-
Run one of the MQPU examples on Perlmutter.
-
Interface the cudaq.observe call with the SLSQP optimizer from scipy.
-
Use the VQE H2 example from the docs and adapt it for LiH. Try different initializations as well as optimizers from other libraries. What parallelization technique can you use here?
-
Create an Ising Hamiltonian simulation. Try various scaling techniques.
-
Implement the QMCMC algorithm from this paper. https://arxiv.org/abs/2203.12497
-
Teleport a quantum state from Alice to Bob.
-
Extend the hybrid qnns example to include a multi-qubit circuit, why not add the 8-qubit aca dataset to try improve accuracy with quantum and classical nodes?
-
For the qnn example, code up a function to perform gradient descent by shifting parameters individually rather than simultaneously and compare the number of circuit evaluations as a function of epochs for both approaches.
-
Demonstrate the barren plateau phenomena - show that as you increase qubit sizes, the variance in the gradient decreases.
https://arxiv.org/abs/1803.11173 -
Evaluate the implementation of the quantum natural gradient optimization technique to accelerate VQE convergence.
https://arxiv.org/abs/1909.02108 -
Using a parameterized quantum circuit and a neural network, create a hybrid quantum GAN. https://arxiv.org/abs/2212.11614
https://pennylane.ai/qml/demos/tutorial_QGAN.html
https://pennylane.ai/qml/demos/tutorial_quantum_gans.html -
Circuit cutting is a powerful tool for running large quantum circuits on small computers. Implement this in CUDA Quantum.
https://arxiv.org/abs/2012.02333 -
Modify the qnn example, use a sci-kit learn dataset and implement the data reupload technique. https://arxiv.org/abs/1907.02085
-
Classical shadows is a powerful technique to reconstruct quantum states by measuring the circuit in random bases and postprocessing the data. Implement this in CUDA Quantum. https://arxiv.org/abs/2002.08953
https://pennylane.ai/qml/demos/tutorial_classical_shadows.html