Figure 4
From: Device quantization policy in variation-aware in-memory computing design
Weight distribution of 3-bit linear quantization (S-MLC, GH/GL = 10, σ = 0.05). (a) N-IMC mapping. Variation is higher at + w with high conductance in the linear scale. (b) WS-IMC mapping. Variation is symmetric at + w and − w.
