Pith. sign in

REVIEW 4 major objections 5 minor 19 references

Two-parameter superposable S-curves

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that any non-uniform pattern can be represented as a singular perturbation of a uniform distribution.

desk verdict A well-meaning but overclaimed paper: the cubic S-curve is fine, but the universal representation claim collapses because the fitted superpositions are not constrained to be CDFs. read the letter →

arxiv 2504.19488 v3 pith:VCAWBH7K submitted 2025-04-28 stat.ME cs.LG

Vijay Prakash S This is my paper · ORCID
classification stat.MEcs.LG
keywords singularperturbationS-curvescumulativedistributionfunctionuniformsuperpositionpatternrecognitionirisdatasetnonlinearitymeasure
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes that cumulative distribution data can be modelled by a two-parameter algebraic S-curve obtained by singularly perturbing the straight line $y=mx$ into $a y^3 + y = mx$. As the perturbation parameter $a$ tends to zero the curve reduces to the uniform cumulative distribution; as $a$ tends to infinity it approaches a degenerate single-value distribution. The paper then superposes such curves, sharing one value of $a$, and fits the weighted sum to empirical cumulative distributions. On the iris flower measurements the fitted parameters separate species, and the paper concludes that any non-uniform pattern can be represented as a singular perturbation of a uniform distribution.

What carries the argument

The load-bearing object is the Sa-m curve, the real solution $y(x)$ of $y-y_c = \frac{m(x-x_c)}{1+a(y-y_c)^2}$, a singular perturbation of the straight line $y=mx$; its derivative $\frac{dy}{dx} = \frac{m}{1+a(y-y_c)^2}$ is the associated bell curve. The superposition combines $n$ such curves with a single shared parameter $a$ but different slopes $m_i$, weights $p_i$, and inflection points $(x_{c_i}, y_{c_i})$: $y_{\mathrm{net}} = \sum_i p_i\left[\frac{m_i(x-x_{c_i})}{1+a(y_i-y_{c_i})^2} + y_{c_i}\right]$. The shared $a$ keeps the nonlinear adjustment common, and the linear superposition of otherwise linear-in-$m$ terms is what makes the family bounded and superposable; fitted values of $a$ and of the maximum slope $m$ are then used as pattern descriptors.

What would settle it

Fit the superposed family (Eq. 9) to a known heavy-tailed cumulative distribution such as a Pareto or log-normal over a wide range, with n increasing from 1 to, say, 11, and check whether the residual error decreases toward zero and the fitted curve remains nondecreasing within [0,1]; if residuals plateau or monotonicity fails, the representation claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the real solution of $y-y_c = \frac{m(x-x_c)}{1+a(y-y_c)^2}$ — the two-parameter Sa-m curve — together with its linearly superposed form $y_{\mathrm{net}} = \sum_i p_i \left[\frac{m_i(x-x_{c_i})}{1+a(y_i-y_{c_i})^2} + y_{c_i}\right]$, provides a statistical model for empirical cumulative distributions. The parameter $a$ acts as a shared nonlinearity: at $a\to 0$ the model is the line of a uniform CDF, and at $a\to\infty$ the derived bell curve concentrates at a single point, resembling a degenerate distribution. The paper shows that these superpositions fit the logistic sigmoid, the Gaussian error function, and the cumulative flower measurements of iris plants, and that the fitted maximum slope $m$, the ratio $m/(1+a)$, and a percentage-nonlinearity measure can be used to characterise and compare patterns. The paper's strongest claim is that every non-uniform pattern can be represented as a singular perturbation of a uniform distribution.

Load-bearing premise

The load-bearing premise is that the fitted curve behaves like a probability's running total: it never decreases and stays between 0 and 1; the fitting procedure does not enforce this, and several reported fits use negative weights, so if the premise fails the model is a curve-fitting device rather than a probability distribution.

