Circles datasets

To assess performance on nonlinearly separable structures, we generated geometric datasets using the make_circles functions from scikit-learn, supplemented by custom transformations to introduce structural diversity. The Circles dataset consisted of a two-cluster nested circular structure with 20,000 samples, where the factor parameter was set to 0.5, ensuring the inner circle radius was half that of the outer circle. The five cluster variant comprised concentric rings with 800 samples per cluster and radial scaling factors of 1.0, 2.0, 3.5, 5.0, and 7.0, ensuring distinct separation between rings.