Figure 9
From: Device quantization policy in variation-aware in-memory computing design
Weight distribution of 3-bit linear quantization (GH/GL = 10, σ = 0.05 for 1-bit cell) using (a) the WS-IMC + D-MLC mapping scheme and (b) the WS-IMC + A-MLC mapping scheme, showing tighter distributions than S-MLC in Fig. 4b.
