{"id":"03ad60f6-005a-4e7a-9951-b8523be40b23","arxiv_id":"2403.20200","paper_version":5,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Derives deterministic equivalents for ridge regression predictive risk under a variance profile in high-dimensional non-i.i.d. data, showing double descent for some profiles and different shapes for others.","lead":"This paper derives deterministic equivalents for the predictive risk and degrees of freedom of ridge regression when predictors have a variance profile instead of being i.i.d. A smart generalist might read it to see how non-uniform variance changes the shape of double descent in high-dimensional regression.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"The reader's weakest-assumption note already flags the RMT technical conditions; the abstract gives no indication that those conditions are violated or that the derivation contains an unstated step that would invalidate the equivalents. Because the full proofs are not reproduced here, an honest non-finding is the appropriate posture rather than manufacturing a concern about uninspectable algebra.","tokens_in":1750,"tokens_out":298,"duration_ms":10468,"concrete_test":"Re-derive the deterministic equivalent for the predictive risk (the expression obtained after applying the variance-profile RMT resolvent) for the special case of a constant (i.i.d.) profile and verify that it reduces exactly to the known Marchenko-Pastur ridge formula; any discrepancy would indicate a gap in the application of the new RMT tools.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is the derivation of deterministic equivalents for ridge risk and degrees of freedom under a random-effects model with variance-profile covariates, using RMT tools not previously applied to regression. The abstract states that the profile satisfies the proportional-growth and moment conditions required by those RMT results, and that double descent appears for a subclass of profiles while other profiles yield qualitatively different risk curves. No internal inconsistency, hidden circularity, or unsupported step is visible in the stated claims; the work is explicitly conditional on the technical hypotheses of the underlying RMT machinery.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript derives deterministic equivalents for the predictive risk and the degrees of freedom of the ridge estimator in high-dimensional linear regression under a random-effects model, where the design matrix has independent but non-identically distributed entries governed by a variance profile whose dimensions grow proportionally. It shows that, for certain classes of variance profiles satisfying the requisite technical conditions, the minimum-norm least-squares estimator exhibits the double-descent phenomenon as the ridge parameter tends to zero, while other profiles produce qualitatively different risk curves. The derivations rely on random-matrix-theory results for variance profiles that have not previously been applied to regression; numerical experiments are provided to illustrate the accuracy of the equivalents.","tokens_in":1838,"tokens_out":363,"duration_ms":14318,"significance":"If the derivations hold under the stated conditions, the work extends the RMT analysis of ridge regression from the iid setting to a substantially more general class of heterogeneous data. The explicit dependence of risk shape on the variance profile, including both double-descent and non-double-descent regimes, supplies a concrete mechanism for understanding generalization behavior beyond the classical iid case. The application of previously unused RMT tools to regression and the provision of numerical validation are clear strengths.","major_comments":[],"minor_comments":[{"comment":"Abstract: the phrase 'for certain class of variance profile' is imprecise; the introduction or Section 2 should explicitly reference the precise technical conditions (proportional growth, moment bounds) under which double descent is recovered.","section":"Abstract"},{"comment":"The manuscript would benefit from a short table or figure in the main text that contrasts the risk curves for at least two concrete variance-profile families (one yielding double descent, one not) rather than relegating all examples to the numerical section.","section":"Introduction"}],"recommendation":"minor_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the careful reading and positive assessment of the manuscript, including the accurate summary of the contributions and the recommendation for minor revision. No specific major comments were provided in the report.","responses":[],"tokens_in":1287,"tokens_out":58,"duration_ms":5509,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper's main result is a set of deterministic equivalents for the predictive risk of ridge regression and its degrees of freedom when the features have a variance profile. Under a random effects model and proportional dimension growth, the equivalents let the authors track how the risk behaves as a function of the regularization parameter and highlight that double descent appears for some profiles but not others when the ridge parameter goes to zero. The proofs rely on random matrix theory tools for variance profiles that the abstract says had not been applied to regression before. Numerical experiments are included to check that the equivalents track the finite-sample risk closely. This is a clean extension of the i.i.d. ridge analysis that has been standard in the literature. The technical conditions on the profile (moment bounds and growth rates) are the usual ones from the RMT side, so the scope is narrower than fully general non-i.i.d. data but still broader than the i.i.d. case. No circularity or internal inconsistency appears in the claims. The random effects assumption is standard for this type of risk calculation and keeps the derivations tractable. The work is aimed at researchers in high-dimensional statistics who want explicit risk formulas beyond i.i.d. assumptions or