sp
/
tempest
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

fixed weak bisim for dtmc 

Former-commit-id: 05462b99ec
main
dehnert 10 years ago
parent
29597e014f
commit
9475d29164
5 changed files with 51 additions and 84 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 5
      src/storage/bisimulation/BisimulationDecomposition.cpp
  2. 5
      src/storage/bisimulation/Block.cpp
  3. 110
      src/storage/bisimulation/DeterministicModelBisimulationDecomposition.cpp
  4. 5
      src/storage/bisimulation/DeterministicModelBisimulationDecomposition.h
  5. 10
      src/storage/bisimulation/Partition.cpp

5
src/storage/bisimulation/BisimulationDecomposition.cpp
View File

@ -160,9 +160,6 @@ namespace storm {
this->performPartitionRefinement();
std::chrono::high_resolution_clock::duration refinementTime = std::chrono::high_resolution_clock::now() - refinementStart;
std::cout << "final partition: " << std::endl;
this->partition.print();
std::chrono::high_resolution_clock::time_point extractionStart = std::chrono::high_resolution_clock::now();
this->extractDecompositionBlocks();
std::chrono::high_resolution_clock::duration extractionTime = std::chrono::high_resolution_clock::now() - extractionStart;
@ -197,7 +194,7 @@ namespace storm {
void BisimulationDecomposition<ModelType, BlockDataType>::performPartitionRefinement() {
// Insert all blocks into the splitter queue as a (potential) splitter.
std::deque<Block<BlockDataType>*> splitterQueue;
std::for_each(partition.getBlocks().begin(), partition.getBlocks().end(), [&] (std::unique_ptr<Block<BlockDataType>> const& block) { splitterQueue.push_back(block.get()); } );
std::for_each(partition.getBlocks().begin(), partition.getBlocks().end(), [&] (std::unique_ptr<Block<BlockDataType>> const& block) { block->data().setSplitter(); splitterQueue.push_back(block.get()); } );
// Then perform the actual splitting until there are no more splitters.
uint_fast64_t iterations = 0;

5
src/storage/bisimulation/Block.cpp
View File

@ -37,6 +37,11 @@ namespace storm {
template<typename DataType>
void Block<DataType>::print(Partition<DataType> const& partition) const {
std::cout << "block [" << this << "] " << this->id << " from " << this->beginIndex << " to " << this->endIndex << " with data [" << this->data() << "]" << std::endl;
std::cout << "states ";
for (auto stateIt = partition.begin(*this), stateIte = partition.end(*this); stateIt != stateIte; ++stateIt) {
std::cout << *stateIt << ", ";
}
std::cout << std::endl;
}
template<typename DataType>

110
src/storage/bisimulation/DeterministicModelBisimulationDecomposition.cpp
View File

