BIE::MultipleChains Class Reference

Runs the user-specified number of chains in parallel using the user-specified Monte Carlo algorithm (mca). More...

#include <MultipleChains.h>

Inheritance diagram for BIE::MultipleChains:

[legend]

Collaboration diagram for BIE::MultipleChains:

[legend]

List of all members.

Global variables
enum	Control { parallel, serial }
	Parallelization type. More...
static Control	cntrl
	Current control.
static unsigned	nfifo
	Number of states for acceptance rate queue.
static int	state_iter
	Number of iterations allowed for finding a good initial state.
unsigned	ncount
	Convergence check.
unsigned	totalnum
	Number of accepted states.
unsigned	totaltry
	Number of attempted states.
void	SampleNewState (int k)
	Attempt to find a good state from the prior distribution for chain k.
void	SampleNewStateParallel (int k)
	Sample new state (parallel implementation, called by generic member).
void	SampleNewStateSerial (int k)
	Sample new state (serial implementation, called by generic member).
Public Member Functions
	MultipleChains (int max, int ndim, int number, MHWidth *width, BaseDataTree *d, Model *m, Integration *i, Converge *c, MixturePrior *mp, LikelihoodComputation *l, MCAlgorithm *mca)
	Constructor for top level invocation.
	MultipleChains (int max, int ndim, int number, MHWidth *width, BaseDataTree *d, Model *m, Integration *i, Converge *c, MixturePrior *mp, LikelihoodComputation *l, MCAlgorithm *mca, Simulation *last)
	Constructor for invocation at higher levels of resolution.
virtual	~MultipleChains ()
	Destructor.
virtual void	Initialize ()
	Initialize all the chains.
virtual void	Reinitialize (MHWidth *width, MixturePrior *mp)
	Reinitialize with new class instances.
void	EnableLogging ()
	Enable caching per chain.
void	PrintState ()
	Print current state (override).
virtual void	PrintStepDiagnostic ()
	Override of Simulation method to provide diagnostics specific to TemperedSimulation.
virtual void	PrintStateDiagnostic ()
	Override of Simulation method to provide state specific diagnostics.
vector< double >	GetStat (void)
	Return vector with fraction of acceptances for each chain.
void	AdditionalInfo ()
	Print out simulation specific diagnostics.
void	NewState (int Mcur, vector< double > &w, vector< vector< double > > &p)
	Initialize new state (override from Simulation).
void	NewState (Chain &ch)
	Compute from a current Chain state.
void	NewState (State &s)
	Compute from a supplied State vector.
void	SetControl (int)
	Set global parameters.
Members which report and diagnose current state
double	GetValue ()
	Return current posterior probability value (override).
double	GetPrior ()
	Return current prior value (override).
double	GetLikelihood ()
	Return current likelihood probability value.
State	GetState ()
	Return current state (override).
virtual void	ReportState ()
	Print passed state in ascii with labels (override).
virtual void	LogState (string &logfile)
	Add state to log file.
virtual void	LogState (int level, int iterno, string &outfile)
	WITH level and iteration values.
Protected Member Functions
virtual void	MCMethod ()
	Compute one step in the Markov Chain.
void	CreateChains ()
	Create the chain structures.
void	ResetChainStats ()
	For gathering diagnostic data.
void	getChainCounts ()
Protected Attributes
MHWidth *	width
	The Metropolis-Hastings width instance.
bool	caching
	True if chain data is to be invidually logged.
vector< int >	stat0
vector< int >	stat1
bool	chains_initialized
	True if state values have been computed for at least the first time.
vector< unsigned char >	update_flag
	State variables.
vector< unsigned char >	update_flag1
	Chain values from this node need to be communicated.
vector< unsigned long >	nstat0
	Swaps proposed tallied for each chain.
vector< unsigned long >	nstat1
	Swaps accepted tallied for each chain.
deque< vector< unsigned char > >	arate
	Acceptence rates for each chain.
Friends
class	boost::serialization::access

---

Detailed Description

Runs the user-specified number of chains in parallel using the user-specified Monte Carlo algorithm (mca).

Each state will be written into the output log. Per chain statistics are computed and printed to stdout if the global variable mstat=1, as usual. This class will probably be used with a parallel chain convergence class such as GelmanRubinConverge.

Configuration

There are two obvious possibilities for arranging the computation.

1. The parallel likelihood classes should be used with parallel control only. The parallel likelihood classes LikelihoodComputationMPI and LikelihoodComputationMPITP assume that each process will enter the likelihood computation code.
2. A serial likelihood class should be used with serial control only. A serial likelihood computation is only entered by the process which calls it.

Diagnostic output

A call to the EnableLogging member function will cause the state of each individual chain to be logged to a separate file. The chain files will have names of the form nametag.chainlog.k where nametag is the global nametag variable and k is the chain number.

Example

See script8 in examples/User/parallel

---

Member Enumeration Documentation

enum BIE::MultipleChains::Control

Parallelization type.

Enumerator:

parallel	root controls all chain updates.
serial	chains are run at each node.

---

Member Function Documentation

void BIE::MultipleChains::NewState	(	int	Mcur,
		vector< double > &	w,
		vector< vector< double > > &	p
	)			[virtual]

Initialize new state (override from Simulation).

Compute from mixture cardinality, component weight vector and component parameter vectors

Reimplemented from BIE::Simulation.

void BIE::MultipleChains::SetControl

(

int

)

Set global parameters.

Choose parallelization control (parallel=0, serial=1)

virtual void BIE::MultipleChains::LogState

(

string &

logfile

)

[virtual]

Add state to log file.

WITHOUT level and iteration values

Reimplemented from BIE::Simulation.

virtual void BIE::MultipleChains::PrintStateDiagnostic

(

)

[virtual]

Override of Simulation method to provide state specific diagnostics.

Here, the method provides mixing statistics.

Reimplemented from BIE::Simulation.

void BIE::MultipleChains::SampleNewState

(

int

k

)

[protected]

Attempt to find a good state from the prior distribution for chain k.

Sample new state for chain k

---

Member Data Documentation

vector<unsigned char> BIE::MultipleChains::update_flag [protected]

State variables.

Chain values need to be communicated globally

unsigned BIE::MultipleChains::ncount [protected]

Convergence check.

Number of steps so far

---

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

• /home/weinberg/src/BIE/include/MultipleChains.h