Normalized Gradient Descent for Variational Quantum Algorithms
arXiv:2106.10981
Abstract
Variational quantum algorithms (VQAs) are promising methods that leverage noisy quantum computers and classical computing techniques for practical applications. In VQAs, the classical optimizers such as gradient-based optimizers are utilized to adjust the parameters of the quantum circuit so that the objective function is minimized. However, they often suffer from the so-called vanishing gradient or barren plateau issue. On the other hand, the normalized gradient descent (NGD) method, which employs the normalized gradient vector to update the parameters, has been successfully utilized in several optimization problems. Here, we study the performance of the NGD methods in the optimization of VQAs for the first time. Our goal is two-fold. The first is to examine the effectiveness of NGD and its variants for overcoming the vanishing gradient problems. The second is to propose a new NGD that can attain the faster convergence than the ordinary NGD. We performed numerical simulations of these gradient-based optimizers in the context of quantum chemistry where VQAs are used to find the ground state of a given Hamiltonian. The results show the effective convergence property of the NGD methods in VQAs, compared to the relevant optimizers without normalization. Moreover, we make use of some normalized gradient vectors at the past iteration steps to propose the novel historical NGD that has a theoretical guarantee to accelerate the convergence speed, which is observed in the numerical experiments as well.
References in corpus (18)
- Adam: A Method for Stochastic Optimization
- Quantum Computing in the NISQ era and beyond
- A variational eigenvalue solver on a quantum processor
- Hardware-efficient Variational Quantum Eigensolver for Small Molecules and Quantum Magnets
- Barren plateaus in quantum neural network training landscapes
- The theory of variational hybrid quantum-classical algorithms
- A Quantum Approximate Optimization Algorithm
- Quantum Circuit Learning
- Cost Function Dependent Barren Plateaus in Shallow Parametrized Quantum Circuits
- Evaluating analytic gradients on quantum hardware
- Circuit-centric quantum classifiers
- From the Quantum Approximate Optimization Algorithm to a Quantum Alternating Operator Ansatz
- Quantum Natural Gradient
- Quantum Supremacy through the Quantum Approximate Optimization Algorithm
- Equivalence of quantum barren plateaus to cost concentration and narrow gorges
- Learning to learn with quantum neural networks via classical neural networks
- Beyond Convexity: Stochastic Quasi-Convex Optimization
- On the natural gradient for variational quantum eigensolver