By Ze Quantum Gros Bigs

Inspiration

Inspiration comes from classical machine learning algorithms and performance, as well as from the emergence of quantum computing and its interesting properties than can be used to create quantum layers and modules.

What it does

It can create hybrid models using classical image classifying (like with the MNIST dataset) and quantum components using convolutional kernels and qubits.

How I built it

We build it using well known alorithms provided by the module pytorch and databases provided by scikit-learn. For the quantum aspects, we used the pennylane module which provide wrapper for useful modules and where we can create various quantum circuits.

Challenges I ran into

The computation of quantum convolutions takes a really long time, we were not able to optimize it and Python doesn't help with looping efficiency. We ran into some problems related to GPU vs CPU computations: some Tensors were running on the CPU while components required it on GPU.

Accomplishments that I'm proud of

We are proud of what we accomplished durng this 24h period. We learned a lot about quantum machine learning and quantum algorithms in general. We also learned about what can currently be done in QML.