Editorial extensions

If this is right

  • If the representation claim is correct, the fitted parameter $a$ directly quantifies how far an empirical distribution departs from uniformity: small $a$ means near-uniform spread, large $a$ means concentration near a single value.
  • The model supplies density estimates without kernel smoothing: the derivative of the fitted cumulative curve peaks at the maximum slope $m$, so $m$ and the normalised peak $\bar{m}$ can serve as probability-density summary values.
  • The percentage-nonlinearity measure compares datasets on a common scale; the paper finds, for instance, that the Gaussian error function is more nonlinear than the logistic sigmoid over the same interval, which would transfer to comparing empirical distributions.
  • For the iris data, the fitted parameters yield usable pattern-recognition rules, such as sepal length below 5.5 cm pointing to Iris setosa and petal length around 5 cm or more pointing to Iris virginica.
  • The superposed model can represent multimodal distributions and can locate points of inflection within the data, whereas a single Sa-m curve is unimodal.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the natural next step is to prove a universal-approximation statement for the superposed family as $n\to\infty$; the paper demonstrates good fits for finite $n$ but does not establish convergence.
  • Because the reported fits include negative weights and large parameter magnitudes, a version of the model that enforces monotonicity and the $[0,1]$ bound would turn the fitted curve into a genuine probability distribution; that constraint is not imposed in the paper.
  • The parameter $a$ behaves like an inverse-scale or temperature parameter, so the family could be compared against standard two-parameter distributions, such as generalised logistic or Weibull, on the same empirical CDFs; the paper does not make that comparison.
  • The iris analysis stops short of constructing a classifier; a direct testable extension is to use the fitted $(a, m, \bar{m}, NL)$ values as feature vectors for a nearest-neighbour or decision rule and measure classification accuracy.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces the algebraic S-curve defined by the singularly perturbed line equation ay^3 + y = mx, generalizes it to a two-parameter form with inflection point (x_c, y_c), and treats its derivative as a bell-shaped curve. It then proposes that finite linear superpositions of such S-curves with a common parameter a, Eq. (9), can serve as a statistical model for empirical cumulative distribution functions, and applies this model to sepal and petal measurements in the iris dataset. The authors introduce two measures, the maximum slope m and a 'percentage nonlinearity' measure defined in Eq. (12), and conclude that any non-uniform pattern can be represented as a singular perturbation to a uniform distribution. The manuscript includes Python code for fitting and reproducibility.

Significance. If the representation claim were established, the model would offer a simple algebraic alternative to logistic and Gaussian CDFs with an interpretable nonlinearity parameter a. The paper ships runnable fitting code and reports fitted parameter tables for all iris attributes, which is a strength for reproducibility. However, the central statistical claim is not supported: the fitted superpositions are not constrained to be cumulative distribution functions, no approximation theorem is given, and the iris fits are used both to fit the model and to infer the pattern, so the explanatory claims are partly restatements of the fit. As it stands, the significance is limited to a descriptive curve-fitting exercise rather than a validated statistical model.

major comments (4)
  1. [Section 3, Eq. (9); Tables 2, 4, 6, 8] The fitted superposition ynet is not constrained to be a cumulative distribution function. The weights p_i are unconstrained, and Tables 2, 4, 6, and 8 report negative p_i values (e.g., p1 = -0.136528 in Table 2, p1 = -7.976545 in Table 6, p1 = -11142.446584 in Table 8). Consequently ynet can decrease, leave the interval [0,1], and its derivative (claimed as the probability density) can take negative values. No theorem or constraint is provided to ensure monotonicity or range. Therefore the fitted object cannot be interpreted as a CDF, and the iris fits support at most a descriptive smoothing claim, not the universal representation claim.
  2. [Section 1, Eq. (5); Section 3 code sam_der] The displayed derivative in Eq. (5) is missing the factor 3. Implicit differentiation of ay^3 + y = mx gives y' = m/(1 + 3a y^2), not m/(1 + a y^2). The paper's own code sam_der uses the correct form, 1 + 3*a*sam(...)**2, so the displayed equation contradicts the code. This inconsistency must be corrected before the bell-curve plots and any quantitative statements about tail behavior are reliable.
  3. [Conclusions] The claim that 'any non-uniform pattern can be represented as a singular perturbation to uniform distribution' is not supported by the evidence presented. The paper fits four attributes of the iris dataset and does not provide an approximation theorem, a density result, or error bounds for the superposed family. The analogy with Cybenko/Hornik universal approximation, cited in the Introduction, does not carry over without a proof that finite superpositions of Sa-m curves with a common a can approximate arbitrary CDFs. As stated, this conclusion is a conjecture rather than a demonstrated result.
  4. [Section 1, Fig. 1(b)] The limiting argument from uniform to degenerate distribution is not established. The derivative as a function of y is m/(1 + 3a y^2); as a → infinity, the integral of this function over the real line tends to 0, and its pointwise limit is zero for y ≠ 0 and m at y = 0, which is not a probability density. The claim that the support becomes finite at y = 0 requires a rigorous limiting construction (for example, a rescaled density), which the paper does not provide.
minor comments (5)
  1. [Fig. 2(b)] The axis label 'y, = 1' appears garbled; it should read 'dy/dx' or a similar notation for the derivative.
  2. [Section 3.1, item 3] There is a typo: 'legnth' should be 'length'.
  3. [Section 3, histogram code] The histograms are normalized by c/sum(c) after calling density=True, which is not a proper probability-density normalization; this normalization is inconsistent with the derivative curves, which are normalized by a sum over bin edges rather than by an integral.
  4. [Section 2.1, Eq. (12)] The quantity in Eq. (12) is called a percentage nonlinearity measure, but it can exceed 100 (for example, NL = 894.998545 in Table 6), so 'percentage' is a misnomer unless the measure is rescaled.
  5. [Tables 4, 6, 8] Several fitted values are extreme (for example, a = 225514.913420, a = 1.185 × 10^-7, p1 = -11142.446584) and may indicate non-identifiability or overfitting; the paper does not discuss numerical conditioning or report standard errors.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the universal representation claim is an unsupported generalization, not a circular reduction.

