Session R11: V: Computer Algorithms to Advance Quantum Technology

Investigating Training Dynamics of Quantum Generative Adversarial Networks

The recent advent of Quantum Generative Adversarial Networks (QGANs) has marked a significant milestone in the integration of quantum computing with deep learning. QuGANs, with their reduced parameter sets and quantum-state-based gradients, offer a promising solution to some of the traditional limitations faced by classical GANs, such as computational intensity and mode collapse. This work focuses on investigating the training dynamics of the discriminator within the QGAN framework, a component critical to the overall performance yet challenging in terms of training stability and efficiency. Additionally, I provide insights into the practical aspects of QGAN implementation, including training duration and computational resources, which are often overlooked in theoretical models.   

A Quantum-Classical Hybrid Algorithm for Solving Satisfiability Problems

We design and test a hybrid algorithm to solve Boolean Satisfiability Problems (SAT). SAT problems are NP-complete and the algorithms used to solve them require exponential time at worst case. A quadratic speedup over classical algorithms can be obtained by using Quantu Amplitude Amplification. The algorithm we consider is built with IBM Quantum lab and Qiskit.   

Applications of Quantum Machine Learning (QML) for Quantum Simulation

In recent years, machine learning has revolutionized data analysis, extrapolation, and simulation capabilities. However, many of these classical simulation techniques face significant challenges when applied to quantum mechanical systems, primarily due to their complexity and high dimensionality. Traditionally, these methods often require the use of approximations or simplifications in order to be completely functional. Quantum computation provides a more suitable foundation for representing these systems, offering algorithmic increases in speed and efficiency for specific tasks. Leveraging this feature set, quantum machine learning models can be developed without the limitations inherent in classical analogs, significantly enhancing the potential efficiency and accuracy of certain simulations.   

Machine learning for efficient generation of universal photonic quantum computing resources

We present numerical results from simulations using deep reinforcement learning to control a measurement-based quantum processor—a time-multiplexed optical circuit sampled by photon-number-resolving detection—and find it generates squeezed cat states quasi-deterministically, with an average success rate of 98%, far outperforming all other proposals. Since squeezed cat states are deterministic precursors to the Gottesman-Kitaev-Preskill (GKP) bosonic error code, this is a key result for enabling fault tolerant photonic quantum computing.   

Pattern-based Functional Testing of Quantum Memories

With the growing number of qubits of quantum information processing devices, the task of fully characterizing these chips becomes increasingly unfeasible. From a practical perspective, one wants to find possible errors in the functioning of the device in as short time as possible, or otherwise establish its correct functioning with high confidence. In response to these challenges, we propose a pattern-based approach inspired by classical memory testing algorithms to evaluate the functionality of a quantum memory, based on plausible failure mechanisms. We demonstrate the method's capability to extract important qubit characteristics, such as T1 and T2 times, and to identify and analyse interactions between adjacent qubits. Additionally, our approach enables the recognition of different types of crosstalk effects and the detection of signatures indicating non-Markovian dynamics in individual qubits.   

Generation of entanglement via single-qubit rotation in torn Hilbert space

We present an efficient and simple protocol for generating arbitrary symmetric entangled states in a torn Hilbert space using only global single-qubit rotations.

The system is based on spin 1/2 qubits in resonators, such as atoms in optical cavities or superconducting qubits coupled to metallic microwave resonators. By sending light or microwaves into the resonator, it induces an AC Stark shift in the qubit's specific angular momentum eigenstate (Dick state). We can then generate barriers that impede transitions between adjacent Dick states,<br _istranslated="1" _mstmutation="1" /> And tear the original Hilbert space into pieces. Therefore, a simple global single-qubit rotation becomes very non-trivial, creating entanglement between many-body systems.

By optimally controlling the energy shift of the Dick state, we can generate any symmetric entangled state. We also illustrate that in just one or a few steps we can create a variety of useful states with near-uniform fidelity, including W states, spin-squeezed states (SSS), and Greenberger-Horne-Zeilinger (GHZ) states. In particular, SSS can only be created in one step with a compression parameter $\xi_R^2\sim1/N^{0.843}$ close to the Heisenberg limit (HL). Our findings establish a method for universal entanglement generation using only single-qubit actuation, where all multi-qubit control is integrated into a simple on/off microwave. It has direct applications in variational quantum optimizers available with existing technology.