>>11370778To be fair to wavenet though, it's by the same authors as some of the papers I listed as things it "stole" from so it's a shameless republishing and not theft. For other examples of theft: see most papers about ordinal classification with neural networks that just copy-paste the same scheme as cheng et al., don't even cite it, and call it new.
One of the early published CNNs in the time domain for language (TCN for all intents and purposes) is probably
https://arxiv.org/pdf/1408.5882.pdfThe original TCN paper was probably
https://arxiv.org/pdf/1803.01271.pdfRather than WaveNet (on account of the former being in a different domain and the latter actually calling them TCN and calling them new).
They themselves note it's basically a 30 year old thing. The paper itself is about the fact that it works and shedding some light on why rather than claiming it's new (and better, they merely show that it works, not that it's the best shit ever and will solve all your problems), as most people have been doing recently. Regardless, as usual, it is just a new brand for an ancient thing.
