{"id":"f40c4993-1eb3-4221-b659-68a5e73dd16c","arxiv_id":"2507.02732","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"An entropy-minimizing classifier learns a separating hyperplane (or polynomial surface) together with per-point positive distances, and is tested against standard baselines.","lead":"This paper introduces a classifier that finds a separating hyperplane by minimizing an entropy-like function over the plane's orientation and per-point positive distances, and extends the same idea to polynomial decision surfaces. It reports accuracy comparisons against SVM, KNN, logistic regression, and perceptron on four synthetic datasets and one real-world dataset.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The method's core theorem only covers strictly separable data; on nonseparable datasets like Circles the dual objective has no true minimizer, so the reported Entropic-Linear result is outside the theorem's scope and may be a degenerate artifact.","rationale":"The reader's weakest-assumption analysis already identifies the infeasible/nonseparable case as the central gap, and my reading agrees: the convex-duality derivation is sound under a nonempty feasible set, but the paper applies it to datasets where no separating hyperplane exists. The concrete failure mode is sharper than just 'the theorem does not apply': for infeasible data the dual minimization can have an unattained infimum along a direction where the gradient decays to zero, so the numerical stopping rule is satisfied by a degenerate solution with w≈0 and b≈0. This is exactly what the Circles Entropic-Linear result suggests. I do not see an internal inconsistency that invalidates Theorem 7 for feasible inputs, nor evidence that the polynomial experiments (which are feasible for the chosen degrees on Circles, Spirals, and Moons) are fabricated. The missing feasibility check, absent baselines, and lack of error bars are all addressable and do not overturn the conditional mathematical claim. Hence the reader's CONDITIONAL verdict remains appropriate; the manuscript should be revised to state the feasibility hypothesis, add a feasibility test before running the method, and either omit or clearly interpret degenerate results on nonseparable linear problems.","tokens_in":8175,"tokens_out":10555,"duration_ms":144323,"concrete_test":"Run the released code on the Circles dataset with Entropic-Linear, logging λ, w, b, M(A^tλ), and the gradient norm at each iteration. If ||λ|| grows without bound, w→0, and the Section 4.1 criterion triggers anyway, the experiment is outside Theorem 7. Independently, solve the feasibility LP 'exists w∈[-1,1]^2 with Dw>0' for the Circles training points; infeasibility would directly falsify the theorem's premise for that experiment and require removing or relabeling the Entropic-Linear result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 7 solves the equality-constrained problem min{Ψ(x) | x∈K, Ax=0}. With A=[D,-I], any feasible x must satisfy b=Dw with b>0, i.e. the data must be strictly linearly separable by some w∈[-1,1]^N. The paper never checks this feasibility condition before invoking the theorem. On the Circles dataset no such w exists, so the primal feasible set is empty. In that case the dual objective g(λ)=M(A^tλ)=Σ_j ln(2cosh((D^tλ)_j))+Σ_i e^{-λ_i} is bounded below but its infimum is generically not attained: because the two disks overlap, there is a positive vector v with D^t v=0, and along λ=t v→∞ one has g(λ)→M ln2 and ∇g(λ)=A x(λ)→0, with x(λ)→(0,0). Thus the spectral-gradient stopping criterion of Section 4.1 is satisfied by an unbounded λ sequence producing an essentially zero weight vector, not by a separating hyperplane. Consequently the Entropic-Linear row of Table 3 (accuracy 0.46) is not a valid instance of Theorem 7; it is an artifact of optimizing a dual problem that has no minimizer. The same caveat applies to any nonseparable dataset on which Entropic-Linear is run. The theorem itself is conditionally correct, but the claimed scope of the method and the experimental protocol need to explicitly restrict to, or test for, feasibility.