partitioned quantum reinforcmenet learning

Dimensionality reduction in quantum reinforcement learning agents.

Quantum reinforcement learning is typically done with a reuploading PQC. This scheme has been proven to work quite well, however, for a n dimensional state space one would need an n qubit circuit. This can become problematic, especially with the currnt size of quantum computers.

how to run?

Install everything etcetera and run the main.ipynb for a simple run

why is this so slow

I think the main problem that causes speed issues is the custom tensorflow model/layers. Simon used the regular tensorflow model which u compile and then run. Here we implement the circuit as a custom tensorflow layer which sits inside a custom tensorflow model.(this is a must i think if you want to assign optimizers to each trainable variable)
This makes it, unfortunatly, very slow.