✗ Failed on benchmark 2026

Proximal-Mismatch Fine-Tuning

Implementation & benchmark of arXiv:2607.14894 — Domain Adaptation of Mismatched Proximal Denoiser for Plug-and-Play Image Reconstruction

Usefulness7/10

Difficulty5/10

Novelty6/10

Regularization Accuracy Stability Sample efficiency Loss Fine tuning Regularization Optimizer Convex analysis Optimization Numerical analysis

Source paper: Domain Adaptation of Mismatched Proximal Denoiser for Plug-and-Play Image Reconstruction arXiv:2607.14894 ⓘ · analyzed Aug 30, 2026

AI-generated research hypothesis, automatically tested. Not peer-reviewed.

Idea description

Fine-tune a denoiser by matching its action to a target-domain proximal operator, instead of minimizing only pixelwise denoising error. Apply the loss on the intermediate states and noise levels actually encountered by the downstream iterative solver, so the adaptation directly reduces the error that controls PnP reconstruction stability.

Formulas

$$F(\boldsymbol{x})=\eta f(\boldsymbol{x})+R(\boldsymbol{x})$$

$$\mathsf{D}_{\star}=\operatorname{prox}_{R_{\star}},\qquad \operatorname{prox}_{\gamma R_{\star}}(\boldsymbol{v})=\arg\min_{\boldsymbol{z}}\left\{R_{\star}(\boldsymbol{z})+\frac{1}{2\gamma}\|\boldsymbol{z}-\boldsymbol{v}\|_2^2\right\}$$

$$\delta(\boldsymbol{v})=\left\|\widehat{\mathsf{D}}(\boldsymbol{v})-\mathsf{D}_{\star}(\boldsymbol{v})\right\|_2$$

$$\min_{0\le k<K}\operatorname{Stat}(\boldsymbol{x}_k)\;\lesssim\;\frac{C}{K}+\frac{1}{K}\sum_{k=0}^{K-1}\left\|\widehat{\mathsf{D}}(\boldsymbol{v}_k)-\mathsf{D}_{\star}(\boldsymbol{v}_k)\right\|_2^2$$

Mathematical statement

The paper formulates the regularized target problem as $F(\boldsymbol{x})=\eta f(\boldsymbol{x})+R(\boldsymbol{x})$, where $f$ is the data-fidelity function, $R$ is the target-domain regularizer, and $\eta>0$ weights data consistency. The ideal denoising operator is $\mathsf{D}_{\star}=\operatorname{prox}_{R_{\star}}$, with $\operatorname{prox}_{\gamma R_{\star}}(\boldsymbol{v})=\arg\min_{\boldsymbol{z}}\{R_{\star}(\boldsymbol{z})+(2\gamma)^{-1}\|\boldsymbol{z}-\boldsymbol{v}\|_2^2\}$; $\gamma>0$ is the proximal step size. The deployed network is $\widehat{\mathsf{D}}$. Define the proximal mismatch at input $\boldsymbol{v}$ as $\delta(\boldsymbol{v})=\|\widehat{\mathsf{D}}(\boldsymbol{v})-\mathsf{D}_{\star}(\boldsymbol{v})\|_2$. The paper's key transfer principle is that PnP stationarity has an $\mathcal{O}(1/K)$ optimization term plus an additive mismatch term proportional to the average squared error: $\min_{0\le k<K}\operatorname{Stat}(\boldsymbol{x}_k)\lesssim C/K+(1/K)\sum_{k=0}^{K-1}\delta(\boldsymbol{v}_k)^2$, where $K$ is the number of PnP iterations, $C$ collects initial-objective and step-size factors, and $\boldsymbol{v}_k$ is the denoiser input. We use this as the training objective: minimize empirical $\delta^2$ over target-domain solver states.

Implementation notes

(1) Integration point: use an image denoiser $D_\theta$ inside a PnP proximal-gradient reconstruction loop. For measurements $y$, forward operator $A$, and differentiable data loss $f(x)=\tfrac12\|Ax-y\|_2^2$, form $v_k=x_k-\gamma\eta A^T(Ax_k-y)$ and update $x_{k+1}=D_\theta(v_k,\sigma_k)$. During adaptation, modify the denoiser training loss, not the PnP optimizer.

(2) Pseudocode: sample a target-domain clean image $x$, generate a measurement $y=Ax+e$, and initialize $z_0$. For $k=0,\ldots,K-1$, compute $v_k=z_k-\gamma\eta A^T(Az_k-y)$, compute $u_k=D_\theta(v_k,\sigma_k)$, and compute a target proximal output $t_k$. If a differentiable target proximal teacher $D_\star$ is available, set $t_k=D_\star(v_k,\sigma_k)$. Otherwise obtain $t_k$ from a strong frozen teacher denoiser, a converged domain-specific PnP solver, or an inner optimization $t_k\approx\arg\min_z\{R_\star(z)+(2\gamma)^{-1}\|z-v_k\|^2\}$. Optimize $L_{prox}=\frac1{BK}\sum_{b,k}w_k\|D_\theta(v_{b,k},\sigma_k)-t_{b,k}\|_2^2+\lambda L_{MSE}$, then update $z_{k+1}=\operatorname{stopgrad}(u_k)$ while collecting states. Use $w_k$ to emphasize late or high-mismatch states.