@ -256,7 +256,7 @@ namespace storm {
}
template <typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::exploreRemainingStatesOfSplitter(bisimulation::Block<BlockDataType>& splitter) {
void DeterministicModelBisimulationDecomposition<ModelType>::exploreRemainingStatesOfSplitter(bisimulation::Block<BlockDataType>& splitter, std::list<bisimulation::Block<BlockDataType>*>& predecessorBlocks) {
for (auto splitterIt = this->partition.begin(splitter), splitterIte = this->partition.begin(splitter) + (splitter.data().marker2() - splitter.getBeginIndex()); splitterIt != splitterIte; ++splitterIt) {
storm::storage::sparse::state_type currentState = *splitterIt;
@ -268,7 +268,7 @@ namespace storm {
if (!possiblyNeedsRefinement(predecessorBlock)) {
continue;
}
// We keep track of the probability of the predecessor moving to the splitter.
increaseProbabilityToSplitter(predecessor, predecessorBlock, predecessorEntry.getValue());
@ -277,6 +277,8 @@ namespace storm {
if (predecessorPosition >= predecessorBlock.data().marker1()) {
moveStateToMarker1(predecessor, predecessorBlock);
}
insertIntoPredecessorList(predecessorBlock, predecessorBlocks);
}
}
@ -286,14 +288,14 @@ namespace storm {
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::updateSilentProbabilitiesBasedOnProbabilitiesToSplitter(bisimulation::Block<BlockDataType>& block) {
// For all states that do not have a successor in the block itself, we need to set the silent probability to 0.
std::for_each(silentProbabilities.begin() + block.data().marker1(), silentProbabilities.begin() + block.getEndIndex(), [] (ValueType& val) { val = storm::utility::zero<ValueType>(); } );
// For all states that do have a successor in the block, we set the silent probability to the probability
// stored in the vector of probabilities going to the splitter.
auto it = boost::make_zip_iterator(boost::make_tuple(silentProbabilities.begin() + block.getBeginIndex(), probabilitiesToCurrentSplitter.begin() + block.getBeginIndex()));
auto ite = boost::make_zip_iterator(boost::make_tuple(silentProbabilities.begin() + block.data().marker1(), probabilitiesToCurrentSplitter.begin() + block.data().marker1()));
std::for_each(it, ite, [] (boost::tuple<ValueType&, ValueType&> const& tuple) { boost::get<0>(tuple) = boost::get<1>(tuple); } );
// For all predecessors, we can set the probability to the current probability of moving to the splitter.
for (auto stateIt = this->partition.begin(block), stateIte = this->partition.begin() + block.data().marker1(); stateIt != stateIte; ++stateIt) {
silentProbabilities[*stateIt] = probabilitiesToCurrentSplitter[*stateIt];
}
// All non-predecessors have a silent probability of zero.
for (auto stateIt = this->partition.begin() + block.data().marker1(), stateIte = this->partition.end(block); stateIt != stateIte; ++stateIt) {
silentProbabilities[*stateIt] = storm::utility::zero<ValueType>();
}
}
template<typename ModelType>
@ -313,13 +315,11 @@ namespace storm {
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::computeConditionalProbabilitiesForNonSilentStates(bisimulation::Block<BlockDataType>& block) {
auto it = boost::make_zip_iterator(boost::make_tuple(silentProbabilities.begin() + block.getBeginIndex(), probabilitiesToCurrentSplitter.begin() + block.getBeginIndex()));
auto ite = boost::make_zip_iterator(boost::make_tuple(silentProbabilities.begin() + block.data().marker1(), probabilitiesToCurrentSplitter.begin() + block.data().marker1()));
std::for_each(it, ite, [this] (boost::tuple<ValueType&, ValueType&> const& tuple) {
if (!this->comparator.isZero(boost::get<0>(tuple))) {
boost::get<1>(tuple) /= storm::utility::one<ValueType>() - boost::get<0>(tuple);
for (auto stateIt = this->partition.begin() + block.getBeginIndex(), stateIte = this->partition.begin() + block.data().marker1(); stateIt != stateIte; ++stateIt) {
if (!this->comparator.isOne(getSilentProbability(*stateIt))) {
probabilitiesToCurrentSplitter[*stateIt] /= storm::utility::one<ValueType>() - getSilentProbability(*stateIt);
}
} );
}
}
template<typename ModelType>
@ -334,13 +334,11 @@ namespace storm {
// Now that we have the split points of the non-silent states, we perform a backward search from
// each non-silent state and label the predecessors with the class of the non-silent state.
std::vector<storm::storage::BitVector> stateLabels(block.getNumberOfStates(), storm::storage::BitVector(nonSilentBlockIndices.size() - 1));
std::cout << "creating " << block.getNumberOfStates() << " labels " << std::endl;
std::vector<storm::storage::sparse::state_type> stateStack;
stateStack.reserve(block.getNumberOfStates());
for (uint_fast64_t stateClassIndex = 0; stateClassIndex < nonSilentBlockIndices.size() - 1; ++stateClassIndex) {
for (auto stateIt = this->partition.begin() + nonSilentBlockIndices[stateClassIndex], stateIte = this->partition.begin() + nonSilentBlockIndices[stateClassIndex + 1]; stateIt != stateIte; ++stateIt) {
std::cout << "moving backward from state " << *stateIt << std::endl;
stateStack.push_back(*stateIt);
stateLabels[this->partition.getPosition(*stateIt) - block.getBeginIndex()].set(stateClassIndex);
@ -351,20 +349,13 @@ namespace storm {
for (auto const& predecessorEntry : this->backwardTransitions.getRow(currentState)) {
storm::storage::sparse::state_type predecessor = predecessorEntry.getColumn();
std::cout << " found pred " << predecessor << " with value " << predecessorEntry.getValue() << std::endl;
if (this->comparator.isZero(predecessorEntry.getValue())) {
continue;
}
// Only if the state is in the same block, is a silent state and it has not yet been
// labeled with the current label.
std::cout << "pred is at " << this->partition.getPosition(predecessor) << std::endl;
std::cout << "begin of block " << block.getBeginIndex() << std::endl;
std::cout << "accessing index " << (this->partition.getPosition(predecessor) - block.getBeginIndex()) << " of " << stateLabels.size() << std::endl;
std::cout << predecessor << " is in the same block as " << block.getId() << " ? " << (this->partition.getBlock(predecessor) == block) << std::endl;
// std::cout << isSilent(predecessor) << ", " << stateLabels[this->partition.getPosition(predecessor) - block.getBeginIndex()].get(stateClassIndex) << std::endl;
