For Quantum Embeddings for Machine Learning research paper, what exactly can be used as the Loss function for training (to reproduce Hybrid Model results)? They mention Linear Loss function and Risk function, but I am not sure which one of them has to be minimized to train.
Also, what is actually the training procedure ?
Hi @VIJAY_KARANJKAR , welcome to the Forum!
You’ll find a lot of PennyLane embedding templates in the docs. The QAOA embedding template is inspired by the ansatz proposed in the paper.
I believe the code for the experiment in Figure 5 of that paper was inspired by the work by Mari et al. (2019). The first author, Andrea Mari, actually wrote a nice PennyLane demo on quantum transfer learning, which you can download and run yourself.
Note that this is not the same code as the one used in the Quantum Embeddings for Machine Learning paper, but you can use it as inspiration too. This demo is several years old so it may not produce the exact same results as the paper, since PennyLane has evolved a lot in these past few years.
If you’re interested in newer approaches to quantum machine learning feel free to explore other PennyLane demos on this topic!
I hope this helps.