{"id":"693debd1-e9be-4e1b-9b68-c4bdba0a9e2d","arxiv_id":"2607.09546","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"Graph-regularized RTRMC (GR-RTRMC) recovers missing matrix entries more accurately than RTRMC and several baselines by enforcing smoothness on row and column graphs via variable projection on the Grassmann manifold.","lead":"The paper adds graph regularization to Riemannian trust-region matrix completion so missing entries can be filled using known row and column relationships. This matters for weather stations and recommender systems where nearby or similar items should behave alike.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified beyond the reader's already-flagged graph-mismatch failure mode.","rationale":"The paper is a clean, correctly derived engineering extension of RTRMC. The variable-projection reduction, the extra Kronecker term λ_w(L_w⊗I_r) inside A, and the projected Hessian are algebraically consistent with the stated objective (11). Empirical gains are modest but consistent across four matrices, code is released, and the authors themselves document the failure modes that arise when spatial/temporal coherence is violated. The reader’s weakest_assumption already isolates the precise modeling risk; nothing more load-bearing appears on a second pass. Hence the CONDITIONAL verdict and the identified soft spot stand unchanged.","tokens_in":15940,"tokens_out":482,"duration_ms":6070,"concrete_test":"Re-run the Belgian Block experiment of Table 1 after replacing the fixed spatial Laplacian L_u by a time-varying graph that zeros edges between stations whose precipitation differs by more than a threshold during the storm window of Fig. 6; if the reported 0.45 °C RMSE rises above RTRMC’s 0.49 °C, the graph-mismatch concern is confirmed as performance-critical rather than merely anecdotal.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (graph-regularized RTRMC via variable projection yields lower test RMSE than RTRMC/GRALS/SoftImpute on the reported weather and MovieLens matrices) is supported by the supplied Riemannian gradient/Hessian (Eqs. 24–30, 43), the closed-form W_U solve that correctly absorbs the column Laplacian (Eq. 35), and the tabulated gains (Tables 1–3). The only place the claim is fragile is already identified by the reader: when the chosen graphs (k-NN haversine+altitude for stations; k-NN rating vectors for users/movies) do not encode the latent smoothness that actually holds, the Dirichlet terms become harmful and produce artifacts (§3.5 storm/rain cases) unless C is manually down-weighted. That is a genuine limitation of the modeling assumption, not an internal inconsistency or derivation error. No stronger load-bearing flaw (incorrect derivatives, non-reproducible numbers, or hidden circularity) is visible.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper extends Riemannian Trust-Region Matrix Completion (RTRMC) by adding Dirichlet graph-regularization terms on the row and column factors, yielding GR-RTRMC. After variable projection of the right factor W_U, the problem is optimized over the Grassmannian Gr(m,r). The authors derive the reduced cost, the Euclidean and Riemannian gradients (Eqs. 24–25), the Riemannian Hessian (Eq. 30), and the closed-form linear system for W_U that correctly absorbs the column Laplacian via the Kronecker term λ_w(L_w ⊗ I_r) (Eq. 35). Experiments on Belgian and French weather matrices (Block and Spread missingness) and on MovieLens 100K/1M report lower average test RMSE than RTRMC, GRALS, SoftImpute and other baselines (Tables 1–3). Two case studies (§3.5) document failure modes under localized storms and rain when the spatial graph assumption is violated.","tokens_in":16208,"tokens_out":1033,"duration_ms":10876,"significance":"If the reported gains hold under independent re-implementation, the work supplies a clean, usable extension of a well-known Riemannian matrix-completion method that incorporates side-information graphs while retaining second-order trust-region geometry. The explicit gradient/Hessian formulae and the Kronecker-structured W_U solve are concrete technical contributions that practitioners can implement. Code is stated to be available. The weather and MovieLens results, together with the honest storm/rain failure analysis, make the paper useful both as a method paper and as a cautionary study of when graph smoothness helps or hurts. The contribution is incremental rather than foundational, but it is solid and of clear interest to the matrix-completion and Riemannian-optimization communities.","major_comments":[{"comment":"§3.5 (storm and rain cases) shows that when the chosen graphs do not encode the true latent smoothness, the Dirichlet terms produce artifacts unless the confidence matrix C is manually lowered over the anomalous window. The abstract and introduction claim improved “robustness,” yet the only mitigation demonstrated is this hand-tuned down-weighting of C. A load-bearing claim of the paper is therefore only partially supported: either a quantitative sensitivity study (varying graph construction / λ_u,λ_w and reporting when GR-RTRMC degrades relative to plain RTRMC) or an automatic anomaly-aware