In this paper, an unsupervised deep learning-based framework based on dual-path model-driven auto-encoders (AE) is proposed for angle-of-arrivals (AoAs) estimation in massive MIMO systems.

Existing learning-based arbitrary-scale point cloud upsampling methods are usually challenged with limited point cloud feature representation and noise-sensitive refinement of coarse point cloud. In ...

Key features: Bidirectional Autoencoder → same weights for encoder and decoder (50% fewer trainable params). Adaptive Wavelet Packet Decomposition → uses learnable filters (LFDWPT) to capture both ...