Mitigating Noise-Induced Gradient Vanishing in Variational Quantum Algorithm Training
This addresses a critical bottleneck for quantum computing practitioners by mitigating training issues on realistic hardware, representing a domain-specific incremental improvement.
The paper tackles the problem of noise-induced gradient vanishing in training variational quantum algorithms on near-term noisy quantum hardware by proposing a novel training scheme with a new cost function that augments gradients using traceless observables in truncated subspace, and experiments show it is highly effective for major algorithms across various tasks.
Variational quantum algorithms are expected to demonstrate the advantage of quantum computing on near-term noisy quantum computers. However, training such variational quantum algorithms suffers from gradient vanishing as the size of the algorithm increases. Previous work cannot handle the gradient vanishing induced by the inevitable noise effects on realistic quantum hardware. In this paper, we propose a novel training scheme to mitigate such noise-induced gradient vanishing. We first introduce a new cost function of which the gradients are significantly augmented by employing traceless observables in truncated subspace. We then prove that the same minimum can be reached by optimizing the original cost function with the gradients from the new cost function. Experiments show that our new training scheme is highly effective for major variational quantum algorithms of various tasks.