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a classification method based on minimizing a strictly convex, entropy-like function Psi over a box containing the weight vector and a positive orthant containing per-point margins, subject to equality constraints Ax=0. In the linear case the constraints require every training point to satisfy an exact equality with a positive per-point margin; the polynomial extension maps the data through monomial features and places the coefficient vector in [-e,e]^{T(p)}. Section 3 introduces the Legendre-Fenchel dual M, and Section 4 derives closed-form expressions for the primal variables in terms of a dual vector lambda, culminating in Theorem 7, which states that the solution is obtained by minimizing lambda |-> M(A^t lambda). Section 5 compares the resulting Entropic-Linear and Entropic-Polynomial classifiers against SVM, KNN, logistic regression, and perceptron on four synthetic datasets and on the Breast Cancer dataset.","tokens_in":8513,"tokens_out":10234,"duration_ms":128995,"significance":"If Theorem 7 holds under the appropriate hypotheses, the paper offers an attractive convex-duality formulation: minimizing Psi subject to affine equality constraints is equivalent to minimizing a smooth convex function of the dual variable, with simple gradient formulas. The availability of reproducible code and the inclusion of several standard baselines are positive features. However, the central theorem is conditional on the feasible set {x in K : Ax=0} being nonempty, a condition that is never stated, checked, or enforced; the experiments include datasets for which that set is empty, and the reported degenerate result for Entropic-Linear on Circles is consistent with an unbounded dual sequence rather than a separating solution. The polynomial extension also contains an unresolved scaling gap between the coefficient box [-e,e] and the entropy function derived for [-1,1]. As a result, the paper currently establishes a conditional theoretical result plus an illustrative implementation, but not the broad empirical and methodological claim made in the abstract.","major_comments":[{"comment":"The theorem presupposes that the feasible set {x in K : Ax=0} is nonempty, but the paper never states or tests this condition. With A=[D,-I], feasibility is equivalent to b=Dw with b_i>0 and w in [-1,1]^N, i.e. strict linear separability by a hyperplane through the origin. For the Circles dataset in Section 5.3.2 this condition fails, so the Entropic-Linear row of Table 3 (accuracy 0.46) is not a valid instance of Theorem 7. The degenerate confusion matrix, with all test points assigned to one class, is what one expects from an unbounded dual sequence lambda=t v with v>0 and D^t v=0, along which grad_lambda M(A^t lambda)=A x(lambda) tends to 0 while x(lambda) tends to (0,0); the stopping criterion in Section 4.1 can therefore be satisfied without approaching a separating solution. The paper should add a feasibility check before applying the method, or explicitly restrict all claims and experiments to data satisfying Problem 3's equality constraint.","section":"Section 4, Theorem 7 and Eq. (6)"},{"comment":"The polynomial formulation places coefficients in the box [-e,e]^{T(p)}, but the entropy function Psi in Eq. (12) and the resulting dual formulas in Eqs. (17) and (20)-(21) are derived only for the box [-1,1]^N. The paper does not provide the rescaled entropy or the corresponding dual formulas (after a change of variables one would expect expressions containing e, such as w_j = e tanh(e tau_j)). Unless the implementation silently sets e=1, the Entropic-Polynomial classifier is not justified by the theory in Sections 3 and 4; if it does set e=1, then e is not an adjustable parameter as claimed. This gap needs to be fixed by generalizing the formulas or by explicitly normalizing the coefficients.","section":"Section 2, Problem 6 and Eq. (12)"},{"comment":"All numerical comparisons are based on a single 70:30 train-test split, with no standard errors, confidence intervals, or repeated-seed evaluation. Given that the behavior of Entropic-Linear on nonseparable data is the central empirical question raised by Theorem 7's feasibility condition, single-split accuracy values cannot support the paper's comparative claims. In addition, the polynomial degree is selected by grid search on the same split used for reporting accuracy, which can optimistically bias the reported results. Please report repeated splits or cross-validation, and include a feasibility diagnosis for each dataset.","section":"Section 5.2 and Tables 3-10"}],"minor_comments":[{"comment":"The expression 'G S D' should read 'G union D'; the union symbol appears to have been lost in typesetting.","section":"Section 1, Problem 2"},{"comment":"The symbol M is overloaded: M denotes the number of samples K1+K2 in Eq. (5), and also denotes the dual function M(tau,zeta) in Eq. (13). This makes Section 4 harder to read; consider using a different symbol, such as m, for the sample size.","section":"Eq. (5) and Eq. (13)"},{"comment":"The