Maximum number of QNN circuits allowed

PennyLane Help
1

Hi, I am implementing the QNN architecture proposed in “Continuous variable quantum neural networks” on the Breast Cancer Wisconsin data.

In order to create as many learnable parameters as possible, I want to set the number of layers to the highest that is possible. Is there a maximum number of layers that StrawberryFock simulators can handle?

2

Hi @sophchoe, there’s no programatic limit to the number of layers. This will be limited by your CPU.

3

Hi @CatalinaAlbornoz Does that mean I can run 10,000 layers, for example?

I want a network to have as many parameters as possible for better training. With 4 qumodes, that will bring the number of parameters to 320,000.

I am finishing up my experiment on the 4-qumode MNIST classifier. I will let you know when it’s done.

Thank you for all your help!

4

Hi @sophchoe, in theory you could have 10,000 layers. In practice your computer will probably run out of memory and your computation will fail. But if you had access to a supercomputer then there’s no line in StrawberryFields that says you can’t run more than a certain number of layers.

Let us know how it goes with this experiment!

5

@CatalinaAlbornoz Fabulous as usual! Thank you so much!

1 Like