full rationale

The paper's derivation chain is algebraic and self-contained: Eq. (4) defines the Sa-m curve, Eq. (6) gives its explicit real solution, and Eq. (9) defines the superposition. Fitting this superposition to iris CDFs and taking derivatives to obtain density curves is a transparent transformation of the fitted model, not a prediction equivalent to the input by construction. The nonlinearity measure in Eq. (12) and the parameters a and m are post-hoc summaries of the same fit, which may limit their interpretive value but does not make the derivation circular. The central claim that 'any non-uniform pattern can be represented as a singular perturbation to uniform distribution' is an inductive generalization based on successful fits to a few datasets; it lacks a proof or held-out validation, but the model family is not defined in terms of the iris data, and the fits are not used to redefine the model. The self-citations ([10], [11], [12], [15], [16], [17]) are used as provenance for the S-curve equation and as examples of prior applications; the superposition formula and derivative are derived in the present paper, and no uniqueness theorem or external constraint is imported to force the choice of model. The paper also acknowledges limitations such as sensitivity to initial conditions and possible overfitting. The apparent inconsistency between Eq. (5), which omits the factor 3 in the derivative, and the code's sam_der, which includes it, is a correctness issue, not a circularity issue. Likewise, negative fitted weights and possible non-monotonicity of ynet concern whether the fitted object is a valid CDF; these are statistical validity concerns, not circular reasoning. Therefore no load-bearing circular step is present.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central representation claim rests on the cubic equation, the superposition ansatz, and the uniform-to-degenerate interpretation. The two parameters named in the title are accompanied by inflection coordinates, weights, and the number of components in any real fit, so the effective parameter count is higher than two. The fitted values in Tables 1-8 are the empirical content; no independent validation is provided.

free parameters (5)
  • a = per dataset, e.g., 1.519780 (Iris setosa sepal length n=1)
    Fitted nonlinearity parameter; enters Eq. (3)/(4) and controls curvature. Values in Tables 1-8.
  • m (slope at inflection; m_i for each component) = per dataset, e.g., 1.086830
    Fitted slope at inflection; for superpositions, one slope per S-curve.
  • p_i superposition weights = per component, e.g., -0.136528
    Fitted weights in Eq. (9); unconstrained, can be negative.
  • inflection points xc_i, yc_i = chosen from data, e.g., 5.4 cm, 0.9
    Selected from data by highest absolute slope (Section 2) or most frequent values (Section 3); effectively data-derived rather than independently justified.
  • number of components n = n = 1, 2, 3, 4, ... (chosen by user)
    Model complexity chosen by user; not selected by a data-driven criterion.
assumptions (3)
  • standard math The cubic equation a y^3 + y = m x has a unique real solution defining an S-shaped function.
    Relied on throughout; standard algebra and calculus, but no proof is given in the paper.
  • ad hoc to paper A linear combination of Sa-m curves with a common parameter a remains a valid model for a cumulative distribution.
    Eq. (9) is used as a statistical CDF model; monotonicity and range properties are not established for superpositions with arbitrary weights.
  • ad hoc to paper As a -> 0 and a -> infinity, the derivative ranges continuously from a uniform to a degenerate distribution.
    Stated in the Abstract and Section 1 as motivation; it is not derived, and the derivative is not normalized as a probability density over the real line.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Two-parameter superposable S-curves." pith.science (2026). https://pith.science/paper/VCAWBH7K

