![[2408.12739] Quantum Convolutional Neural Networks are Effectively Classically Simulable](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2408.12739] Quantum Convolutional Neural Networks are Effectively Classically Simulable
Summary
The paper explores Quantum Convolutional Neural Networks (QCNNs) and their ability to be classically simulated, revealing insights about their performance on simple datasets.
Why It Matters
This research is significant as it challenges the perceived advantages of QCNNs in quantum machine learning, suggesting that their success may stem from benchmarking on simpler problems. This insight could guide future research towards more complex datasets necessary for advancing quantum machine learning.
Key Takeaways
- QCNNs can be classically simulated when initialized randomly.
- Their performance is tied to low-bodyness measurements of input states.
- Benchmarking on simple datasets may misrepresent their true capabilities.
- The study emphasizes the need for non-trivial datasets in quantum machine learning.
- Insights may extend to simulating other quantum architectures.
Quantum Physics arXiv:2408.12739 (quant-ph) [Submitted on 22 Aug 2024 (v1), last revised 20 Feb 2026 (this version, v2)] Title:Quantum Convolutional Neural Networks are Effectively Classically Simulable Authors:Pablo Bermejo, Paolo Braccia, Manuel S. Rudolph, Zoë Holmes, Lukasz Cincio, M. Cerezo View a PDF of the paper titled Quantum Convolutional Neural Networks are Effectively Classically Simulable, by Pablo Bermejo and 5 other authors View PDF Abstract:Quantum Convolutional Neural Networks (QCNNs) are widely regarded as a promising model for Quantum Machine Learning (QML). In this work we tie their heuristic success to two facts. First, that when randomly initialized, they can only operate on the information encoded in low-bodyness measurements of their input states. And second, that they are commonly benchmarked on "locally-easy'' datasets whose states are precisely classifiable by the information encoded in these low-bodyness observables subspace. We further show that the QCNN's action on this subspace can be efficiently classically simulated by a classical algorithm equipped with Pauli shadows on the dataset. Indeed, we present a shadow-based simulation of QCNNs on up-to $1024$ qubits for phases of matter classification. Our results can then be understood as highlighting a deeper symptom of QML: Models could only be showing heuristic success because they are benchmarked on simple problems, for which their action can be classically simulated. This insight points to th...
