The TasNet authors have written a follow-up paper: [TasNet: Surpassing Ideal Time-Frequency Masking for Speech Separation](https://arxiv.org/abs/1809.07454) Do you have any plans to implement the algorithm described there?