@misc{pith2026250419488,
  author       = {Pith},
  title        = {Pith review of: Two-parameter superposable S-curves},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VCAWBH7K}},
  note         = {Machine review of arXiv:2504.19488}
}
abstract

Straight line equation $y=mx$ with slope $m$, when singularly perturbed as $ay^3+y=mx$ with a positive parameter $a$, results in S-shaped curves or S-curves on a real plane. As $a\rightarrow 0$, we get back $y=mx$ which is a cumulative distribution function of a continuous uniform distribution that describes the occurrence of every event in an interval to be equally probable. As $a\rightarrow\infty$, the derivative of $y$ has finite support only at $y=0$ resembling a degenerate distribution. Based on these arguments, in this work, we propose that these S-curves can represent maximum entropy uniform distribution to a zero entropy single value. We also argue that these S-curves are superposable as they are only parametrically nonlinear but fundamentally linear. So far, the superposed forms have been used to capture the patterns of natural systems such as nonlinear dynamics of biological growth and kinetics of enzyme reactions. Here, we attempt to use the S-curve and its superposed form as statistical models. We fit the models on a classical dataset containing flower measurements of iris plants and analyze their usefulness in pattern recognition. Based on these models, we claim that any non-uniform pattern can be represented as a singular perturbation to uniform distribution. However, our parametric estimation procedure have some limitations such as sensitivity to initial conditions depending on the data at hand.

Figures

Figures reproduced from arXiv: 2504.19488 by the authors.

Figure 1
Figure 1. S- and bell- shaped curves of modified exponential function and modified (singularly perturbed) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Fitting superposed Sa-m curves on S- and bell- shaped curves obtained by modifying the exponential function. We get better fits as more Sa-m curves are added to the superposition. we get ynet = p1 m1(x − x1) 1 + a(y1 − yc1) 2 + p2 m2(x − x2) 1 + a(y2 − yc2) 2 = p1m1(x − x1)(1 + a(y2 − yc2) 2 ) + p2m2(x − x2)(1 + a(y1 − yc1) 2 ) (1 + a(y1 − yc1) 2)(1 + a(y2 − yc2) 2) . (10) It can be seen that as we superpose more yi… view at source ↗
Figure 3
Figure 3. Variation of parameters a and m based on the fits on Sexp for two different intervals x ∈ [−3, 3] (shown as ‘×’) and x ∈ [−5, 5] (shown as ‘•’) with n as the x− axis. 0.5 1.0 1.5 2.0 1 2 3 4 5 6 7 8 9 10 11 Number of Sa m curves in superposition 0.00 0.02 0.04 0.06 0.08 0.10 (a) Estimation of a stabilizes to some extent as n increases. 1 2 3 4 5 6 7 8 9 10 11 Number of Sa m curves in superposition 0.400 0.425 0.450 … view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Variation of parameters a and m based on the fits on Gauss error function whose derivative is ˜y for two different initial conditions pi = 1, mi = 1 (shown as ‘×’) and pi = 0, mi = slope at xci between data points (shown as ‘•’) with n as the x− axis. Pn i=1 pimi is dy…
Figure 5
Figure 5. Figure 5: Percentage nonlinearity measures against the number of [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Fitted models on cumulative distributions of sepal length (in cm) of iris plants. The derived [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Fitted models on cumulative distributions of sepal width (in cm) of iris plants. The derived [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Fitted models on cumulative distributions of petal length (in cm) of iris plants. The derived [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Fitted models on cumulative distributions of petal width (in cm) of iris plants. The derived [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Fitted models on cumulative distributions of petal width (in cm) of Iris setosa after introducing [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 16 canonical work pages