(3) Computed from the paper's mathematics: the training signal is squared proximal mismatch and the predicted reconstruction floor is its empirical average. Estimated empirically: $D_\star$ or its proxy, the state distribution $v_k$, and $\operatorname{Stat}(x_k)$, measured using the data-gradient norm or reconstruction-objective decrease. Compare proximal matching against ordinary clean-target MSE and a mixed loss.

(4) First experiment: adapt a small DnCNN or U-Net denoiser with 1-10 target-domain images for Gaussian deblurring on BSD68 or a shifted medical or satellite subset. Run 20-step PnP-PGD with identical $A$, noise, step sizes, and parameter count. Test ordinary MSE adaptation, proximal matching, and a 50/50 mixture. Success means higher final reconstruction PSNR and SSIM, faster decrease of $F(x_k)$, and a lower measured stationarity or data-consistency residual at equal PnP iterations, especially in the few-shot regime.

Verification

Failed on benchmark

Stage 1 · Toy mechanism gate: Passed ✓

Stage 2 · Mechanism transferred to benchmark: Not reproduced ✗

Stage 2 · Practical benchmark result: No measurable effect

Methodology: 8 paired seeds, permutation test (p<0.05) against a learning-rate-tuned baseline. How verification works

Stage 1 — Mechanism check agent confidence 9/10

Built and ran a reproducible scalar PnP proximal-mismatch experiment in experiment.py. All three quantitative predictions were confirmed: proximal coefficient error was 3.3e-16, mismatch scaling had log-log slope 2.000 versus predicted 2, zero-regularization mismatch was 0, and the stability transition bracketed the predicted boundary a=2.0 between 1.99 and 2.01. Proximal matching also beat clean-MSE in this exactly solvable quadratic toy, but this does not establish real-image performance.

Agent confidence: 9/10

Baseline: Clean-MSE coefficient=0.68604; final F=0.4410949; final stationarity=0.5887514; MSE to exact optimum=0.0581222; proximal mismatch=0.0129127
Idea: Proximal-match coefficient=0.454545; final F=0.3539115; final stationarity=2.01e-16; MSE to exact optimum=8.12e-33; proximal mismatch=9.81e-34. Mechanism checks: coefficient max error=3.33e-16, mismatch slope=2.000, predicted stability boundary=2.0, observed bracket=[1.99,2.01].

Limitations: No image denoiser, real BSD/medical data, learned nonlinear network, approximate teacher, noisy inverse operator, SSIM, GPU training, or few-shot generalization was tested. The proximal teacher was analytically known, so the practical gain is evidence only for the isolated quadratic mechanism.

How to run: python3 experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2607.14894) Sep 2, 2026

No significant difference vs. the tuned baseline (+0.6%, p=0.8863, wins 5 / 8 paired seeds).

Benchmark: Tabular regression (Friedman)
Model: mlp_tiny
Paired seeds: 8
Baseline mean: 10.9309
Idea mean: 10.996
Effect (Δ): 0.0651 (+0.6%; negative = idea better)
Wins: 5 / 8 paired seeds
p-value: 0.8863 (permutation test, 20 000 shuffles)
Smallest detectable effect: ±5.8%
Mechanism: Not confirmed ✗
Practical effect: No effect
Baseline tuning: swept over 4 configs

Limitations:

Only the registered tabular Friedman#1 track was tested. The intended image-denoising/PnP reconstruction setting, real target-domain proximal teacher, noisy inverse operator, SSIM, and few-shot image adaptation were not evaluated. The benchmark implementation used a quadratic proximal teacher and solver-like input perturbations, so transfer to the intended image reconstruction task remains untested.

How to run: python3 bench_experiment.py

Verdict computed by deterministic test code from paired-seed statistics — not by the language model.

Stage 2 — Benchmark Aug 31, 2026 latest Failed ✗

Agent confidence: 9/10

Implemented Proximal-Mismatch Fine-Tuning as a modified training loss for the registered tabular Friedman#1 track using the shared mlp_tiny architecture. The idea was slightly worse than the tuned baseline (mean test MSE 10.996044 vs 10.930936; paired delta +0.065108; permutation p=0.8863), so it did not yield a significant improvement. The mechanism signature was not confirmed.

Baseline: Mean test MSE 10.930936, std 1.398331, best configuration lr=0.006 and weight_decay=0.0
Idea: Mean test MSE 10.996044, std 1.169448, best configuration lr=0.006 and weight_decay=0.0; paired delta +0.065108; permutation p=0.8863; 5/8 paired wins

Benchmark result