Commit 5d10d792 authored Oct 01, 2025 by Joe Grimes

Checkpoint before switching to masking

parent 5180b6c4

src/loss_function/LossFunctionClustering.cpp

+11 −0

Original line number	Diff line number	Diff line
		@@ -76,6 +76,17 @@ LossFunctionClustering::LossFunctionClustering(std::vector<double> prop_train,
		set_nfeat(_n_feat);
		}

		//void LossFunctionClustering::populate_mask() {
		////Mask should be flattened or n class per task * n sample size of each prop train cluster
		//for (auto x : _prop_train){
		//
		// int len =
		//
		//}


		}

		LossFunctionClustering::LossFunctionClustering(const std::shared_ptr<LossFunction>& o)
		: LossFunctionClustering(
		o->prop_train(),

src/loss_function/LossFunctionClustering.hpp

+2 −0

Original line number	Diff line number	Diff line
		@@ -20,6 +20,7 @@ protected:
		std::vector<double> _a; //!< matrix to copy values from the D_Matrix
		std::map<int, std::vector<size_t>> _cluster_indices; //!< HashMap of the cluster label to its indices inside all feature values
		std::map<int, std::vector<double>> _centroids; //!< centroids of clusters to use for Davies Bouldin and Calinski
		std::vector<int> label_mask; //!< Label mask for prop train

		public:
		/**
		@@ -54,6 +55,7 @@ public:
		virtual double operator()(const std::vector<int>& inds) override;
		virtual double operator()(const std::vector<model_node_ptr>& feats) override;
		virtual double test_loss(const std::vector<model_node_ptr>& feats) override;
		// virtual void populate_mask();
		};

src/loss_function/LossFunctionDaviesBouldin.cpp

+1 −2

Original line number	Diff line number	Diff line
		@@ -44,9 +44,8 @@ double LossFunctionDaviesBouldin::get_dbi() {
		std::vector<double> mean(_n_feat, 0.0);
		for (size_t i = 0; i < indices.size(); i += _n_feat) {
		size_t samp = indices[i] / _n_feat;
		const double* p = &_a[samp * _n_feat];
		for (size_t f = 0; f < _n_feat; ++f) {
		mean[f] += p[f];
		mean[f] += _a[samp * _n_feat + f];
		}
		}
		for (double& x : mean){