Layerwise learning for quantum neural networks

By Andrea Skolik, Jarrod R. McClean, Masoud Mohseni, Patrick van der Smagt, Martin Leib. Quantum Machine Intelligence, 2020

355 citations recorded by the source corpus.

With the increased focus on quantum circuit learning for near-term applications on quantum devices, in conjunction with unique challenges presented by cost function landscapes of parametrized quantum circuits, strategies for effective training are becoming increasingly important. In order to ameliorate some of these challenges, we investigate a layerwise learning strategy for parametrized quantum circuits. The circuit depth is incrementally grown during optimization, and only subsets of parameters are updated in each training step. We show that when considering sampling noise, this strategy can help avoid the problem of barren plateaus of the error surface due to the low depth of circuits, low number of parameters trained in one step, and larger magnitude of gradients compared to training the full circuit. These properties make our algorithm preferable for execution on noisy intermediate-scale quantum devices. We demonstrate our approach on an image-classification task on handwritten digits, and show that layerwise learning attains an 8% lower generalization error on average in comparison to standard learning schemes for training quantum circuits of the same size. Additionally, the percentage of runs that reach lower test errors is up to 40% larger compared to training the full circuit, which is susceptible to creeping onto a plateau during training.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/9fbf61fa1dec29f31eff0386f798734f5ef63db2

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

Citations in the corpus

Absence of Barren Plateaus in Quantum Convolutional Neural Networks (2020, background): “[49] Andrea Skolik, Jarrod R McClean, Masoud Mohseni, Patrick van der Smagt, and Martin Leib, “Layerwise learning for quantum neural networks,” arXiv preprint arXiv:2006.”
Noise-induced barren plateaus in variational quantum algorithms (2020, background): “Indeed, strategies to avoid noise-free barren plateaus [34, 37, 40– 43] do not appear to solve the NIBPs issue.”
Variational quantum algorithms (2020, methodology): “[178] introduced a method where one employs layerby-layer training.”
Diagnosing Barren Plateaus with Tools from Quantum Optimal Control (2021, background): “[54] Andrea Skolik, Jarrod R McClean, Masoud Mohseni, Patrick van der Smagt, and Martin Leib, “Layerwise learning for quantum neural networks,” arXiv preprint arXiv:2006.”
Connecting ansatz expressibility to gradient magnitudes and barren plateaus (2021, background): “[40] Andrea Skolik, Jarrod R McClean, Masoud Mohseni, Patrick van der Smagt, and Martin Leib, “Layerwise learning for quantum neural networks,” arXiv preprint arXiv:2006.”
Cost function dependent barren plateaus in shallow parametrized quantum circuits (2021, background): “Finally, we remark that some strategies have been developed to mitigate the effects of barren plateaus 32,33,53,54 .”
Quantum variational algorithms are swamped with traps (2022, neutral): “This is particularly exciting given the variety of no-go results via statistical learning theory, which demonstrate that in the worst case these models are not trainable via stochastic gradient-based methods [9, 52, 24, 51].”
General parameter-shift rules for quantum gradients (2021, background): “Recent studies have shown such coordinate descent methods to work well on many tasks [73, 50, 45, 74], although there are limited cases where these methods fail [75].”