{"id":"284f7f7d-9272-498a-a7e7-b8b98e5a2db8","arxiv_id":"2411.17728","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"FL-net, an autoencoder-based neural network, maps imaginary-time Green's functions to spectral functions with roughly 20% lower loss than MEM and previous networks, but larger hidden layers make it more noise-sensitive.","lead":"This paper describes FL-net, a neural network that reconstructs spectral functions from imaginary-time Green's function data by first learning a compact feature representation. It reports about 20% lower prediction error than standard analytic continuation methods and shows that larger feature spaces reduce accuracy robustness.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The robustness trade-off in Fig. 6 rests on unvalidated SVD assumptions (Appendix D) and is never checked against actual noisy inputs, so the paper's secondary claim is not established.","rationale":"The paper's primary benchmark is internally consistent for synthetic noiseless data, and the reported 20% loss improvement is plausible within that setting. The more fragile argument is the robustness claim, because it derives from an approximate SVD analysis rather than from direct testing with noisy inputs. The reader flagged the synthetic/noiseless limitation and the equal-coefficient assumption, which is related but not identical to my focus. I emphasize the validity of Eq. D5: if the weighted cross terms are non-negligible, Sj does not measure what Fig. 6 claims, independent of how realistic the benchmark is. The proposed check—comparing exact and approximate KL responses, plus quantifying the weighted off-diagonal terms—settles this directly. A conditional verdict remains appropriate, but the conditions should require validation of the robustness metric, not just additional realistic benchmarks.","tokens_in":9397,"tokens_out":6391,"duration_ms":69772,"concrete_test":"For a trained FL-net with nh ∈ {8, 23, 128, 256} and representative multi-peak spectra, numerically compute the exact second-order KL response DKL ≈ 1/2 Σi (δAi)²/Ai for many random small perturbations δG drawn as Gaussian noise in the original basis, and compare its distribution with the approximate Eq. D5 built from the SVD of M. Also compute the off-diagonal weighted terms Σi uij uik / Ai for the top modes in Fig. 6. If the approximation deviates substantially, or if off-diagonal terms are comparable in magnitude to diagonal terms, Sj is not a reliable sensitivity measure and the Fig. 6 ranking should be re-derived.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's second claim—higher hidden dimension lowers loss but reduces robustness—is supported only by the SVD sensitivity Sj (Eq. 15, Fig. 6). That derivation requires two unvalidated assumptions (Appendix D): (i) left singular vectors uj are approximately orthogonal under the 1/Ai weight so cross terms in the quadratic KL expansion vanish, even though U is orthonormal only in the Euclidean metric; and (ii) noise coefficients cj are equal across modes. For multi-peak spectra with small Ai between peaks, the weighted inner products Σi uij uik / Ai can be substantial, so Eq. D5 may not approximate Eq. D1. The paper provides no test injecting Gaussian noise into G at inference to confirm that the ordering of FL-net variants by Sj matches the actual error degradation. If those assumptions fail, Fig. 6 does not demonstrate a real robustness trade-off; it may reflect a property of the approximate metric. The primary accuracy claim on synthetic noiseless benchmarks is less affected, but the advertised robustness analysis is a central part of the paper.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes FL-net, a neural network for analytic continuation that first maps the imaginary-time Green's function G to a low-dimensional latent feature h learned by an autoencoder on spectral functions, and then decodes h to the spectral function A. The authors report that FL-net achieves lower normalized prediction loss than MEM, Fournier's network, D-net, and a fully connected network on synthetic multi-peak Gaussian and Lorentzian spectra, with the abstract claiming an improvement of at least 20%. They also introduce an SVD-based sensitivity metric S_j to argue that increasing the hidden dimension n_h reduces test loss but increases noise sensitivity. The manuscript includes implementation details, comparisons with baselines, and a code repository.","tokens_in":9669,"tokens_out":4194,"duration_ms":45796,"significance":"If the accuracy and robustness claims hold, FL-net would be a useful addition to the toolkit for analytic continuation, with the feature-learning idea being intuitive and the SVD-based sensitivity analysis a novel attempt to quantify robustness in a problem where noise sensitivity is central. The paper provides a reproducible code repository and compares against several established baselines, including MEM and the Fournier network. However, the current evidence is limited to noiseless synthetic spectra drawn from the