In July this year, IBM organized their already traditional Qiskit Summer School, and I was lucky enough to be able to register and follow.
Qiskit is a programming framework for coding quantum algorithms on IBM quantum computers. It is developed and maintained not only by IBM developers, but also a large community of external developers and Qiskit Advocates (to which I also belong). Qiskit summer school this year (GQSS21) was focused on quantum machine learning, a very exciting field at the intersection between quantum computing and machine learning. During two weeks, we learned about topics like:
- Understanding noise in quantum computers, how to identify and how to model it, so to be able to clean up the results (error mitigation).
- Building quantum classifiers and different methods how to encode the classical data, how to create an expressive variational model that can work on NISQ computers, how to extract the labels and perform optimization of parameters, and how to make your model converges.
- Quantum approximate optimization algorithm, how it can be implemented through QUBO model Hamiltonian operator into variational quantum circuit and how it can be adiabatically evolved to get the optimal solution for your problem.
- Quantum kernels and how support vector machines can be implemented on quantum computers, but we have to design quantum kernels as non-trivial and classically hard to estimate.
- Different types of quantum neural networks and the practical aspects of their training
- Generative models like Quantum Boltzmann Machines and qGANs
After successful completion of the courses and lab exercises, I decided to share what I have learned, so I gave a presentation based on the course material. If you are interested in these topics, you can watch the recording here (QBM and qGANS are not covered):
Sasha Lazarevic
August 2021