This paper proposes a multi-layer perceptron (MLP)-based three-track detection method for dual-layer bit-patterned magnetic recording systems. Three architectures are explored: one MLP for three tracks, two MLPs for upper and lower layers, and three individual MLPs per track. Simulation results show that all MLP-based methods outperform conventional partial response maximum likelihood detection, especially under high areal density and complex interference. Among them, the two-MLP system achieves the best bit-error rate performance by effectively separating detection tasks across layers.
