This article proposes a multilayer perceptron (MLP)-based three-track detection method for dual-layer bit-patterned magnetic recording (BPMR) systems. Three different MLP architectures are explored and evaluated, namely: 1) a single MLP detecting all three tracks simultaneously; 2) three MLPs, each detecting one track independently; and 3) two MLPs dedicated to upper and lower recording layers. Simulation results show that the proposed MLP-based systems outperform the conventional partial-response maximum-likelihood (PRML) detection scheme, particularly under severe interferences and high areal density (AD). Among the proposed systems, the two-MLP architecture offers the optimal balance between detection accuracy and computational complexity, making it the most promising solution for future high-density magnetic recording systems.
