GNN-Assisted BiG-AMP: Joint Channel Estimation and Data Detection for Massive MIMO Receiver

In massive MIMO systems, joint channel estimation and data detection (JCD) is essential for high spectral efficiency. Traditional bilinear generalized approximate message passing (BiG‑AMP) algorithms often lose performance because they use inaccurate marginal‑likelihood approximations and their variance estimates converge too early. This work introduces a graph‑neural‑network (GNN)‑assisted bilinear inference framework. The first version, called GNN‑BiGAMP, inserts a GNN module directly into the data‑detection loop to correct those approximation errors. Building on that, a bilinear GNN‑assisted BiG‑AMP (BiGNN‑BiGAMP) receiver is proposed. It places two similar GNNs inside the bilinear posterior‑estimation loops. The dual‑GNN design helps prevent the performance drop that occurs when variance converges prematurely. To train the system, a multitask

learning scheme coordinates the two GNNs across the channel‑estimation and data‑detection loops. Simulations reported in the paper show that these GNN‑assisted JCD receivers achieve much better channel‑estimation accuracy and data‑detection reliability than conventional JCD methods. For M.Tech researchers, the framework offers a solid base for exploring deep‑learning‑enhanced message‑passing algorithms in next‑generation wireless receivers.

Message: Hi FE, I need help with “GNN-Assisted BiG-AMP: Joint Channel Estimation and Data Detection for Massive MIMO Receiver” in “Communication Systems”