A reinforcement learning-based channel denoising method is proposed to improve the accuracy of least squares channel estimation in MIMO OFDM systems without requiring prior channel knowledge or labeled training data.
