tensorflow
Ansatz library brainstorming
The impression I got from tutorials was that in general, other than for random gates, the convention is that quantum circuit ansatz are constructed manually. (If I am mistaken about this point then please feel free to disregard.) Would it be possible to assemble a library of common ansatz architecture initialization functions that can be initialized in simpler fashion by just designating a few points like qubit count and etc? Please consider this brainstorming.
This is similar to something I previously discusses, with relation to input embeddings (see https://github.com/tensorflow/quantum/issues/391). Maybe having something like a template class (similar to pennylane https://pennylane.readthedocs.io/en/stable/introduction/templates.html) which has common data encoders and circuit structures is an idea. Not sure how much work that would require, so it may or may not be worthwhile.
Yeah is not clear to me how varied are types of common ansatz. The tensorflow quantum paper demonstrates a few, wasn’t sure if these forms are like authoritative accepted mainstream architectures or just starting points.