relationship to reference [8] is not made precise. If the dual minimizer coincides with the maxent-in-the-mean solution of [8], the paper should explicitly state the novel contribution, such as the dual optimization algorithm or the polynomial extension, so that the reader can assess the advance over that prior work.","section":"Introduction and reference [8]"},{"comment":"Theorem 7 is stated without the convex-analysis hypotheses needed to guarantee that the dual minimizer exists and that strong duality holds, such as feasibility together with a Slater-type condition. A short remark stating these hypotheses would make the theorem rigorous rather than a formal first-order derivation.","section":"Theorem 7"}],"recommendation":"major_revision","confidential_remarks":"The editor may wish to ask the authors for a detailed comparison with reference [8] during revision. As written, the paper cites [8] but does not identify the mathematical difference between the earlier maxent-in-the-mean solution and the present dual minimization, and one author is a co-author of both works. The degenerate Entropic-Linear result on the Circles dataset should also be investigated with the provided code, since it appears to be a direct consequence of running the algorithm outside the theorem's feasibility regime."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know: the convex-duality derivation in Section 4 is correct for the feasible case, and the method's uncertainty band is a nice touch. But the linear classifier is the maxent-in-the-mean method from reference [8], dressed in new notation, and the experiments on nonseparable data operate outside the theorem's scope because the equality constraint Ax=0 requires strict homogeneous separability.\n\nWhat is actually new: the polynomial extension (feature expansion followed by the same dual minimization) and the explicit uncertainty band between b+ and b-. The derivation of Theorem 7 is sound when a feasible x exists. The dual objective λ↦M(A^tλ) is convex, and the gradient stopping rule is a sensible way to approach the constraint for separable data.\n\nWhere it gets soft. First, the novelty claim: equations (15)-(16) identify Ψ as the Fenchel dual of the log-Laplace transform M, and the solution x*=∇M(A^tλ*) is exactly the maxentropic mean used in [8]. The paper cites [8] but never says the linear algorithm is equivalent to it. That is a disclosure problem.\n\nSecond, and more load-bearing: Theorem 7 assumes Ax=0 is satisfiable. For a linear classifier with K=[-1,1]^N×R_{++}^M, feasibility means some w in the box satisfies Dw>0, i.e., the data are strictly separable by a hyperplane through the origin. The Circles dataset is not separable this way, so the Entropic-Linear row in Table 3 is not a valid instance of the theorem. The stress-test note is right: the dual objective is bounded below but its infimum is not attained; along an unbounded λ direction the gradient goes to zero with w and b shrinking to zero. The spectral-gradient stopping rule then reports a degenerate solution, which explains the 0.46 accuracy and zero precision/recall. The paper never checks feasibility before running the method, and Section 5.3.2 treats the result as just another linear classifier's failure.\n\nThird, the experiments use a single train-test split with no error bars, degree selection by grid search, and no polynomial-kernel SVM baseline. The Entropic-Polynomial results are encouraging but not persuasive without repeated splits.\n\nWho this is for: someone working on entropy-based inverse problems and classification might find the dual formulation useful, but the added value over [8] is modest. The paper deserves a serious referee because the theorem is non-trivial and the infeasibility issue is addressable with an explicit feasibility test or a relaxed formulation. My recommendation: send it to peer review, but flag that the experimental protocol and the novelty disclosure need fixing. If the authors can show feasibility checks, add error bars, and position the contribution honestly against [8], it could be a decent methods note.","headline":"Correct convex dual for separable data, but the experiments on nonseparable data run outside the theorem's scope and the linear method is essentially the maxent-in-mean approach of [8].","tokens_in":9031,"tokens_out":3536,"would_cite":false,"duration_ms":36791,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["90C05","90C25","90C47","90C52","68T01","68T05","68T07","68T20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Binary classification can be recast