same family as the training data, the inverse temperature is never specified, and the robustness analysis rests on unvalidated approximations, so the significance is conditional on substantial additional validation.","major_comments":[{"comment":"The inverse temperature β is never specified for the synthetic datasets, although the Matsubara frequencies ω_n=(2n+1)π/β and the kernel in Eq. (1) depend on it. Without β the benchmark is not reproducible, and it is unknown whether the reported gains persist at physically relevant temperatures. Please state the value of β used and, ideally, test at least one other β to show the method is not tuned to a particular temperature.","section":"Section II and Section III, Eq. (1)"},{"comment":"The abstract's claim of 'an improvement of at least 20%' is not quantitatively supported by the reported evidence. Fig. 4(a) shows approximately 20% loss reduction relative to the Fournier network, but with no error bars, no multiple-seed statistics, and no systematic numeric table for the Gaussian datasets; the only numeric table (Table I) is for Lorentzian spectra. Please provide mean and standard deviation over retraining seeds for MEM, FC, Fournier, D-net, and FL-net on each dataset and state the exact margins.","section":"Abstract and Section III, Fig. 4"},{"comment":"The benchmarks use only noiseless Green's functions generated from the same synthetic families used for training, yet the motivation of the paper is analytic continuation of noisy quantum Monte Carlo data. The claimed superiority over MEM and previous networks is therefore demonstrated only in an idealized setting. Please add experiments with injected noise on G at inference and, if possible, out-of-distribution spectral shapes, so that the robustness discussion is connected to actual noisy inputs.","section":"Section III and Table I"},{"comment":"The robustness analysis is built on two unvalidated assumptions: that the left singular vectors u_j are approximately orthogonal under the 1/A_i weighting so that cross terms vanish in Eq. (D5), and that the projection coefficients c_j are equal across modes. These assumptions are not checked numerically, and no experiment injects Gaussian noise into G to compare the actual KL divergence with the S_j ordering. Please quantify the weighted inner products Σ_i u_{ij} u_{ik}/A_i for the trained matrices and test whether the S_j ranking predicts the measured error degradation under realistic noise.","section":"Section IV and Appendix D, Eqs. (D1)-(D7)"},{"comment":"The claim that increasing n_h lowers loss but decreases robustness is based on averaged S_j values with no error bars and no statistical test. Fig. 6 plots curves for different n_h but does not report variability across the dataset or across retraining seeds, and the averaging procedure over modes and samples is described only qualitatively. Please provide error bars and a clear scalar summary, such as mean or maximum S_j with uncertainty, to make the trade-off quantitative.","section":"Section IV, Fig. 6"}],"minor_comments":[{"comment":"The entry 'Dent' appears to be a typo for 'D-net', and the table caption says 'Table reftab', which should be resolved to a proper reference.","section":"Table I"},{"comment":"The citation '[27] Appendix, see appendix...' is unhelpful as a reference; please point to the specific appendix sections where the encoder/decoder architectures, dataset generation, and sensitivity analysis are described.","section":"Reference [27]"},{"comment":"Equation (14) calls D_j the logarithm of the Fisher information along the u_j direction, but Fisher information is not defined in the manuscript; please add a definition or rephrase to avoid overloading the term.","section":"Section II, Eq. (14)"},{"comment":"Some panels in Fig. 5 lack axis labels or units; in particular, the horizontal axis in panel (b) is labeled only as 'mode range', which is not self-explanatory. Please add clear labels and a description in the caption.","section":"Figure 5"},{"comment":"The acknowledgment section writes 'ACKNOWLEDGEMENT' and 'Grant Nos.' in a nonstandard way; please make the wording consistent with the journal style.","section":"Section VI"}],"recommendation":"major_revision","confidential_remarks":"The core idea is promising and the comparison against MEM and previous neural networks is a useful starting point, but the missing inverse temperature, the lack of noise-injection tests, and the unvalidated SVD assumptions in Appendix D are load-bearing for the advertised robustness claim. These issues are fixable within the manuscript's scope, so major revision rather than rejection is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth a serious look. The core idea—learn a low-dimensional latent feature space from spectral functions, then map G to h and h to A—is simple and effective. On synthetic Gaussian and Lorentzian benchmarks it beats MEM, a fully connected network, and the Fournier network by a clear margin, and the loss curves in Fig. 4(a) support the main claim. The Jacobian check for the single-peak case is a nice touch, showing that the learned latent variables are locally equivalent to μ and σ. Code is public, which helps reproducibility. This is a legitimate step beyond the prior neural-network work it cites, and the subfield could plausibly adopt it if the results hold up on real QMC data.