C schedule should be added so that the robustness claim can be evaluated rather than asserted.","section":null},{"comment":"Tables 1–3 report only average RMSE (and wall-clock time). No standard deviations across folds, no statistical significance tests, and no ablation that isolates the contribution of the row graph versus the column graph versus the Riemannian geometry are provided. Because the central empirical claim is that GR-RTRMC “yields lower test RMSE,” the absence of variability measures and of a controlled ablation makes it impossible to judge whether the observed gains are reliable or driven by a particular hyper-parameter / fold combination. Adding fold-wise statistics and a minimal ablation would substantially strengthen the claim.","section":null}],"minor_comments":[{"comment":"Notation for the unobserved set is inconsistent: the text after Eq. (11) writes Ω for the unobserved entries while earlier Ω denotes the observed set; a bar or complement symbol would avoid confusion.","section":null},{"comment":"In the vectorized derivation (around Eqs. 32–36) the symbols λ_u^{2} / λ_w^{2} appear in places where the original objective uses λ_u / λ_w; a short consistency check would help readers re-implement the linear system.","section":null},{"comment":"Figure 2 captions list station IDs but the corresponding geographic locations are not indicated on the map of Fig. 1; a small inset or legend would improve readability.","section":null},{"comment":"The computational-complexity discussion (§3.4) is only empirical; a short big-O statement for the dominant cost of forming/solving the Kronecker-augmented system for W_U would be useful.","section":null},{"comment":"A few typographical issues remain (e.g., “M´ et´ eo-France”, missing spaces around some equation references). A careful proof-reading pass is recommended.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is a natural and carefully executed extension of RTRMC; novelty relative to the authors’ own BNAIC/BeNeLearn 2024 abstract and EUSIPCO 2023 graph-based weather paper is incremental but sufficient for a full journal version. Fit for a methods-oriented ML / optimization journal is good. I see no integrity concerns; the storm/rain analysis is refreshingly candid."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is a clean engineering paper that does exactly what the title promises: it folds row- and column-graph Dirichlet penalties into the RTRMC variable-projection setup and derives the new gradient, Hessian, and non-block-diagonal W_U system that result. The math is standard manifold calculus done carefully (eqs. 24–30, 35, 43); the Kronecker term from L_w is handled correctly. Code is released. That is the real contribution.\n\nWhat is new is modest but concrete. Boumal–Absil already gave the Grassmann geometry and variable projection; GRALS and related work already used graph smoothness. The authors supply the previously unpublished closed forms that appear once both ingredients sit inside the same reduced cost, then show that the resulting method beats RTRMC, GRALS, SoftImpute and the usual baselines on Belgian and French temperature matrices and on MovieLens 100K/1M under k-fold CV. Tables 1–3 are consistent; the gains are small but real (roughly 0.04–0.05 °C or a few hundredths of RMSE). The case studies in §3.5 are honest: when a convective front or localized rain violates the spatial graph, the method produces artifacts unless C is manually lowered. That is a modeling limitation, not a derivation error, and they flag it themselves.\n\nSoft spots are proportional. Hyper-parameters (rank, three λ’s, graph k and similarity) are free and chosen by validation—standard, but the paper does not report sensitivity or error bars. Timing is 2–3× slower than plain RTRMC, as expected once the column Laplacian couples the columns. Novelty is incremental; significance is subfield-level for people who already do low-rank + graph side information for weather or collaborative filtering.\n\nThe paper is for practitioners and Riemannian-optimization people who want a ready-to-use, graph-aware RTRMC variant with explicit derivatives. It is formally grounded enough and empirically sharp enough to deserve a serious referee. I would accept it for peer review and would bring it to a reading group if we are currently working on matrix completion or weather imputation. I would cite the derivative expressions and the weather numbers if I were building on the same stack.","headline":"Solid, correctly derived engineering extension of RTRMC with graph Dirichlet terms; real but incremental gains on weather and MovieLens, with the expected graph-mismatch failure modes already documented by the authors.","tokens_in":16771,"tokens_out":551,"would_cite":true,"duration_ms":7232,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15A83","65F55","90C35"],"pacs":[],"model":"grok-4.5","headline":"Adding graph smoothness to Riemannian low-rank matrix completion improves recovery of missing weather and ratings data.","keywords":["low-rank matrix completion","graph regularization","Riemannian optimization","Grassmann manifold","variable projection","weather data imputation","collaborative