"FBGC:Dataset for Fast Bipartite Graph Clustering"

"This dataset is a collection of benchmark datasets used in the paper \"Fast Bipartite Graph Clustering (FBGC)\" for evaluating anchor-based graph clustering algorithms across diverse scales and complexity scenarios. It comprises three synthetic datasets (Flame, Two Moons, and Aggregation) and twelve real-world benchmark datasets (Segment, Mnist05, Waveform, Odr, Mnist10, Isolet, Caltech101, USPS, Pendigits, Letter, Mnist30K, and Sensorless). The sample sizes range from 240 to 58,509, feature dimensions from 16 to 784, and the number of clusters from 2 to 101. The datasets span multiple application domains, including image recognition, handwritten digit classification, speech processing, and sensor data, exhibiting diverse data distributions and geometric structures. The synthetic datasets feature non-convex and complex geometric shapes, effectively testing the ability of clustering algorithms to capture intrinsic data structures. The real-world datasets cover a wide spectrum from small-scale to large-scale scenarios, enabling comprehensive evaluation of clustering performance and computational efficiency under varying data characteristics. This collection provides a standardized benchmark reference for researchers working on anchor-based graph clustering, large-scale spectral clustering, and related unsupervised learning tasks."