{"id":"a6d72f0f-52aa-4000-bfb5-aeff3a2391c3","arxiv_id":"2506.11210","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For 1 < p < infinity, the Lp-modulus, Fourier rate, and step rate parameterizations of the space of p-integrable functions are mutually linearly equivalent.","lead":"The paper proves that three natural ways to measure the complexity of an integrable function, its Lp-modulus, its Fourier series convergence rate, and its step-function approximation rate, are equivalent up to constant factors. This provides a robust definition of computational complexity for spaces of p-integrable functions, which is needed for complexity analysis of partial differential equations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 16(a) proof underabsorbs the Lp Dirichlet-projection norm; the displayed chain only closes if C_p ≤ 1, which fails for p ≠ 2.","rationale":"The reader identifies the reliance on the Lp boundedness of Dirichlet projections as the load-bearing step in Theorem 16(a). I agree that this is the critical analytic input, but the sharpest problem is not that the Riesz theorem is cited rather than proved; it is that the proof fails to incorporate the size of the Riesz constant C_p into the argument shift. The displayed inequality needs ∥f − g_K∥ ≤ 2^{−(n+N)} with N depending on p, otherwise the factor (C_p + 1) prevents the final error from reaching 2^{−n}. Because the theorem’s O(1) notation explicitly allows p-dependent constants, a modest modification of the proof should repair the gap. The false Example 19 claim and the inaccurate Schauder-basis remark are real blemishes but peripheral to the central equivalence. The main theorem is therefore not invalidated, but the written proof is incomplete at a central point, so the reader’s CONDITIONAL verdict remains appropriate.","tokens_in":16943,"tokens_out":17130,"duration_ms":187516,"concrete_test":"Independently re-derive the proof of Theorem 16(a) with explicit constants: set N_p = ⌈log2(C_p + 1)⌉ + 1, choose δ = 2^{−μ(n+N_p)}, and require K ≥ 2^{μ(n+N_p)+n+N_p}. Verify that the chain ∥F_K f − f∥_p ≤ (C_p + 1)2^{−(n+N_p)} ≤ 2^{−n} closes. If it closes, Theorem 16(a) is valid and the proof text needs a correction; if it does not close for some 1 < p < ∞, the stated bound is false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 16(a) in §2.5 first shows ∥f − f∗F_K∥_p ≤ 2^{−n}. It then sets g_K := f∗F_K and writes ∥F_K f − f∥_p ≤ ∥F_K(f − g_K)∥_p + ∥F_K g_K − g_K∥_p + ∥f − g_K∥_p ≤ (C_p + 1)2^{−n}, where C_p is the Lp operator norm of the Dirichlet projection. The desired conclusion is ∥F_K f − f∥_p ≤ 2^{−n}, so the displayed chain is only valid if C_p ≤ 1. For p ≠ 2, C_p is strictly larger than 1 (the Hilbert transform is not an isometry on Lp), so the proof as written does not close. The gap is repairable because the theorem’s O(1) argument shifts may depend on p: one can start from the Fejér approximation with error 2^{−(n+N)} for N ≈ log2(C_p + 1), replacing μ(n+1) by μ(n+N) and taking K ≥ 2^{μ(n+N)+n+N}. Thus the theorem statement is very likely true, but the displayed proof omits exactly the step where the Riesz bound is absorbed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a second-order parameterized complexity theory for the Banach space Lp(T), 1<p<∞. Three parameterizations are considered: the Lp-modulus of translation continuity (from Steinberg), the rate of approximation by step functions, and the rate of convergence of Fourier series. Theorem 16 claims that these three parameters are mutually bounded up to constant factors in both argument and value, and Corollary 17 concludes that they induce the same notion of second-order linear (or polynomial) time complexity. The proofs use the Fréchet–Kolmogorov theorem, Fejér summation, M. Riesz boundedness of the Dirichlet projection, and elementary norm inequalities. The paper also sketches examples of functions with large moduli, generalizations to Sobolev spaces and higher-dimensional tori, and a unifying definition of Schauder-system rates.","tokens_in":17213,"tokens_out":18953,"duration_ms":205155,"significance":"If the proof gaps identified below are repaired, this is a useful and credible contribution. It gives the first unified quantitative treatment of natural encodings of Lp(T), generalizing the well-understood continuous-function case, and it provides a principled answer to the parameterization question for integrable functions. The central derivation is grounded in standard external theorems (Fréchet–Kolmogorov, M. Riesz, Fejér), and the paper does not assume the equivalence it proves. The authors are also explicit about limitations, such as the failure of ball-multiplier Fourier convergence in higher dimensions. These are genuine strengths: the claimed equivalence is plausible and the overall architecture is