Real-time adaptive control of quantum systems requires rapid generation of robust, high-fidelity pulses across a continuous range of operating conditions. Standard optimization algorithms such as gradient-ascent pulse engineering (GRAPE) solve each instance independently, discarding information between runs and requiring costly reinitialization when parameters change. We present an approach to robust optimal quantum control based on model-based reinforcement learning, in which a single neural network -- embedding the Hamiltonian directly into the training pipeline -- generates robust gates across an entire family of gate configurations, without pre-computed training data. Demonstrated on a single-spin (two-level) system, the trained networks produce pulses for arbitrary rotation angles over a range of pulse durations, detunings, and field inhomogeneities in milliseconds, at fidelities comparable to multi-seed GRAPE. The framework is inherently adaptable: any parameter entering the Hamiltonian can serve as a network input, extending the approach to different systems and control settings. Beyond speed, the network reveals structure in the control landscape: it discovers the same structured phase profiles that appear in GRAPE solutions -- made identifiable through fidelity-invariant symmetry transformations -- but more consistently than independent optimization. This consistency enables smooth interpolation across the entire trained parameter space.
Quantum optimal control is a versatile, powerful method to tailor quantum experiments. In NMR, numerical quantum optimal control algorithms such as the gradient ascent pulse engineering (GRAPE) algorithm have sparked many developments in pulse shaping, pushing the limits of pulse selectivity and robustness. To provide an overview of its core principles and the many expansions it has undergone ever since its introduction, the most common quality functions are benchmarked in their ability to guide the spin state transfer for facilitating excitation, inversion, and maximum quantum coherence. Explicit derivations of gradient and Hessian formulas are given, both summarizing what is known from the literature and complementing it. The derivations also give due consideration to the fact that spin states can be represented as density operators or spin operators. Analogously, the most common penalties to guide optimization toward experimentally feasible controls are discussed along with their explicit gradient and Hessian formula derivations. The discussion is complemented by thorough benchmarking for the most commonly used optimization algorithms.
We present a framework combining Lie group theory and feed-forward neural networks to efficiently generate quantum optimal control pulses for arbitrary unitary operations in superconducting qubit systems, bypassing the need for explicit optimization at inference time. The exponential scaling of the Hilbert space dimension with qubit number makes standard optimization approaches computationally prohibitive when large ensembles of distinct propagators must be processed, a bottleneck that is particularly acute in Trotterized quantum simulation. Our method addresses this limitation by pre-computing a representative set of control pulses via Lie group theory and training neural networks to map target propagators to their corresponding pulses. We demonstrate the approach on superconducting qubit systems of 2, 3, and 4 qubits, finding high reconstruction fidelity for specific combinations of Lie algebra parameters. As a physically motivated benchmark, we apply the methodology to reconstruct control pulses for the Trotter propagators of a neutrino system undergoing collective flavor oscillations. The successful generalization across system types demonstrates that a single model -- trained once on hardware-specific random propagators -- can serve as a universal control-pulse generator for any target quantum system of compatible Hilbert space dimension, offering a promising route toward scalable quantum simulation.
The gradient ascent pulse engineering (GRAPE) algorithm is a celebrated control algorithm with excellent converging rates, owing to a piecewise-constant ansatz for the control function that allows for cheap objective gradients. However, the computational effort involved in the exact simulation of quantum dynamics quickly becomes a bottleneck limiting the control of large systems. In this paper, we experiment with a modified version of GRAPE that uses Krylov approximations (K-GRAPE) to deal efficiently with high-dimensional state spaces. Even though the number of parameters required by an arbitrary control task scale linearly with the dimension of the system, we find a constant elementary computational effort (the effort per parameter). Since the elementary effort of GRAPE is superquadratic, this speed up allows us to reach dimensions far beyond. The performance of the K-GRAPE algorithm is benchmarked in the paradigmatic
XXZ spin-chain model.
An exquisite control over the dynamics of a quantum system is essential for realizing any quantum technology. This thesis presents some practical and efficient strategies for coherent quantum control: a paradigm for controlling quantum dynamics using electromagnetic fields. Coherent control is a common framework for describing the physical implementation of quantum logic gates in several quantum information processing platforms including superconducting qubits and trapped ions. For the purpose of demonstration, however, liquid-state Nuclear Magnetic Resonance (NMR) is used in this thesis as it is an ideal test-bed for developing, testing and benchmarking coherent control tools and techniques. Coherent control was revitalised in 2005 by the development of Gradient Ascent Pulse Engineering (GRAPE), a powerful optimal control technique that is widely used for designing shaped pulses for a variety of tasks. GRAPE, like any other closed-loop optimal control algorithm, requires multiple evaluations of the computationally expensive matrix exponential and the full time-evolution propagator. To this end, I present strategies to sidestep the explicit evaluation of matrix exponentials, thereby offering substantial speed-ups over conventional implementations of the GRAPE algorithm. The results are demonstrated experimentally in NMR, but also in simulations of ultracold molecules and superconducting qubits. Beyond GRAPE, this thesis also considers methods for designing spin-echo sequences for implementing any arbitrary network of controlled-Z gates in a system of qubits with zz-couplings. First, a method based on linear-programming provides near time-optimal sequences in fully-coupled systems having up to a few hundred qubits. Further, an analytic method based on graph colouring finds near time-optimal spin echo sequences in engineered systems with any number of qubits having only nearest and next-nearest neighbour-couplings.
