/*----------------------------------------------------------------------------------------------------------------------
| Computes the joint log working prior over all edges in the associated tree.
*/
double TreeScalerMove::recalcWorkingPriorForMove(
bool using_tree_length_prior, /*< true if using the Ranalla-Yang tree length prior */
bool using_vartopol_prior) const /*< true if using the Holder et al. variable topology reference distribution */
{
double ln_ref_dist = 0.0;
if (using_tree_length_prior)
{
ln_ref_dist = likelihood->getTreeLengthRefDist()->GetLnPDF(tree);
}
else if (using_vartopol_prior)
{
// Computes the log of the probability of the tree under Mark Holder's variable tree topology reference distribution
PHYCAS_ASSERT(topo_prob_calc);
std::pair<double, double> treeprobs = topo_prob_calc->CalcTopologyLnProb(*tree, true);
const double ln_ref_topo = treeprobs.first;
const double ln_ref_edges = treeprobs.second;
ln_ref_dist = ln_ref_topo + ln_ref_edges;
}
else
{
// Loop through all EdgeLenMasterParam objects and call the recalcWorkingPrior function of each.
// Each EdgeLenMasterParam object knows how to compute the working prior for the edge lengths it controls.
ChainManagerShPtr p = chain_mgr.lock();
const MCMCUpdaterVect & edge_length_params = p->getEdgeLenParams();
for (MCMCUpdaterVect::const_iterator it = edge_length_params.begin(); it != edge_length_params.end(); ++it)
{
if (!(*it)->isFixed())
ln_ref_dist += (*it)->recalcWorkingPrior();
}
}
return ln_ref_dist;
}
/*----------------------------------------------------------------------------------------------------------------------
| Calls the sample() member function of the `slice_sampler' data member.
*/
bool TreeScalerMove::update()
{
if (is_fixed)
return false;
ChainManagerShPtr p = chain_mgr.lock();
PHYCAS_ASSERT(p);
JointPriorManagerShPtr jpm = p->getJointPriorManager();
bool using_tree_length_prior = jpm->isTreeLengthPrior();
// If first edge length master param returns false for useWorkingPrior(), then we are using
// Mark Holder's variable tree topology reference distribution (see _provideRefDistToUpdaters in MCMCImpl.py)
MCMCUpdaterVect::const_iterator it = p->getEdgeLenParams().begin();
bool using_vartopol_prior = !(*it)->useWorkingPrior();
double prev_ln_prior = jpm->getLogJointPrior();
double prev_ln_like = p->getLastLnLike();
double prev_ln_ref_dist = (use_ref_dist ? recalcWorkingPriorForMove(using_tree_length_prior, using_vartopol_prior) : 0.0);
proposeNewState();
if (jpm->isTreeLengthPrior())
jpm->treeLengthModified("tree_length", tree);
else
jpm->allEdgeLensModified(tree);
curr_ln_prior = jpm->getLogJointPrior();
likelihood->useAsLikelihoodRoot(NULL); // invalidates all CLAs
double curr_ln_like = (heating_power > 0.0 ? likelihood->calcLnL(tree) : 0.0);
double curr_ln_ref_dist = (use_ref_dist ? recalcWorkingPriorForMove(using_tree_length_prior, using_vartopol_prior) : 0.0);
double prev_posterior = 0.0;
double curr_posterior = 0.0;
if (is_standard_heating)
{
prev_posterior = heating_power*(prev_ln_like + prev_ln_prior);
curr_posterior = heating_power*(curr_ln_like + curr_ln_prior);
if (use_ref_dist)
{
prev_posterior += (1.0 - heating_power)*prev_ln_ref_dist;
curr_posterior += (1.0 - heating_power)*curr_ln_ref_dist;
}
}
else
{
prev_posterior = heating_power*prev_ln_like + prev_ln_prior;
curr_posterior = heating_power*curr_ln_like + curr_ln_prior;
}
double ln_hastings = getLnHastingsRatio();
double ln_accept_ratio = curr_posterior - prev_posterior + ln_hastings;
double lnu = std::log(rng->Uniform());
bool accepted = false;
if (ln_accept_ratio >= 0.0 || lnu <= ln_accept_ratio)
{
if (save_debug_info)
{
debug_info = boost::str(boost::format("ACCEPT, forward_scaler = %.5f, prev_ln_prior = %.5f, curr_ln_prior = %.5f, prev_ln_like = %.5f, curr_ln_like = %.5f, lnu = %.5f, ln_accept_ratio = %.5f") % forward_scaler % prev_ln_prior % curr_ln_prior % prev_ln_like % curr_ln_like % lnu % ln_accept_ratio);
}
p->setLastLnLike(curr_ln_like);
accept();
accepted = true;
}
else
{
if (save_debug_info)
{
debug_info = boost::str(boost::format("REJECT, forward_scaler = %.5f, prev_ln_prior = %.5f, curr_ln_prior = %.5f, prev_ln_like = %.5f, curr_ln_like = %.5f, lnu = %.5f, ln_accept_ratio = %.5f") % forward_scaler % prev_ln_prior % curr_ln_prior % prev_ln_like % curr_ln_like % lnu % ln_accept_ratio);
}
curr_ln_like = p->getLastLnLike();
revert();
accepted = false;
}
//POLTMP
lambda = p->adaptUpdater(lambda, nattempts, accepted);
return accepted;
}