sound, even though the written proofs need local repairs.","major_comments":[{"comment":"The displayed chain concludes ∥F_Kf−f∥_p ≤ (C_p+1)2^{−n}, not ≤2^{−n}. For p≠2, the Lp norm of the Dirichlet projection is strictly larger than 1, so the bound at precision n does not satisfy the requirement of Definition 15. The theorem's O(1) argument shift can absorb C_p, but the proof must say so explicitly: start from the Fejér approximation with error 2^{−(n+N)} for N≈log_2(C_p+1) and then set K accordingly. As written, the proof is incomplete at exactly the point where the Riesz bound is used.","section":"§2.5, proof of Theorem 16(a)"},{"comment":"Fact 11e is stated without proof and is applied in parts (b) and (d) with the wrong index. As stated, Fact 11e bounds μ_f(n) using an approximant with error 2^{−(n+2)}. The displayed proofs invoke it with K=2^{φ(n)} and K=2^{σ(n)}, which only guarantee error 2^{−n}, so the estimates do not close as written. The fix is to take K=2^{φ(n+2)} and K=2^{σ(n+2)} respectively (or to prove a variant of Fact 11e with the weaker closeness assumption), and to include the short triangle-inequality proof of Fact 11e.","section":"§2.5, proofs of Theorem 16(b) and (d); Fact 11e"}],"minor_comments":[{"comment":"The symbol F_K is used both for the Fejér kernel and for the Dirichlet projection operator; please use distinct notations (for example V_K for the kernel and S_K for the projection) to avoid confusion.","section":"§2.5(a)"},{"comment":"The claimed asymptotic rates in Example 19 appear incorrect as stated: for f(t)=Σ 2^{−k}e^{2πi2^k t}, the L2 tail after truncating at k=m has norm Θ(2^{−m}), so Definition 15 gives φ(n)=Θ(n), not exponential growth. The double-exponential example is similar; please correct the statements or clarify that the exponential growth refers to the degree 2^{φ(n)} rather than to φ(n) itself.","section":"§3.1, Example 19"},{"comment":"The sentence 'Proceeding from a multitape to a one-tape Turing machine is known to incur an increase in running time increase most polynomial' contains a duplicated and ungrammatical phrase; it should say 'at most polynomial'.","section":"§1.2.1"},{"comment":"The asserted generalization of Corollary 17 to T^d with maximum-norm summation is stated without a proof sketch; given the known dependence on the summation convention (see [Fef71]), please add at least an outline of why the one-dimensional proof adapts.","section":"§3.3"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is very likely correct, and the two proof gaps are local and repairable, so rejection is not warranted. The Section 3.1 examples contain statements that are false as written and should be corrected before final acceptance, as they could mislead readers about the growth of the Fourier rate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Core result is new and very likely true: for 1<p<∞, the Lp-modulus, step rate, and Fourier rate are mutually linearly equivalent on Lp(T). That answers the LSZ open question in the natural setting. The proof is mostly solid, but the printed proof of Theorem 16(a) skips the absorption of the Dirichlet projection norm C_p. The stress-test note is on the mark: the chain gives (C_p+1)·2^{-n}, not 2^{-n}, so the displayed argument does not close. The fix is easy—start from a Fejér approximation with error 2^{-(n+N)} for N≈log_2(C_p+1)—so the theorem statement is not in danger.\n\nWhat is genuinely new is the step-rate and Fourier-rate formalization plus the mutual linear equivalence. The analytic work (Fejér kernel, M. Riesz, Fréchet-Kolmogorov) is clean and mostly well documented. The sketches for Sobolev spaces and higher-dimensional tori are useful pointers.\n\nSoft spots, in order of importance:\n1. The Theorem 16(a) proof gap described above. Minor, but it should be fixed in the text.\n2. Example 19 is inaccurate: the lacunary series with coefficients 2^{-2^k} has L2 error O(2^{-N}) after frequencies up to 2^N, i.e., linear Fourier rate, not exponential; the doubly exponential claim is likewise off.\n3. Fact 11e (modulus transfer under Lp-close functions) is used for parts (b) and (d) but only stated, not proved. It is true and follows from the triangle inequality in two lines.\n4. The remark that monomials form a Schauder basis of C[0,1] is false; they span a dense subspace, but the coordinate functionals are not continuous in sup norm.