chain is simulated |
stats | indicates whether occupancy statistics are collected and returned |
traj | indicates whether a trajectory is returned |
trace | indicates whether the trajectory should be printed along the way (on stdout) |
int * markovChain::simulateHittingTime | ( | int | iState, |
bool * | hittingSet, | ||
int | nbSamples, | ||
int | tMax | ||
) |
Obtain samples of hitting times through Monte Carlo simulation. A trajectory is simulated until it hits the target set, or its length attains a maximum, whichever comes first. Samples with the maximum are returned although they do not represent a proper hitting time. It is the responsibility of the calling party to ignore these values.
iState | the initial state from which trajectories start |
hittingSet | boolean array indicating with true which states are in the target |
nbSamples | number of samples to collect |
tMax | maximum length of trajectories |
simulationResult * markovChain::simulatePSI | ( | int | tMax, |
bool | stats, | ||
bool | Traj, | ||
bool | |||
) |
Simulates the evolution of a Markov Chain. This is a front-end function to the discrete-time simulator. For continuous time, the PSI simulator does not allow to control the simulation horizon. The effect of "Print" is to be handled in the specific simulation function, not here.
tMax | time until which the Markov chain is simulated |
stats | indicates whether occupancy statistics are collected and returned |
Traj | indicates whether a trajectory is returned |
indicates whether values should be printed along the way |
|
inline |
Read accessor to get the number of states in the state space of the Markov chain.
Distribution * markovChain::stationaryDistribution | ( | bool | progress | ) |
Entry point for methods Computing the stationary distribution of chains using the iterative method.
progress | indicates whether the progress of the iterative method should be displayed |
Distribution * markovChain::stationaryDistribution_iterative | ( | string | method, |
int | tmax, | ||
double | precision, | ||
string | initDistribType, | ||
discreteDistribution * | initDistrib, | ||
bool | progress | ||
) |
Multi-purpose entry point for iterative methods for approximating the stationary distribution of discrete-time Markov chains. Offers the maximal flexibility with respect to parameters, as well as the possibility to set defaults.
method | a string describing the method to be called. Possibilities are "Power", "SOR". |
tmax | maximal number of iterations; defaults to 1000 if set to 0 |
precision | precision parameter for stopping rules; defaults to 1e-7 if set to 0 |
initDistribType | a string describing the initial distribution to be used in iterations. Possibilities are "Zero", "Max", "Uniform", "Custom". |
initDistrib | a distribution object in the case of "Custom" initial distribution. |
progress | boolean indicator for tracing the progression |
|
virtual |
Computing the stationary distribution of a DTMC using the standard iterative power method.
tMax | maximal number of iterations |
epsilon | precision parameter for the stopping rule |
iDis | initial distribution to be used |
progress | indicates whether the progress of the iterative method should be displayed |
|
virtual |
Computing the stationary distribution of a CTMC using uniformization and the iterative method.
progress | indicates whether the progress of the iterative method should be displayed |
|
virtual |
Computing the stationary distribution of a CTMC using embedding and the iterative method.
tMax | maximal number of iterations |
epsilon | precision parameter for the stopping rule |
iDis | initial distribution to be used |
progress | indicates whether the progress of the iterative method should be displayed |
Distribution * markovChain::stationaryDistributionGthLD | ( | ) |
Entry point for methods computing stationary distributions using the GTH method for solving the linear system.
Distribution * markovChain::stationaryDistributionR | ( | ) |
Entry point for methods computing stationary distributions using the R package 'markovchain' for solving the linear system.
simulationResult * markovChain::stationaryDistributionSample | ( | int | nbSamples | ) |
Methods to sample from the stationary distribution using backwards coupling. The result is returned in a simulationResult object, but the interpretation is different: it does not represent trajectories. Here, "state" entries are the measured states, and "time" entries are the measured coupling time.
nbSamples | number of samples to collect |
Distribution * markovChain::stationaryDistributionSOR | ( | ) |
Entry point for methods computing stationary distributions using the SOR method for solving the linear system.
std::string markovChain::toString | ( | std::string | format | ) |
String serialization method for a Markov chain.
format | the format/language to be used. |
Distribution * markovChain::transientDistributionDT | ( | int | fromState, |
int | t | ||
) |
Method for computing the transient distribution of a DTMC. Uses the general method evaluateMeasure.
Distribution * markovChain::transientDistributionR | ( | int | fromState, |
double | t | ||
) |
Method for computing the transient distribution of a Markov chain. It wraps the 'solve.uc' method of the R package of L. Cerda-Alabern.
double markovChain::transitionProbability | ( | int | stateFrom, |
int | stateTo | ||
) |
Method to get the transition probabilities from some intial to some destination state. It wraps the 'transitionProbability' method of the R package 'markovchain'.
|
virtual |
Uniformize Markov Chain, by uniformizing the generator. If the chain is already discrete time, a copy is returned.
void markovChain::Vincent | ( | ) |
Entry point for Vincent command of XBORNE.
|
virtual |
Method for writing Markov chains in a file with the ERS format.
out | the file descriptor in which to write the chain |
withReward | specifies if rewards are to be written; not used yet |
|
virtual |
Method for writing Markov chains in files with various formats. The ERS and R formats are supported at this time.
format | a string describing the format |
modelName | a string for naming the model, usually used for file name prefixes |
Reimplemented in homogeneous1DRandomWalk.
|
inline |
Method for writing Markov chains in files with various formats. This version uses the given model name.
format | a string describing the format |
|
protected |
table of abstraction parameters
|
protected |
number of abstraction parameters
|
protected |
internal debugging indicator
|
protected |
format/language of the model
|
protected |
transition structure of the chain
|
protected |
initial distribution of the process
|
protected |
true if the object is "abstract", i.e. a pointer to some files
|
protected |
name of the model
|
protected |
size of the state space (should be a pointer on the state space itself)
|
protected |
time type: discrete or continuous