  1. [1]

    Kucharavy, D., and De Guio, R. (2011). Application of S-shaped curves. Procedia Engineering, 9, 559- 572

    work page 2011

  2. [2]

    G. Cybenko. (1989). Approximation by superpositions of a sigmoidal function. Mathematics of Control, Signals, and Systems, 2 (4), pp.303-314. 10.1007/BF02551274. hal-03753170. 13

    doi:10.1007/bf02551274 1989

  3. [3]

    (1989), Multilayer Feedforward Networks are Universal Approximators, Neural Networks, Vol

    Hornik, K., Stinchcombe, M., and White, H. (1989), Multilayer Feedforward Networks are Universal Approximators, Neural Networks, Vol. 2, pp. 359-366

    work page 1989

  4. [4]

    Tsoularis, A., and Wallace, J. (2002). Analysis of logistic growth models. Mathematical biosciences, 179(1), 21-55

    work page 2002

  5. [5]

    Kyurkchiev, N., and Markov, S. (2015). Sigmoid functions: some approximation and modelling aspects. LAP LAMBERT Academic Publishing, Saarbrucken, 4

    work page 2015

  6. [6]

    Fekedulegn, D., Mac Siurtain, M.P., and Colbert, J.J. (1999). Parameter estimation of nonlinear growth models in forestry. Silva Fennica 33(4): 327–336

    work page 1999

  7. [7]

    E., Bradley Jr, E

    Turner Jr, M. E., Bradley Jr, E. L., Kirk, K. A., and Pruitt, K. M. (1976). A theory of growth. Mathematical Biosciences, 29(3-4), 367-373

    work page 1976

  8. [8]

    Stahl, S. (2006). The evolution of the normal distribution. Mathematics magazine, 79(2), 96-113

    work page 2006

Show all 19 references
  1. [9]

    Taleb, N. N. (2020). Statistical consequences of fat tails: Real world preasymptotics, epistemology, and applications. arXiv preprint arXiv:2001.10488

    2020

  2. [10]

    S., and Vijay Prakash, S

    Shruti, I. S., and Vijay Prakash, S. (2025). A biological growth model using continued fraction of straight lines, bioRxiv doi:10.1101/2025.01.07.631841

    2025 doi

  3. [11]

    Shruti, I. S. (2025). A biological growth curve is a sum of two distinct S-curves, bioRxiv doi:10.1101/2025.02.06.636984

    2025 doi

  4. [12]

    Vijay Prakash S. (2024). Real-valued continued fraction of straight lines. arXiv preprint arXiv:2412.16191

    2024 arXiv

  5. [13]

    A. N. Tikhonov. (1943). On the stability of inverse problems. Doklady Akademii Nauk SSSR, 39(5): 195–198

    1943

  6. [14]

    Tibshirani

    R. Tibshirani. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society. Series B. Methodological, 58(1):267–288

    1996

  7. [15]

    S., and Vijay Prakash, S

    Shruti, I. S., and Vijay Prakash, S. (2025). Fitting multiple bacterial growth data using continued fraction of straight lines, bioRxiv doi:10.1101/2025.01.27.634991

    2025 doi

  8. [16]

    S., and Vijay Prakash, S

    Shruti, I. S., and Vijay Prakash, S. (2025). Modeling Montbeillard’s height data of a human male, bioRxiv doi:10.1101/2025.03.02.641023

    2025 doi

  9. [17]

    S., and Vijay Prakash, S

    Shruti, I. S., and Vijay Prakash, S. (2025). Analyzing enzyme kinetics using a growth model. ChemRxiv. doi:10.26434/chemrxiv-2025-57c99

    2025 doi

  10. [18]

    Fisher, R. (1936). Iris [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C56C76

    1936 doi

  11. [19]

    B., Rawlik, M., Ingargiola, A., and Nelson, A

    Newville, M., Stensitzki, T., Allen, D. B., Rawlik, M., Ingargiola, A., and Nelson, A. (2016). LMFIT: Non-linear least-square minimization and curve-fitting for Python. Astrophysics Source Code Library, ascl-1606. 14

    2016

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.