filtering"],"falsifier":"On a held-out weather or rating matrix whose true row/column graph is known, replace that graph by a random or deliberately misaligned graph of equal density and check whether GR-RTRMC's RMSE advantage over plain RTRMC disappears or reverses.","tokens_in":16877,"feed_emoji":"🌦️","tokens_out":830,"duration_ms":8872,"temperature":0.7,"pith_summary":"This paper claims that low-rank matrix completion becomes more accurate when the factors are forced to vary smoothly along known graphs linking rows and columns. The authors take an existing Riemannian trust-region method that optimizes only over subspaces on the Grassmann manifold and add Dirichlet penalties that pull nearby stations or similar users toward similar latent factors. On temperature matrices from Belgian and French weather stations and on MovieLens rating matrices, the resulting algorithm (GR-RTRMC) reports lower test error than plain RTRMC, graph-regularized alternating least squares, SoftImpute and several other baselines. The practical payoff is better imputation of missing weather readings and movie ratings whenever spatial, temporal or preference graphs are available. The same case studies also show where the assumption breaks: highly local storms produce artifacts unless the confidence weights on those observations are manually reduced.","feed_headline":"Graph smoothness cuts missing-weather and rating errors","feed_subtitle":"Riemannian matrix completion plus row/column graphs beats standard baselines on stations and MovieLens.","key_machinery":"Variable projection of the graph-regularized least-squares objective onto the Grassmann manifold: for each candidate subspace U the optimal factor W is obtained in closed form by solving a linear system that now includes the column-graph Laplacian Kronecker product, after which Riemannian trust-region steps are taken on the reduced cost f(U).","core_discovery":"Incorporating Dirichlet graph-regularization terms into the RTRMC variable-projection formulation yields lower test RMSE than RTRMC, GRALS, SoftImpute and other baselines on Belgian and French weather matrices and on MovieLens 100K/1M, provided the chosen graphs correctly encode the latent smoothness that should be enforced.","pith_inferences":["Automatic detection of confidence-weight drops during anomalous periods would remove the need for the manual storm-window intervention shown in the case study.","Feeding additional meteorological variables (precipitation, radiation) as extra graph edges or multi-layer factors could capture the localized convective events that currently break pure temperature completion.","The same Grassmann-plus-graph construction is a natural candidate for other spatiotemporal imputation tasks such as traffic, air-quality or power-grid matrices."],"forward_implications":["Weather services can impute missing station readings more accurately by supplying a spatial station graph and a temporal smoothness graph.","Collaborative-filtering systems can lower RMSE by feeding user-user and item-item similarity graphs into the same Riemannian solver.","When an anomalous window is known a priori, simply lowering the corresponding confidence weights restores plausible imputations without changing the algorithm.","The same variable-projection Grassmann scheme can be reused for any other matrix-completion task that supplies side-information graphs."],"fun_headline_variants":["Graph-regularized RTRMC lowers RMSE on weather and MovieLens","Dirichlet graphs lift RTRMC accuracy for stations and ratings","Row-column smoothness improves Riemannian matrix completion","GR-RTRMC beats baselines when graphs encode latent structure","Variable projection plus graphs cuts matrix completion errors"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The chosen row and column graphs correctly encode the smoothness that latent factors ought to obey; when local events violate that smoothness the method produces artifacts unless observation confidence is manually lowered.","fun_headline_variants_meta":{"raw":{"variants":["Graph-regularized RTRMC lowers RMSE on weather and MovieLens","Dirichlet graphs lift RTRMC accuracy for stations and ratings","Row-column smoothness improves Riemannian matrix completion","GR-RTRMC beats baselines when graphs encode latent structure","Variable projection plus graphs cuts matrix completion errors"]},"model":"grok-4.5","effort":"low","cost_usd":0.005752,"raw_usage":{"total_tokens":1440,"prompt_tokens":626,"num_sources_used":0,"completion_tokens":63,"cost_in_usd_ticks":57520000,"prompt_tokens_details":{"text_tokens":626,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":751,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":626,"tokens_out":63,"duration_ms":7314,"temperature":1.0,"reasoning_tokens":751,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T02:14:10.127534+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"On a held-out weather or rating matrix whose true row/column graph is known, replace that graph by a random or deliberately misaligned graph of equal density and check whether GR-RTRMC's RMSE advantage over plain RTRMC disappears or reverses.","supporting_citations":[],"review_version":1}