\n\nThe citation pattern is fair: Steinberg is credited for the Lp-modulus, and the new theorem is the authors' own. No circularity. This paper deserves a serious referee. I would send it out with a request to fix the proof gap, correct Example 19, and add the short proof of Fact 11e.","headline":"A new and likely correct equivalence theorem for Lp parameterizations; the proof of part (a) and Example 19 need fixes before publication.","tokens_in":17751,"tokens_out":7553,"would_cite":true,"duration_ms":76149,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["03D78","42A20","46E30","68Q15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Three natural ways to parameterize p-integrable functions are linearly equivalent for 1<p<∞.","keywords":["second-order parameterized complexity","Lp spaces","Lp-modulus","Fourier series convergence","step-function approximation","computable analysis","compactness in function spaces","approximation theory"],"falsifier":"Take the lacunary function $f(t)=\\sum_{k\\ge 0} 2^{-k} e^{2\\pi i 2^k t}$ in $L^p(T)$, say $p=3$, and compute or bound its $L^p$-modulus $\\mu(n)$, step rate $\\sigma(n)$, and Fourier rate $\\varphi(n)$ for increasing $n$. The theorem predicts $\\varphi(n) \\le C\\mu(n+C)+n+C$ and $\\mu(n) \\le C\\sigma(n+C)+pn+C$ for constants $C$ depending only on $p$; if a computation ever shows one parameter growing more than linearly faster than another, such as $\\varphi(n)$ comparable to $2^{\\mu(n)}$, the claimed mutual equivalence is false. The paper's own examples assert exponential growth for all three, so the decisive observation is that no doubly exponential separation appears.","tokens_in":16736,"feed_emoji":"📐","tokens_out":13634,"duration_ms":148720,"temperature":0.7,"pith_summary":"The paper aims to give the space $L^p$ of $p$-integrable complex functions on the unit circle a second-order parameterized complexity theory: a way of measuring the cost of computing with a function that can depend not only on the desired precision $n$ but also on a structural 'modulus' of the function itself, just as the modulus of continuity does for continuous functions. It proposes three natural structural measurements for $L^p$: an $L^p$-modulus based on how much a function changes under small shifts, a rate of approximation by step functions, and a rate of convergence of Fourier series. The main result, Theorem 16, proves that for every fixed $1<p<\\infty$ these three measurements are linearly equivalent, meaning each is bounded by a constant multiple of another up to constant shifts in the argument. The consequence is that the three parameterizations induce the same notion of second-order linear and polynomial time complexity, so algorithms for integrable functions can be analyzed with whichever parameter is most convenient.","feed_headline":"Three Lp complexity parameters prove equivalent","feed_subtitle":"Lp-modulus, Fourier rate, and step rate match up to constants, giving integrable functions one complexity scale.","key_machinery":"The central object is the compact cover of $L^p(T)$ built from the quantitative Fréchet–Kolmogorov theorem: the sets $L^p_\\mu(T,R)$ of functions with $p$-norm at most $R$ and $L^p$-modulus at most $\\mu$ are compact, and every compact subset of $L^p$ is contained in one such set. The four inequalities are proved by moving between a function and its approximations: part (a) uses Fejér means and the uniform boundedness of Fourier partial sums on $L^p$ for $1<p<\\infty$ to control the Dirichlet projection; part (b) bounds the $L^p$-modulus of a finite Fourier series directly and then applies the transfer lemma for $L^p$-close functions; parts (c) and (d) compare against averaged step functions, with the extra $pn$ in part (d) coming from the shift sensitivity of a $K$-step function. The transfer lemma, that functions close in $L^p$ have comparable moduli, is what makes all four directions interchangeable.","core_discovery":"On the space $L^p(T,1)$ of $p$-integrable complex functions on the unit circle with $\\|f\\|_p \\le 1$ and fixed $1<p<\\infty$, let the $L^p$-modulus $\\mu(n)$ be the least shift precision needed to keep the $L^p$ distance between $f$ and its $\\delta$-shift below $2^{-n}$; let the step rate $\\sigma(n)$ be the least $K$ such that a $K$-level step function approximates $f$ to error $2^{-n}$; and let the Fourier rate $\\varphi(n)$ be the least order of a Fourier partial sum achieving the same error. Theorem 16 proves the four inequalities $\\varphi(n) \\le \\mu(n+O(1)) + n + O(1)$, $\\mu(n) \\le 2\\varphi(n+O(1)) + n + O(1)$, $\\sigma(n) \\le \\mu(n+O(1))$, and $\\mu(n) \\le \\sigma(n+O(1)) + pn + O(1)$. Corollary 17 then yields $O(\\varphi(n+O(1)) + n) = O(\\mu(n+O(1)) + n) = O(\\sigma(n+O(1)) + n)$, so the three parameters induce the same second-order linear and polynomial time complexity. This extends the continuous-function parameterization and answers the stated question of how to cover $L^p$ by compact sets parameterized by a modulus.","pith_inferences":["If the equivalence is adopted as a definition, functions with $\\mu(n)=O(n)$ form an $L^p$ analogue of Lipschitz functions, and one could compare that class with Sobolev or Besov smoothness; the paper's Sobolev generalization suggests this comparison