Loop quantum gravity offers a rich non-perturbative framework for quantum gravity, yet its dynamics remain extremely difficult to implement in practice. In the canonical formulation, physical states are defined by quantum constraints acting on spin-network states, whose graph, representation, and intertwiner data lead to a highly complex kinematical state space. In the covariant formulation, transition amplitudes are built from spin foam vertex amplitudes whose exact evaluation is computationally demanding. This dissertation develops a computational programme for the dynamics of loop quantum gravity in both its canonical and covariant formulations. The programme combines structure-preserving numerical methods with machine learning techniques, with the aim of preserving the native graph, gauge, and operator structure of the theory while making its dynamical problems accessible to controlled computation at large scale. In the canonical theory, this means using variational Monte Carlo methods and neural quantum states to approximate solutions of quantum constraints. In the covariant theory, it means using neural network surrogates to approximate spin foam vertex amplitudes. In this way, part of the dynamical problem of loop quantum gravity is brought from the purely formal domain into a regime of explicit and controlled numerical study. On the canonical side, a fixed-graph framework is developed in which graph non-enlarging Hamiltonian constraints are recast as variational problems for positive quadratic forms whose minima encode weak constraint equations. This reformulation turns the search for physical states into a ground state problem on truncated Hilbert spaces. To realise this programme in practice, neuraLQX, an open-source high-performance toolkit for variational canonical loop quantum gravity built around graph-aware Hilbert spaces, observables, and constraint operators, was developed as part of this dissertation. It uses Markov chain Monte Carlo methods, and in particular gauge-invariant sampling, stochastic reconfiguration, and quantum states parametrized by neural networks adapted to the locality of the dynamics. To demonstrate the viability of this framework, high-fidelity near-physical states are obtained for Abelian BF theory and for a weak coupling limit of LQG with structure group
U(1)3 in three and four dimensions. In the four-dimensional model on the complete graph
K5, the analysis shows that the ordering of the vertex Hamiltonian constraint has a strong influence on which solution sectors are found variationally. The relevant operator is the vertex Hamiltonian
H^v, quantized in the spirit of Thiemann's construction and with the corresponding ordering. Numerically, the weak equations associated with
H^v and
H^v† are probed through positive quadratic objectives built from
H^v†H^v and
H^vH^v†. The analysis shows that these two searches do not merely produce different approximations to the same variational state. Instead, they select distinct solution sectors. One sector is peaked on flat connections and appears to describe a non-degenerate spatial quantum geometry. Numerical analysis shows this sector to be consistent with a discrete approximation to a Dittrich-Geiller-type vacuum. The other sector consists of states in the kernel of the volume operator. These states are non-flat and exhibit correlations across the graph, consistent with volume-zero excitations over an Ashtekar-Lewandowski-type vacuum. The relation between the two sectors is asymmetric in the sense that while the Ashtekar-Lewandowski-like states are often solutions for both orderings of the quadratic constraints, the Dittrich-Geiller-like states are not. Thus, one can conclude that the ordering of the Hamiltonian constraint affects which sectors are accessible or selected in the variational search, and it suggests that different orderings may exclude some classes of solutions. On the covariant side, this dissertation develops neural network surrogates for spin foam vertex amplitudes, with the aim of approximating the local building blocks of the path-integral dynamics. Starting with the Euclidean Barrett-Crane amplitude, the numerical problem is shown to separate naturally into two physically distinct tasks. The first is to determine which boundary spin configurations are admissible and therefore lie in the support of the amplitude. The second is to approximate the numerical value assigned to each admissible configuration by the vertex amplitude itself. The support is learned with near-perfect accuracy and generalises across cutoffs, showing that the network can infer and extrapolate the known, algorithmically verifiable
SU(2) recoupling conditions that determine admissibility in the Barrett-Crane case. By contrast, the amplitude values can be reproduced with high precision at fixed cutoff, but do not extrapolate reliably to larger cutoffs. This shows that the dependence of the vertex amplitude on the boundary quantum geometry is considerably more intricate than the structure of its support. The analysis also demonstrates that the representation of discrete spin data is not a merely technical modelling choice but directly affects how well the relevant quantum-geometric information is exposed to the surrogate model. This observation motivates graph-based surrogate architectures in which the spin foam boundary data and their adjacency structure are built into the model more explicitly. These architectures are then applied to the Lorentzian EPRL amplitude whereby in this case, the thesis first addresses the computational bottleneck of exact evaluation by reducing the runtime of the `sl2cfoam-next` exact evaluator sufficiently to generate large training datasets. Preliminary surrogate results show that the support, as well as the sign, of the EPRL amplitude can also be learned effectively, giving first evidence that graph-based learning can capture part of the admissibility structure of the Lorentzian theory.