TT-net: Quantum Inspired Tensor Network Denoising in Conditional GANs

TT-Net replaces single-cut SVD decomposition in conditional GAN image denoising with a two-cut tensor-train decomposition. This architecture allows the generator to access cross-channel information directly during feature map processing. In controlled comparisons, this method outperforms standard per-channel SVD denoising blocks.