as the minimization of an entropy-like convex function over a box, with the optimal separating hyperplane given by a closed-form dual solution.","keywords":["Classification","Linear discrimination","Ill-posed inverse problems","Convex optimization","Entropy minimization","Separating hypersurfaces","Lagrange duality","Polynomial classifiers"],"falsifier":"Take a linearly separable dataset with a known separating hyperplane, run the dual minimization until the gradient norm is below ε, and compute the primal residual ‖Ax*‖. If Theorem 7 is right, this residual should tend to zero as ε decreases; if it stalls above zero, the asserted equivalence between the primal and dual minimizers fails. The same test on the Circles dataset with a degree-1 polynomial will show a residual bounded away from zero, confirming that the infeasible-data case does not satisfy the theorem's premise.","tokens_in":7983,"feed_emoji":"🎯","tokens_out":6888,"duration_ms":75316,"temperature":0.7,"pith_summary":"This paper proposes a new way to train a binary classifier: instead of solving the usual separation inequalities, it turns them into equalities by giving each training point its own parallel separating plane with a positive distance. The training problem becomes the minimization of a strictly convex, entropy-like function Ψ over a bounded box, subject to exact linear constraints Ax = 0. The paper proves (Theorem 7) that the optimal weights and distances have an explicit closed form in terms of a Lagrange multiplier that minimizes a dual objective, so training reduces to a smooth convex optimization problem. The same construction is extended to polynomial decision surfaces by lifting the data into monomial features. If the result is correct, classification becomes a parameter-free convex program that also outputs an uncertainty band between the two classes, and numerical experiments show it is competitive with SVM, KNN, logistic regression, and perceptron on several datasets.","feed_headline":"Entropy dual turns classification into convex optimization","feed_subtitle":"A new training rule gives closed-form separating hyperplanes plus an uncertainty band.","key_machinery":"The load-bearing object is the pair of convex functions Ψ(x) and M(τ, ζ) in equations (12) and (13). Ψ is defined on the constraint box: each weight component w_j contributes a binary entropy term and each distance b_j contributes b_j(ln b_j − 1), making the box constraints automatic. M is the logarithm of the Laplace transform of a measure that places unit mass at the corners of the weight box and at all nonnegative integers for the distances, and it is the Legendre–Fenchel dual of Ψ. The key identity ∇M = (∇Ψ)^{-1} lets the authors solve the constrained primal problem through its unconstrained dual: the Lagrange multiplier λ* of the equality constraints minimizes λ ↦ M(A^t λ), and the primal variables are recovered by evaluating ∇M at A^t λ*. This dual substitution is what turns classification into a smooth convex optimization problem that standard spectral gradient methods can handle.","core_discovery":"The central claim is that the separation problem can be restated as finding x* = (w*, b*) that minimizes Ψ(x) over K = [-1,1]^N × R_{++}^M subject to Ax = 0, where the data matrix A encodes the requirement that every good point g_i satisfies ⟨g_i, w⟩ = b_1(i) and every defective point d_i satisfies ⟨d_i, w⟩ = -b_2(i) with all b components positive. The paper proves that this primal solution is given in closed form: the weight components are hyperbolic-tangent-like functions of the first N components of A^t λ*, while the distances are exponentials of the remaining components, and λ* minimizes the dual function M(A^t λ) defined in (13). This is Theorem 7, and it holds because the entropy function Ψ and the log-Laplace transform M are Legendre–Fenchel duals whose gradients are inverses of each other. A byproduct of the solution is the separation measure Δ(w*) = b_+ − b_- and an uncertainty band between the two hyperplanes H_g and H_d, which gives the classifier a natural region of uncertain prediction.","pith_inferences":["The dual formulation suggests a probabilistic reading of the distances b_i = e^{-λ_i}, which could be used to turn the uncertainty band into a calibrated probability of class membership; the paper does not develop this.","The same entropy–dual scheme could be applied to other feature maps, such as radial basis functions, by changing the matrix A; only polynomial features are tested, so the comparison to kernel SVMs is not direct.","A natural extension is to relax the