if (this->partition.getBlock(predecessor) == block && isSilent(predecessor) && !stateLabels[this->partition.getPosition(predecessor) - block.getBeginIndex()].get(stateClassIndex)) {
std::cout << "assigning label " << stateClassIndex << std::endl;
stateStack.push_back(predecessor);
stateLabels[this->partition.getPosition(predecessor) - block.getBeginIndex()].set(stateClassIndex);
}
@ -378,38 +369,20 @@ namespace storm {
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::refinePredecessorBlockOfSplitterWeak(bisimulation::Block<BlockDataType>& block, std::deque<bisimulation::Block<BlockDataType>*>& splitterQueue) {
// First, we need to turn the one-step probabilities to go to the splitter to the conditional probabilities
// for all non-silent states.
computeConditionalProbabilitiesForNonSilentStates(block);
// First, we need to compute a labeling of the states that expresses which of the non-silent blocks they can reach.
std::vector<uint_fast64_t> nonSilentBlockIndices = computeNonSilentBlocks(block);
std::cout << "indices: " << std::endl;
for (auto el : nonSilentBlockIndices) {
std::cout << el << std::endl;
}
for (int ind = 0; ind < nonSilentBlockIndices.size() - 1; ++ind) {
for (int inner = nonSilentBlockIndices[ind]; inner < nonSilentBlockIndices[ind + 1]; ++inner) {
std::cout << this->partition.getState(inner) << " is in class " << ind << std::endl;
}
}
// Then, we need to compute a labeling of the states that expresses which of the non-silent blocks they can reach.
std::vector<uint_fast64_t> nonSilentBlockIndices = computeNonSilentBlocks(block);
std::vector<storm::storage::BitVector> weakStateLabels = computeWeakStateLabelingBasedOnNonSilentBlocks(block, nonSilentBlockIndices);
std::cout << "labels: " << std::endl;
for (auto el : weakStateLabels) {
std::cout << el << std::endl;
}
// Then split the block according to this labeling.
// CAUTION: that this assumes that the positions of the states in the partition are not update until after
// the sorting is over. Otherwise, this interferes with the data used in the sorting process.
storm::storage::sparse::state_type originalBlockIndex = block.getBeginIndex();
auto result = this->partition.splitBlock(block,
auto split = this->partition.splitBlock(block,
[&weakStateLabels,&block,originalBlockIndex,this] (storm::storage::sparse::state_type state1, storm::storage::sparse::state_type state2) {
std::cout << "comparing states " << state1 << " and " << state2 << std::endl;
std::cout << (this->partition.getPosition(state1) - originalBlockIndex) << " and " << (this->partition.getPosition(state2) - originalBlockIndex) << std::endl;
std::cout << "size: " << weakStateLabels.size() << std::endl;
std::cout << "begin is " << block.getBeginIndex() << std::endl;
std::cout << "orig begin " << originalBlockIndex << std::endl;
return weakStateLabels[this->partition.getPosition(state1) - originalBlockIndex] < weakStateLabels[this->partition.getPosition(state2) - originalBlockIndex];
},
[this, &splitterQueue] (bisimulation::Block<BlockDataType>& block) {
@ -420,33 +393,41 @@ namespace storm {
splitterQueue.emplace_back(&block);
});
// If the block was split, we also update the silent probabilities of the
if (result) {
// If the block was split, we also update the silent probabilities.
if (split) {
updateSilentProbabilitiesBasedOnTransitions(block);
// Insert the new block as a splitter.
block.data().setSplitter();
splitterQueue.emplace_back(&block);
} else {
block.resetMarkers();
if (!block.data().splitter()) {
// Insert the new block as a splitter.
block.data().setSplitter();
splitterQueue.emplace_back(&block);
}
}
}
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::refinePredecessorBlocksOfSplitterWeak(bisimulation::Block<BlockDataType>& splitter, std::list<bisimulation::Block<BlockDataType>*> const& predecessorBlocks, std::deque<bisimulation::Block<BlockDataType>*>& splitterQueue) {
for (auto block : predecessorBlocks) {
std::cout << "found predecessor block " << block->getId() << std::endl;
if (*block != splitter) {
refinePredecessorBlockOfSplitterWeak(*block, splitterQueue);
} else {
// If the block to split is the splitter itself, we must not do any splitting here.
block->resetMarkers();
}
block->resetMarkers();
block->data().setNeedsRefinement(false);
}
}
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::insertIntoPredecessorList(bisimulation::Block<BlockDataType>& predecessorBlock, std::list<bisimulation::Block<BlockDataType>*>& predecessorBlocks) {
// Insert the block into the list of blocks to refine (if that has not already happened).
if (!predecessorBlock.data().needsRefinement()) {
predecessorBlocks.emplace_back(&predecessorBlock);
predecessorBlock.data().setNeedsRefinement();
}
}
template<typename ModelType>
void DeterministicModelBisimulationDecomposition<ModelType>::refinePartitionBasedOnSplitter(bisimulation::Block<BlockDataType>& splitter, std::deque<bisimulation::Block<BlockDataType>*>& splitterQueue) {
// The outline of the refinement is as follows.
@ -464,11 +445,6 @@ namespace storm {
//
// Finally, we use the information obtained in the first part for the actual splitting process in which all
// predecessor blocks of the splitter are split based on the probabilities computed earlier.
std::cout << "current partition is" << std::endl;
this->partition.print();
std::cout << "refining using splitter " << splitter.getId() << std::endl;
std::list<Block<BlockDataType>*> predecessorBlocks;
storm::storage::sparse::state_type currentPosition = splitter.getBeginIndex();
bool splitterIsPredecessorBlock = false;
@ -500,11 +476,7 @@ namespace storm {
moveStateInSplitter(predecessor, predecessorBlock, currentPosition, elementsToSkip);
}
// Insert the block into the list of blocks to refine (if that has not already happened).
if (!predecessorBlock.data().needsRefinement()) {
predecessorBlocks.emplace_back(&predecessorBlock);
predecessorBlock.data().setNeedsRefinement();
}
insertIntoPredecessorList(predecessorBlock, predecessorBlocks);
}
}
@ -516,7 +488,7 @@ namespace storm {
// If the splitter was a predecessor block of itself, we potentially need to explore some states that have
// not been explored yet.
if (splitterIsPredecessorBlock) {
exploreRemainingStatesOfSplitter(splitter);
exploreRemainingStatesOfSplitter(splitter, predecessorBlocks);
}
// Finally, we split the block based on the precomputed probabilities and the chosen bisimulation type.