who study how covariate heterogeneity changes phenomena like double descent. A reader already comfortable with RMT in statistics will get the most from the explicit expressions and the profile comparisons. The combination of new application, explicit results, and numerical checks is enough to justify sending it to a serious referee rather than desk rejecting it.","headline":"The paper derives deterministic equivalents for ridge risk and degrees of freedom under a variance-profile covariate model and shows the risk curve shape depends on the profile.","tokens_in":2307,"tokens_out":374,"would_cite":true,"duration_ms":17205,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[],"headline":"Standard RMT analysis of ridge regression under heteroscedasticity; no RS-shaped structure","alignment":"orthogonal","rationale":"The paper derives deterministic equivalents for ridge risk/DOF via resolvent analysis, Dyson equations (HLN07/AEK17b), and Marchenko-Pastur limits for quasi-doubly-stochastic profiles, showing profile-dependent risk curves (double/triple descent). This is classical high-dimensional statistics with no ratio-symmetric cost J, golden-ratio identities, 8-tick periodicity, or parameter-free constant derivations. RS theorems (reality_from_one_distinction, Jcost uniqueness, phi-ladder constants) are absent and have no bearing on the statistical setting.","tokens_in":63530,"confidence":"high","tokens_out":160,"duration_ms":5766,"cache_read_input_tokens":38528,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Ridge regression on data with a variance profile admits deterministic equivalents for predictive risk and degrees of freedom.","keywords":["high-dimensional regression","ridge estimator","variance profile","deterministic equivalents","double descent","random matrix theory","predictive risk","non-identically distributed data"],"falsifier":"A direct numerical comparison in which the empirical risk of ridge regression on simulated data with a qualifying variance profile deviates from the deterministic equivalent by more than sampling error.","tokens_in":2650,"feed_emoji":"📈","tokens_out":579,"duration_ms":31583,"temperature":0.7,"pith_summary":"The paper studies high-dimensional linear regression where predictors follow a variance profile rather than being identically distributed. Under a random effects model it derives deterministic equivalents for the ridge estimator's predictive risk and its degrees of freedom. These equivalents are obtained via random matrix theory adapted to variance profiles. For some profiles the risk of the minimum-norm least-squares estimator exhibits double descent as the regularization parameter tends to zero; for other profiles the risk curve takes a different shape.","feed_headline":"Ridge risk on variance-profile data gets exact high-dim formulas","feed_subtitle":"Deterministic equivalents cover predictive risk and degrees of freedom; double descent emerges for some profiles but not others","key_machinery":"The variance profile of the random predictor matrix, analyzed through random matrix theory results that handle non-identical variances.","core_discovery":"Assuming a random effect model, the predictive risk of the ridge estimator and its degrees of freedom admit deterministic equivalents when the data matrix has a variance profile and dimensions grow proportionally. For certain classes of variance profiles, the minimum norm least-squares estimator (ridge parameter to zero) shows double descent in the predictive risk, while other profiles yield different risk shapes.","pith_inferences":["The formulas could be inverted to choose the ridge parameter that minimizes risk for a given estimated variance profile.","Similar deterministic equivalents might be derived for generalized linear models or kernel ridge regression under the same variance-profile assumption.","Real-data applications would require consistent estimation of the variance profile entries from the observed matrix."],"forward_implications":["The deterministic equivalents allow exact high-dimensional computation of ridge risk without Monte Carlo simulation.","Double descent appears in the minimum-norm estimator for some non-iid variance profiles.","Certain variance profiles produce predictive-risk curves that do not follow the double-descent shape.","The same random-matrix machinery can be applied to other linear estimators beyond ridge."],"fun_headline_variants":["Ridge risk equivalents derived for variance profile data in high dim","Deterministic equivalents for ridge risk and df with variance profiles","Ridge regression double descent depends on variance profile shape","High dim ridge gains exact risk formulas via variance profile analysis"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The observations follow a random effects model and the variance profile satisfies the moment and growth conditions required for the random matrix theory tools.","fun_headline_variants_meta":{"raw":{"variants":["Ridge risk equivalents derived for variance profile data in high dim","Deterministic equivalents for ridge risk and df with variance profiles","Ridge regression double descent depends on variance profile shape","High dim ridge gains exact risk formulas via variance profile analysis"]},"model":"grok-4.3","cost_usd":0.008765,"raw_usage":{"total_tokens":3948,"prompt_tokens":669,"num_sources_used":0,"completion_tokens":65,"cost_in_usd_ticks":87649500,"prompt_tokens_details":{"text_tokens":669,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3214,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":669,"tokens_out":65,"duration_ms":19255,"temperature":1.0,"reasoning_tokens":3214,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-24T02:44:38.723271+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A direct numerical comparison in which the empirical risk of ridge regression on simulated data with a qualifying variance profile deviates from the deterministic equivalent by more than sampling error.","supporting_citations":[],"review_version":1}