is within reach.","The Fourier leg of the theorem is the only place where the restriction $1<p<\\infty$ is truly load-bearing; a natural test is whether the step-rate and $L^p$-modulus equivalence survives at $p=1$, where Fourier partial sums no longer converge.","The higher-dimensional extension works only for max-norm partial Fourier sums, since Euclidean-ball sums fail to converge in $L^p$ for $p\\neq 2$ in dimension two or more; this suggests the parameterization is sensitive to the geometry of the summation rule, and other summation geometries could define inequivalent rates.","The paper's basis-change example shows the Fourier rate is not invariant under a change of orthonormal basis, so a useful follow-up would be to characterize which Schauder bases of $L^p$ yield rates linearly equivalent to the $L^p$-modulus."],"forward_implications":["Any second-order runtime bound for an algorithm on $L^p(T)$ stated with one of these parameters translates to the same linear or polynomial bound stated with either of the other two.","The compact cover $L^p(T)=\\bigcup_{r,\\mu} L^p_\\mu(T,2^r)$ gives each integrable function a finite structural parameter, opening the way to bit-complexity statements for operators on $L^p$ data, such as solution operators of PDEs.","Step-function approximation, which is computationally the simplest of the three encodings, can be used as a canonical parameter without losing information about smoothness measured by the $L^p$-modulus or by Fourier convergence.","Because the equivalence is uniform over the unit ball with constants depending only on $p$, complexity classifications based on these parameters do not depend on the particular integrable function being studied."],"supporting_citations":[{"why":"Introduces the $L^p$-modulus and its compactness justification, which the paper adopts as the first parameter.","marker":"[Ste17]"},{"why":"Supplies the uniform $L^p$ boundedness of Fourier partial sums for $1<p<\\infty$ used to prove part (a) of Theorem 16.","marker":"[Duo01]"},{"why":"Provides the Fréchet–Kolmogorov and Arzelà–Ascoli compactness theorems behind the compact cover and the modulus-based parameterization.","marker":"[Br´e11]"},{"why":"Gives the quantitative coding and complexity theory for compact metric spaces that the present work extends from continuous to integrable functions.","marker":"[LiZ25]"},{"why":"Defines second-order polynomial-time computability for operators in analysis, the complexity notion the new parameterizations are meant to support.","marker":"[KaC12]"},{"why":"Poses the question of how to parameterize $L^p$ that Corollary 17 answers.","marker":"[LSZ20]"}],"fun_headline_variants":["Lp-modulus, Fourier rate, step rate: all equivalent","Three Lp complexity measures collapse into one","Integrable functions: one complexity scale rules them all","Second-order complexity of Lp: three parameters, one rate","Lp modulus, Fourier, step: equivalent rates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is a deep analytic fact the paper cites without proof: for each exponent $p$ strictly between 1 and infinity, taking the $K$-th partial sum of a Fourier series is a bounded operation on $p$-integrable functions, uniformly in $K$; if that uniformity fails, the claimed equivalence of Fourier rate with the other two parameters breaks.","fun_headline_variants_meta":{"raw":{"variants":["Lp-modulus, Fourier rate, step rate: all equivalent","Three Lp complexity measures collapse into one","Integrable functions: one complexity scale rules them all","Second-order complexity of Lp: three parameters, one rate","Lp modulus, Fourier, step: equivalent rates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000668,"raw_usage":{"total_tokens":3040,"prompt_tokens":933,"completion_tokens":2107,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":549,"completion_tokens_details":{"reasoning_tokens":2028}},"tokens_in":549,"tokens_out":2107,"duration_ms":18407,"temperature":1.0,"reasoning_tokens":2028,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:13:16.282830+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the lacunary function $f(t)=\\sum_{k\\ge 0} 2^{-k} e^{2\\pi i 2^k t}$ in $L^p(T)$, say $p=3$, and compute or bound its $L^p$-modulus $\\mu(n)$, step rate $\\sigma(n)$, and Fourier rate $\\varphi(n)$ for increasing $n$. The theorem predicts $\\varphi(n) \\le C\\mu(n+C)+n+C$ and $\\mu(n) \\le C\\sigma(n+C)+pn+C$ for constants $C$ depending only on $p$; if a computation ever shows one parameter growing more than linearly faster than another, such as $\\varphi(n)$ comparable to $2^{\\mu(n)}$, the claimed mutual equivalence is false. The paper's own examples assert exponential growth for all three, so the decisive observation is that no doubly exponential separation appears.","supporting_citations":[],"review_version":1}