Multi-Task Learning for massive MIMO CSI Feedback

Deep learning-based massive MIMO CSI feedback has received a lot of attention in recent years. Now, there exists a plethora of CSI feedback models that exploit a wide variety of deep learning models and techniques ranging from convolutional neural networks (CNNs) to the recent attention-based transformer networks. Most of the models are based on auto-encoders (AE) architecture with an encoder network at the user equipment (UE) and a decoder network at the gNB (base station). However, these models are trained for a single user in a single channel scenario, making them ineffective in scenarios where a gNB is addressing various users while each user has different abilities and may employ a different CSI feedback encoder network and also in scenarios where the users are employing the same encoder network but are experiencing different channel conditions. In this work, we address these specific issues by exploiting the techniques of multi-task learning (MTL) in the context of massive MIMO CSI feedback.