5
src/storage/bisimulation/DeterministicModelBisimulationDecomposition.h
View File

@ -88,7 +88,7 @@ namespace storm {
void increaseProbabilityToSplitter(storm::storage::sparse::state_type predecessor, bisimulation::Block<BlockDataType> const& predecessorBlock, ValueType const& value);
// Explores the remaining predecessors of the splitter.
void exploreRemainingStatesOfSplitter(bisimulation::Block<BlockDataType>& splitter);
void exploreRemainingStatesOfSplitter(bisimulation::Block<BlockDataType>& splitter, std::list<bisimulation::Block<BlockDataType>*>& predecessorBlocks);
// Updates the silent probabilities of the states in the block based on the probabilities of going to the splitter.
void updateSilentProbabilitiesBasedOnProbabilitiesToSplitter(bisimulation::Block<BlockDataType>& block);
@ -113,6 +113,9 @@ namespace storm {
// Computes a labeling for all states of the block that identifies in which block they need to end up.
std::vector<storm::storage::BitVector> computeWeakStateLabelingBasedOnNonSilentBlocks(bisimulation::Block<BlockDataType> const& block, std::vector<uint_fast64_t> const& nonSilentBlockIndices);
// Inserts the block into the list of predecessors if it is not already contained.
void insertIntoPredecessorList(bisimulation::Block<BlockDataType>& predecessorBlock, std::list<bisimulation::Block<BlockDataType>*>& predecessorBlocks);
// A vector that holds the probabilities of states going into the splitter. This is used by the method that
// refines a block based on probabilities.
std::vector<ValueType> probabilitiesToCurrentSplitter;

