In deep clustering, the number of trainable parameters is different from [here](https://arxiv.org/abs/1809.07454).