\n\nThe soft spots are real but mostly fixable. First, all benchmarks are synthetic and noiseless; the inverse temperature β is never specified, so the Matsubara grid is underdefined. That matters because noise is the whole point of analytic continuation. The abstract's \"at least 20%\" is also a bit loose—the figure shows roughly 20% against the Fournier network, not necessarily against every baseline. Second, the robustness analysis is the weakest part. The SVD sensitivity Sj in Eq. (15) depends on two assumptions in Appendix D: that the left singular vectors are approximately orthogonal under the 1/Ai weighting, and that the noise coefficients cj are equal across modes. Neither is justified numerically. The stress-test note lands: for multi-peak spectra with small Ai between peaks, the weighted cross terms can be large, so Eq. D5 may not approximate Eq. D1. The paper never checks the predicted sensitivity ordering against actual injected noise in G. Until that test is done, the claim that higher hidden dimension reduces robustness is not established.\n\nThat said, the primary accuracy claim does not rest on the robustness analysis, and it looks solid on the synthetic benchmarks. The paper would benefit from adding noisy-input experiments, stating β explicitly, and either validating the SVD metric or softening the robustness conclusion. For a serious referee, I'd send it out. It's a competent contribution to an important practical problem, and the weaknesses are addressable in revision rather than fatal. I would bring it to a reading group if your group works on analytic continuation or ML for quantum many-body problems.","headline":"A genuinely useful neural-network architecture for analytic continuation with a solid accuracy claim on synthetic benchmarks, but the advertised robustness trade-off rests on unvalidated approximations and needs a noisy-input check before it can be believed.","tokens_in":10110,"tokens_out":1097,"would_cite":false,"duration_ms":13422,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a neural network which first learns a compact latent feature representation of spectral functions and then maps imaginary-time Green's functions into that feature space reconstructs real-frequency spectra with at…","keywords":["analytic continuation","spectral function","neural network","feature learning","robustness analysis","singular value decomposition","ill-posed inverse problem","maximum entropy method"],"falsifier":"Train FL-net and the comparison methods on one of the paper's synthetic Gaussian datasets, then evaluate on spectra with substantially different line shapes (e.g., asymmetric or strongly overlapping peaks) or on Green's functions with added noise of known variance; if the loss advantage over MEM and the previous neural network falls below 20% or reverses, the central claim fails outside the training distribution.","tokens_in":9188,"feed_emoji":"🧠","tokens_out":5387,"duration_ms":93985,"temperature":0.7,"pith_summary":"Analytic continuation, recovering a real-frequency spectral function from imaginary-time data, is an ill-posed inverse problem where standard methods like the maximum entropy method over-smooth sharp or multi-peak features. This paper proposes FL-net, which trains an autoencoder on spectral functions to find a low-dimensional latent feature vector, then trains a network to map the Green's function into that latent space before decoding back to the spectrum. On synthetic Gaussian and Lorentzian multi-peak datasets, FL-net attains roughly 20% lower prediction loss than MEM, a previously published neural network, and a direct-mapping baseline with comparable parameters. The paper also derives an SVD-based sensitivity measure showing that increasing the latent dimension lowers loss but makes the network more sensitive to input noise.","feed_headline":"Feature-learning net cuts analytic continuation loss by 20%","feed_subtitle":"It maps imaginary-time data through a learned latent space, beating MEM and earlier networks on multi-peak spectra.","key_machinery":"FL-net is a two-encoder, one-decoder architecture: an autoencoder trained on spectral functions defines a latent feature vector h; a second encoder maps the imaginary-time Green's function G(iωn) to h, and a decoder maps h back to the spectral function A(Ω). The robustness machinery is the SVD of the Jacobian M = ∂A/∂G, whose singular values τj and left singular vectors uj yield Dj = ln(Σ_i u_{ij}^2/A_i), the log of the inverse Fisher information along mode j, and Sj = 2 ln τj + Dj, a mode-wise sensitivity that lets the authors compare networks of different hidden dimension.","core_discovery":"The paper's central claim is that the