Expansion strategy that always picks the label with the greatest energy reduction. More...
#include <GreedyStrategy.hpp>
Inheritance diagram for GreedyStrategy< T >:
Collaboration diagram for GreedyStrategy< T >:
Public Member Functions | |
| GreedyStrategy (int max_cycles=100) | |
| Constructs the strategy. | |
| int | execute (AlphaExpansion< T > &optimizer, EnergyModel< T > &model) const override |
Runs greedy alpha-expansion until convergence or max_cycles. | |
 Public Member Functions inherited from ExpansionStrategy< T > | |
| virtual | ~ExpansionStrategy ()=default |
Detailed Description
template<typename T>
class GreedyStrategy< T >
class GreedyStrategy< T >
Expansion strategy that always picks the label with the greatest energy reduction.
Each cycle evaluates all labels and applies only the expansion move that yields the largest energy decrease. This is more expensive per cycle than SequentialStrategy (O(K) max-flow solves per cycle instead of one) but may converge in fewer cycles on some instances.
- Template Parameters
-
T Numeric cost type.
Constructor & Destructor Documentation
◆ GreedyStrategy()
template<typename T >
|
inline |
Constructs the strategy.
- Parameters
-
max_cycles Maximum number of greedy cycles before stopping (default: 100).
Member Function Documentation
◆ execute()
template<typename T >
|
inlineoverridevirtual |
Runs greedy alpha-expansion until convergence or max_cycles.
- Returns
- Number of full cycles completed.
Implements ExpansionStrategy< T >.
The documentation for this class was generated from the following file:
- src/strategies/GreedyStrategy.hpp