hard equality constraints with slack variables or a norm penalty, which would let the method handle non-separable and noisy data while retaining the entropic dual structure; the paper notes the Ax = y case in passing but does not implement it.","The separation measure Δ(w*) could be used as a data-dependent margin for model selection, analogous to the margin in SVMs, but the paper does not explore this connection."],"forward_implications":["Training the entropic classifier reduces to minimizing a smooth convex function of the Lagrange multiplier λ, so no quadratic programming or nonsmooth subgradient methods are needed.","The construction extends to polynomial decision surfaces by enlarging the feature vector with monomials, and the reported experiments show near-perfect accuracy on the Circles, Spiral, and Moons datasets for degree 3 or 4.","Every solution comes with an explicit uncertainty band between the two boundary hyperplanes, so predictions can be assigned a confident region and an uncertain region.","Because the primal objective and constraints are convex, the method has a unique global minimizer whenever the feasibility assumption holds, avoiding the local-minimum issues of gradient-descent-trained classifiers."],"supporting_citations":[{"why":"Supplies the Legendre–Fenchel duality facts used to connect Ψ and M.","marker":"[2]"},{"why":"States the classical separation problem (Problem 1) and the convex optimization background.","marker":"[3]"},{"why":"Provides the separation measure Δ(w) that the authors compute from their solution.","marker":"[4]"},{"why":"Supplies the maxentropic inversion methodology that the entropy-minimization approach generalizes.","marker":"[7]"},{"why":"Previous application of maximum entropy in the mean to classification, which the paper adapts to a deterministic convex-entropy objective.","marker":"[8]"}],"fun_headline_variants":["Entropy dual gives closed-form hyperplanes","Separation via entropy minimization","Bounded-box entropy classifier with uncertainty band","Closed-form solution for separating surfaces","Entropy approach yields hyperplane and uncertainty"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole construction assumes that the training data admit an exact vector x in the box with positive distances satisfying Ax = 0, meaning a separating surface of the chosen degree that passes through every training point with a positive margin; when the data are not separable (as with a linear classifier on the concentric Circles dataset), the claimed primal–dual solution lies outside the theorem's scope.","fun_headline_variants_meta":{"raw":{"variants":["Entropy dual gives closed-form hyperplanes","Separation via entropy minimization","Bounded-box entropy classifier with uncertainty band","Closed-form solution for separating surfaces","Entropy approach yields hyperplane and uncertainty"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000156,"raw_usage":{"total_tokens":1222,"prompt_tokens":951,"completion_tokens":271,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":210}},"tokens_in":567,"tokens_out":271,"duration_ms":4208,"temperature":1.0,"reasoning_tokens":210,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:22:53.670539+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a linearly separable dataset with a known separating hyperplane, run the dual minimization until the gradient norm is below ε, and compute the primal residual ‖Ax*‖. If Theorem 7 is right, this residual should tend to zero as ε decreases; if it stalls above zero, the asserted equivalence between the primal and dual minimizers fails. The same test on the Circles dataset with a degree-1 polynomial will show a residual bounded away from zero, confirming that the infeasible-data case does not satisfy the theorem's premise.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Legendre–Fenchel duality facts used to connect Ψ and M."},{"cited_title":"Boyd and L","cited_arxiv_id":null,"evidence_quote":"States the classical separation problem (Problem 1) and the convex optimization background."},{"cited_title":"Cucker and D.-X","cited_arxiv_id":null,"evidence_quote":"Provides the separation measure Δ(w) that the authors compute from their solution."},{"cited_title":"Golan and H","cited_arxiv_id":null,"evidence_quote":"Supplies the maxentropic inversion methodology that the entropy-minimization approach generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous application of maximum entropy in the mean to classification, which the paper adapts to a deterministic convex-entropy objective."}],"review_version":1}