Proceedings of the AAAI Symposium Series · 6 Nov 2026
D4-Equivariant Hybrid Quantum–Classical Learning for Strong Gravitational Lens Classification
Sushmanth Reddy Mereddy, Pranath Reddy Kumbam, Michael W. Toomey, Sergei Gleyzer
Citations by database
Each database indexes a different set of journals, preprints and books, so they are shown side by side rather than blended.
- OpenCitations0
- Crossref0
Citations per year
The current year is still counting.
- 2026 so far0
Abstract
Strong gravitational lensing provides a powerful probe of small-scale dark matter structure, but extracting this information requires distinguishing weak morphological perturbations from source structure, instrumental effects, and noise while remaining insensitive to physically irrelevant image orientations. Although group-equivariant classical networks can encode rotational and reflection symmetry, it remains unclear whether a parameterized quantum circuit can provide additional representational capacity beyond a strong symmetry-aware classical model. We hypothesize that a symmetry-preserving quantum circuit can refine equivariant classical features. Our hybrid architecture lifts each image into its eight-view D₄ orbit, processes all views with a shared morphology-aware MBConv encoder, and maps the resulting features to four eight-qubit quantum heads through learned angular encoding. Symmetry-tied single- and two-qubit operations, repeated data encoding, and orbit-averaged Pauli measurements produce an invariant representation. We compare this quantum branch with a parameter-matched classical orbit control while holding the encoder, data partitions, and model-selection protocol fixed. Across three simulated lensing regimes containing cold dark matter, axion-like dark matter, and no-substructure classes, the hybrid model achieves three-seed mean test accuracies of 0.9855, 0.9987, and 0.9998. These scores exceed the classical control by 2.42, 1.41, and 0.85 percentage points, respectively, while also yielding lower negative log-likelihood on every dataset. The findings indicate that symmetry-structured quantum state evolution can measurably refine a strong equivariant classical representation. They establish a controlled benchmark for hybrid quantum learning in scientific imaging without constituting a claim of quantum advantage.
What the data shows
- No database records a citation yet, as is usual for a paper this new.
Graph
This paper in the middle, its authors and their most-cited papers below, the papers it cites to the left and the papers citing it to the right. The timeline lays every paper out by year.
Select a node for its details. Drag to pan, scroll to zoom, drag a node to move it.
Start with
Cited by 0 papers
Who cites this paper, from open citation data, with Crossref's record of each; the 100 newest when there are more.
References (0)
What it cites, from open citation data, with Crossref's record of each.