arXiv:2609.17991
Fourier Analysis of Parametrized Interactive Quantum Classifiers
Fábio Novaes, Fernando M. de Paula Neto, João V. M. Cardoso
Abstract
Interactive Quantum Classifiers (IQCs) constitute a family of quantum machine learning models inspired by open quantum systems, in which the interaction between a target qubit and an environment is described by a Hamiltonian. Previous works introduced alternative Hamiltonian parameterizations and showed empirically that they can improve classification performance, but the role of these parameters in the resulting classifier remains poorly understood. In this work, we derive a closed-form expression for the reduced quantum channel generated by a parametrized IQC with a single target qubit. The analytical solution explicitly reveals how the Hamiltonian parameters control the constant, sine, and cosine components of the classifier output, establishing a Fourier interpretation of the induced feature map. This analysis motivates a generalized family of Hamiltonian encodings, including matrix-parameterized environmental Hamiltonians whose Fourier components depend on linear combinations of input features, thereby enabling non-separable Fourier structures. Numerical experiments on synthetic and real-world datasets show that the proposed models can improve classification performance on several nonlinear benchmarks. The generalized matrix encoding achieves the strongest aggregate performance in the evaluated benchmark, while a simpler four-parameter extension often attains comparable performance with substantially fewer trainable parameters. We additionally characterize the generated state ensembles using the standard fidelity-based expressibility measure, finding that global expressibility does not directly predict classification performance. Our results provide an analytical characterization of parametrized Hamiltonians in Interactive Quantum Classifiers and establish Fourier analysis as a useful framework for understanding and designing open-system-inspired quantum learning models.
AI Summary
Interactive Quantum Classifiers (IQCs) encode classical inputs through a target–environment interaction and predict from the reduced target state. This paper asks how their Hamiltonian parameters determine the represented function class, focusing on a single target qubit and a diagonal environmental Hamiltonian. For $H_t=\alpha_0 I+\boldsymbol\alpha\cdot\boldsymbol\sigma$ and $H_e=\operatorname{diag}(W\boldsymbol x)$, the reduced dynamics is derived as a random-unitary channel. A Bloch-vector calculation gives the output $f(\boldsymbol x)=C+\sum_l[A_l\cos(\boldsymbol\omega_l\cdot\boldsymbol x)+B_l\sin(\boldsymbol\omega_l\cdot\boldsymbol x)]$, with $\boldsymbol\omega_l=2\kappa\boldsymbol w_l$. The environment dimension $N_e$ bounds the number of indexed frequency components; their coefficients share constraints imposed by the target rotation axis, initialization, measurement, and environment probabilities. This motivates IQC-Multi, which replaces feature-wise encoding with trainable linear combinations and makes $N_e$ independent of input dimension. The amplitude-encoded variant is separately characterized as a diagonal quadratic form in normalized inputs. Classical simulations compare six variants on eleven datasets using ten train/test partitions and Adam or particle-swarm optimization. IQC-Multi with $N_e=8$ obtains aggregate accuracy $0.863$ versus $0.795$ for the original IQC; on XOR, the $N_e=4$ variant reaches $0.996$ while feature-wise variants remain near chance. Rankings depend on the dataset, and many pairwise differences are not statistically significant. Fidelity-distribution analyses find that global Haar-based expressibility does not directly predict classification performance. Stated limitations include shared Fourier coefficients and the mismatch between global-state expressibility and reduced-state predictions. Multiple target qubits, task-relevant expressibility measures, and scalable hardware implementations remain future directions.
Error found: The reported aggregate classification table is inconsistent with the arithmetic means of the eleven dataset results in the appendix, although IQCmulti[8] remains the highest-accuracy model on that calculation. The trace-normalized target Hamiltonian's identity term induces an environment-local unitary rather than a global phase, so omitting it preserves classifier predictions but can change the global-state fidelity distributions used to assess expressibility.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 34+9 pages, 8 figures