10
src/storage/bisimulation/Partition.cpp
View File

@ -211,8 +211,6 @@ namespace storm {
std::pair<typename std::vector<std::unique_ptr<Block<DataType>>>::iterator, bool> Partition<DataType>::splitBlock(Block<DataType>& block, storm::storage::sparse::state_type position) {
STORM_LOG_THROW(position >= block.getBeginIndex() && position <= block.getEndIndex(), storm::exceptions::InvalidArgumentException, "Cannot split block at illegal position.");
std::cout << "splitting block " << block.getId() << " at pos " << position << std::endl;
// In case one of the resulting blocks would be empty, we simply return the current block and do not create
// a new one.
if (position == block.getBeginIndex() || position == block.getEndIndex()) {
@ -242,21 +240,13 @@ namespace storm {
auto originalBegin = block.getBeginIndex();
auto originalEnd = block.getEndIndex();
std::cout << "sorted block:" << std::endl;
for (auto stateIt = this->begin(block), stateIte = this->end(block); stateIt != stateIte; ++stateIt) {
std::cout << *stateIt << std::endl;
}
auto it = this->states.cbegin() + block.getBeginIndex();
auto ite = this->states.cbegin() + block.getEndIndex();
std::cout << "splitting between " << block.getBeginIndex() << " and " << block.getEndIndex() << std::endl;
bool wasSplit = false;
std::vector<storm::storage::sparse::state_type>::const_iterator upperBound;
do {
std::cout << "it (" << *it << ") less than ite-1 (" << *(ite-1) << "? " << less(*it, *(ite - 1)) << std::endl;
upperBound = std::upper_bound(it, ite, *it, less);
std::cout << "upper bound is " << std::distance(this->states.cbegin(), upperBound);
if (upperBound != ite) {
wasSplit = true;
auto result = this->splitBlock(block, std::distance(this->states.cbegin(), upperBound));

Write Preview
Loading…
Cancel
Save