REVIEW 3 cited by
Barren plateaus are swamped with traps
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Two main challenges preventing efficient training of variational quantum algorithms and quantum machine learning models are local minima and barren plateaus. Typically, barren plateaus are associated with deep circuits, while shallow circuits have been shown to suffer from suboptimal local minima. We point out a simple mechanism that creates exponentially many poor local minima specifically in the barren plateau regime. These local minima are trivial solutions, optimizing only a few terms in the loss function, leaving the rest on their barren plateaus. More precisely, we show the existence of approximate local minima, optimizing a single loss term, and conjecture the existence of exact local minima, optimizing only a logarithmic fraction of all loss function terms. One implication of our findings is that simply yielding large gradients is not sufficient to render an initialization strategy a meaningful solution to the barren plateau problem.
Forward citations
Cited by 3 Pith papers
-
Near-Optimal Parameter Tuning of Level-1 QAOA for Ising Models
For p=1 QAOA on Ising models, the paper derives analytic bandwidth bounds, eliminates the mixer angle to reduce optimization to a one-dimensional line search, and proves that for regular graphs the global optimum coin...
-
Learning complexity gradually in quantum machine learning models
Self-paced hard-example mining, which trains a quantum convolutional network on its ten highest-loss states each epoch, outperforms standard training on two spin-chain phase recognition benchmarks.
-
Q-MAML: Quantum Model-Agnostic Meta-Learning for Variational Quantum Algorithms
A classical network trained to output PQC initial parameters across Hamiltonian tasks gives faster VQE convergence in small simulations, but the approach is an extension of Meta-VQE rather than a true MAML implementation.
Discussion (0). Continue with ORCID to comment.