bug fix hyper-tune cost-sensitive

README.md

@@ -233,6 +233,9 @@ Note that STreeD provides an optimal decision tree for the given binarization. T
 
 See [examples/binarize_example.py](examples/binarize_example.py) for an example.
 
+See also our latest paper on optimizing trees with continuous features:
+* Briţa, Cătălin E., Jacobus G. M. van der Linden, and Emir Demirović. "Optimal Classification Trees for Continuous Feature Data Using Dynamic Programming with Branch-and-Bound." In _Proceedings of AAAI-25_ (2025). [pdf](https://arxiv.org/pdf/2501.07903) / [source](https://github.com/consol-Lab/contree)
+
 ## Overfitting and tuning
 
 To prevent overfitting the size of the tree can be tuned. This can be done in the standard way using `scikit-learn` methods, see [examples/gridsearch_example.py](examples/gridsearch_example.py). 

include/tasks/cost_sensitive.h

@@ -51,7 +51,13 @@ namespace STreeD {
 		static const int worst = INT32_MAX;
 		static const int best = 0;
 
-		CostSensitive(const ParameterHandler& parameters);
+		CostSensitive(const ParameterHandler& parameters) : Classification(parameters) { UpdateParameters(parameters); }
+		inline void UpdateParameters(const ParameterHandler& parameters) {
+			cost_filename = parameters.GetStringParameter("cost-file");
+		}
+		inline void CopyTaskInfoFrom(const OptimizationTask* task) {
+			UpdateCostSpecifier(static_cast<const CostSensitive*>(task)->cost_specifier);
+		}
 		void InformTrainData(const ADataView& train_data, const DataSummary& train_summary);
 		void UpdateCostSpecifier(const CostSpecifier& cost_specifier) { this->cost_specifier = cost_specifier; }
 

include/tasks/optimization_task.h

@@ -72,6 +72,7 @@ namespace STreeD {
 
 		// Inform the OT on updated parameters (when hypertuning)
 		inline void UpdateParameters(const ParameterHandler& parameters) { return; }
+		inline void CopyTaskInfoFrom(const OptimizationTask* task) {}
 
 		// Addition and subtraction functions
 		inline static int Add(const int left, const int right) { return left + right; }

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pystreed"
-version = "1.3.3"
+version = "1.3.4"
 requires-python = ">=3.8"						 
 description = "Python Implementation of STreeD: Dynamic Programming Approach for Optimal Decision Trees with Separable objectives and Constraints"
 license=  {file = "LICENSE"}

setup.py

@@ -6,7 +6,7 @@
 # Define package metadata
 package_name = 'pystreed'
 extension_name = 'cstreed'
-__version__ = "1.3.3"
+__version__ = "1.3.4"
 
 ext_modules = [
     Pybind11Extension(package_name + '.' + extension_name,

src/solver/solver.cpp

@@ -902,10 +902,11 @@ namespace STreeD {
 	}
 
 	template <class OT>
-	std::shared_ptr<SolverResult> Solver<OT>::HyperSolve(const ADataView& train_data) {
+	std::shared_ptr<SolverResult> Solver<OT>::HyperSolve(const ADataView& _train_data) {
 		using ScoreType = std::shared_ptr<Score>;
 		runtime_assert(parameters.GetBooleanParameter("hyper-tune"));
 		stopwatch.Initialise(parameters.GetFloatParameter("time"));
+		InitializeSolver(_train_data);
 
 		bool verbose = parameters.GetBooleanParameter("verbose");
 		const int max_num_nodes = int(parameters.GetIntegerParameter("max-num-nodes"));
@@ -929,6 +930,7 @@ namespace STreeD {
 				Solver<OT> solver(parameters, rng);
 				solver.solver_parameters.verbose = false;
 				solver.redundant_features = redundant_features;
+				solver.task->CopyTaskInfoFrom(this->task);
 				//ADataView::TrainTestSplitData<typename OT::LabelType>(train_data, sub_train_data, sub_test_data, rng, validation_percentage, true);
 				ADataView& sub_train_data = sub_train_datas[r], &sub_test_data = sub_test_datas[r];
 				solver.InitializeSolver(sub_train_data); // Initialize with max-depth

src/tasks/accuracy/cost_complex_accuracy.cpp

@@ -165,6 +165,7 @@ namespace STreeD {
 			alphas.push_back(base_alpha * a);
 		for(double alpha = 100*base_alpha; alpha < 0.01; alpha += 0.001)
 			alphas.push_back(alpha);
+		std::sort(alphas.begin(), alphas.end(), std::greater<>());
 		for (auto a: alphas) {
 			if (a > 0.1) continue;
 			ParameterHandler params = default_config;

src/tasks/cost_sensitive.cpp

@@ -127,10 +127,6 @@ namespace STreeD {
 		return max;
 	}
 
-	CostSensitive::CostSensitive(const ParameterHandler& parameters) : Classification(parameters) {
-		cost_filename = parameters.GetStringParameter("cost-file");
-	}
-
 	void CostSensitive::InformTrainData(const ADataView& train_data, const DataSummary& train_summary) {
 		OptimizationTask::InformTrainData(train_data, train_summary);
 		runtime_assert(cost_filename != "" || cost_specifier.IsInitialized());