wail-prec-sbgd

WAIL-PREC-SBGD: Pattern Recognition using Score-Based System & Gradient Descent Current C++ Implementation (classificator module) The project now includes a concrete demo pipeline in the classificator folder that does two core tasks: Build a neighbor graph from active points (spatial relationships). Build chains of neighbors (connected components) and classify each chain with a score-based model trained with gradient descent. Data flow Point activation by saturation - isActivePoint(...) converts RGB to HSV-style saturation and marks a point active when saturation is above a threshold. Neighbor generation - neighboorArray(...) treats each point as a graph node. - Two points become neighbors when distance is within either point radius. - Neighbor links are stored symmetrically. Chain generation - generateNeighborChains(...) runs BFS on the neighbor graph. - Each connected component becomes one NeighborChain. Feature extraction per chain - avgSaturation: average saturation across points in the chain. - compactness: inverse-normalized average pairwise distance. - density: local graph connectivity. - sizeNorm: normalized chain size.