mapping from imaginary-time Green's functions to spectral functions becomes both more accurate and more interpretable when it is factored through a learned latent feature space rather than performed directly. For a single Gaussian peak, FL-net's two latent variables are numerically equivalent to the peak's mean and variance, verified through the Jacobian of the encoder. For multi-peak spectra, the latent dimension that minimizes test loss is around three times the peak number minus one, the number of independent parameters in the synthetic data. The robustness analysis defines a per-mode noise sensitivity Sj = 2 ln τj + Dj from the singular value decomposition of the input–output Jacobian, and shows that larger hidden dimensions produce larger Sj, i.e., higher noise sensitivity, while lowering the normalized prediction loss.","pith_inferences":["The latent-feature factorization can be read as learning a data-driven default model, analogous to MEM's prior but fitted from the training set; a natural test is whether its advantage persists on spectra with line shapes absent from training.","The SVD sensitivity analysis suggests a practical regularizer: truncating or penalizing high-Sj modes of the Jacobian could improve robustness without retraining, though the paper does not propose this.","The paper's synthetic benchmarks use noiseless Green's functions; whether the 20% advantage holds for noisy quantum Monte Carlo data with unknown inverse temperature is an open empirical question the paper leaves implicit.","The architecture's principle, learn features on the output distribution and then map inputs to those features, could transfer to other ill-posed inverse problems beyond analytic continuation."],"forward_implications":["FL-net reproduces sharp and multi-peak spectra that MEM smooths out, improving loss by roughly 20% on the tested synthetic families.","The learned latent features provide a compact description of the spectrum; for single-peak Gaussians they recover exactly the mean and variance.","The SVD sensitivity Sj offers a quantitative robustness metric that can be reported alongside prediction loss for any trained continuation network.","The findings imply a tradeoff: raising hidden dimension improves accuracy but increases noise sensitivity, so model selection should balance both.","The method extends to Lorentzian spectra, where it also achieves the lowest loss among the methods compared."],"supporting_citations":[{"why":"Supplies the previously published neural network baseline that FL-net is compared against and outperforms by roughly 20%.","marker":"[22]"},{"why":"Provides the maximum entropy method references that define the traditional baseline FL-net improves on, especially for sharp and multi-peak spectra.","marker":"[17–21]"},{"why":"Adam optimizer used for training all networks in the comparison, placing the performance differences on a common optimization base.","marker":"[28]"}],"fun_headline_variants":["FL-net: 20% better spectral recovery via feature learning","Neural net learns spectral features, beats MEM by 20%","Feature-learning network improves spectral reconstruction by 20%","20% better spectral functions via learned features","Feature learning: 20% better spectra, but robustness drops"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported accuracy gain assumes the test spectra come from the same synthetic families used for training, sums of Gaussian or Lorentzian peaks generated without input noise, so the 20% improvement and the robustness tradeoff are established only for those families.","fun_headline_variants_meta":{"raw":{"variants":["FL-net: 20% better spectral recovery via feature learning","Neural net learns spectral features, beats MEM by 20%","Feature-learning network improves spectral reconstruction by 20%","20% better spectral functions via learned features","Feature learning: 20% better spectra, but robustness drops"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000921,"raw_usage":{"total_tokens":3883,"prompt_tokens":811,"completion_tokens":3072,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":427,"completion_tokens_details":{"reasoning_tokens":2991}},"tokens_in":427,"tokens_out":3072,"duration_ms":23553,"temperature":1.0,"reasoning_tokens":2991,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:57:35.420184+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Train FL-net and the comparison methods on one of the paper's synthetic Gaussian datasets, then evaluate on spectra with substantially different line shapes (e.g., asymmetric or strongly overlapping peaks) or on Green's functions with added noise of known variance; if the loss advantage over MEM and the previous neural network falls below 20% or reverses, the central claim fails outside the training distribution.","supporting_citations":[{"cited_title":"Fournier, L","cited_arxiv_id":null,"evidence_quote":"Supplies the previously published neural network baseline that FL-net is compared against and outperforms by roughly 20%."}],"review_version":1}