← All papers
First page of How Far Can a Trainable Frequency Move? Resolution, Grids and Coupling in Trainable-FrequencyQuantum Models

How Far Can a Trainable Frequency Move? Resolution, Grids and Coupling in Trainable-FrequencyQuantum Models

Michael Poppel, Markus Baumann, Sebastian Wölckert, Claudia Linnhoff-Popien, Jonas Stein

cs.LG Feb 26, 2026 · v4 cs.AI cs.ET quant-ph
All Section 4 results on prefactor gradients and frequency reach are formalised in Lean 4 with Mathlib, with AI-assisted proofs checked by Lean.
Trainable-frequency quantum models train the prefactors of their encoding gates so that the circuit's frequencies can move onto the target's. From the unit start of prior work, training rarely reaches a target more than one resolution step away. We show that the target-driven part of the prefactor gradient is a weighted sum of slopes of the overlap $\sin u/u$, where $u$ is the model-target frequency distance scaled by the window half-width. This gives each frequency a reach: under gradient flow, an isolated real frequency whose coefficient adapts optimally reaches its target from within half a step, for every target at least one step from zero frequency; with a fixed coefficient, a phase-dependent barrier can stop it. A ternary prefactor grid places a model frequency within half a step of every band frequency, with logarithmically many encoding gates. Single-frequency circuits trained with Adam and jointly trained amplitudes behave as predicted: inside the first side lobe, a frequency starts toward its target exactly when in phase with it in 98% of runs, and only 11 of 768 frequencies starting beyond the lobe reached their target. With several frequencies, the grid start raised successful fits on synthetic targets from 69% to 84% against the best random start. With many target frequencies, least squares on the fixed grid fits near-perfectly, but the circuit does not reach the coefficients; training the prefactors narrows this shortfall, but with 8 and 24 target frequencies, training the fixed grid ten times longer fitted better. On real series, training the frequencies improved held-out $R^2$ at equal budget by 0.010–0.045 on four of seven series, but a classical screen did not separate these from the others. A classical model with independently trainable frequencies shows the same dependence on the start; our results describe how to train such models, not an advantage of the circuit.

Trainable-frequency quantum models train the prefactors of their encoding gates so that model frequencies can move onto the target's. Started from unit prefactors, training rarely reaches targets more than one resolution step away. The question is when frequency training succeeds reliably and what it adds.

The target-driven part of the prefactor gradient is written as a weighted sum of slopes of the sinc overlap. This yields reach results under gradient flow for isolated complex and real lines, and a ternary prefactor grid start that places a model frequency within half a step of every band frequency. The results are formalised in Lean 4 with Mathlib in one file. Proofs were produced with AI assistance (Aristotle, Claude), and two sign conditions were checked separately by interval arithmetic. The predictions are tested with Adam-trained circuits on synthetic targets and real time series.

Inside the first side lobe, a frequency starts toward its target exactly when in phase with it in 98% of runs, and only 11 of 768 frequencies starting beyond the lobe reached their target. With several frequencies, the grid start raised successful fits from 69% to 84% against the best random start. On real series, frequency training improved held-out R^2 by 0.010–0.045 on four of seven series, and a classical model with trainable frequencies shows the same dependence on the start.