Novel Discrete Compactness-Based Training for Vector Quantization Networks: Enhancing Automatic Brain Tissue Classification

Table 2

𝐿_𝐶max values achieved with an object that defines a prism of squared base and height 256. The base’s lengths correspond to the mask sizes used in network topologies 𝜏^DC₁ and 𝜏^DC₂.

Segmentation by network	Mask size	𝐿𝐶max
𝜏^DC1	5 × 5	16,680
𝜏^DC2	4 × 4	10,264