Fig. 10
From: Joint high-dimensional soft bit estimation and quantization using deep learning
Quantization performance of EQ-Net evaluated against state-of-the-art methods in large and very large MIMO, 16-QAM, i.i.d. Rayleigh fading, and LDPC-coded with size (324, 648). a \(8 \times 8\) MIMO, using MMSE-SIC soft bits for training and testing. b \(16 \times 16\) MIMO, using MMSE-SIC soft bits for training and testing. c \(32 \times 32\) MIMO, using MMSE-SIC soft bits for training and LMMSE soft bits for testing. d \(64 \times 64\) MIMO, using MMSE-SIC soft bits for training and LMMSE soft bits for testing