Papers
Topics
Authors
Recent
Search
2000 character limit reached

Positive Magnitude: Concepts & Applications

Updated 4 July 2026
  • Positive magnitude is a multifaceted concept defining quantitative thresholds and robust selection criteria across climate, neural network, seismology, and metric space analyses.
  • In climate forecasting, it quantifies ENSO amplitude by linking complexity measures like SysSampEn with highly correlated prediction models.
  • In both neural network binarization and seismic statistics, positive magnitude guides weight selection and robust difference measures, while in metric spaces it guarantees invariant positivity.

ā€œPositive magnitudeā€ is not a single standardized term in the recent arXiv literature. It appears in several technically distinct senses: as a forecasted positive-valued climatic amplitude, as a magnitude-based selection rule in representation learning, as a robust statistic built from positive earthquake magnitude differences, and as part of the theory of Leinster magnitude on positive definite metric spaces. A broader adjacent literature studies positivity of quantities governed by size, scale, or amplitude, including operator forms, EFT couplings, parton distributions, and scattering amplitudes (Meng et al., 2019, Lin et al., 2021, Tian et al., 2021, Lippiello et al., 2024, Meckes, 2010).

1. Positive El NiƱo magnitude as a complexity-prediction relation

In ENSO forecasting, the phrase acquires a directly predictive meaning. The paper ā€œComplexity based approach for El Nino magnitude forecasting before the ā€˜spring predictability barrierā€™ā€ introduces System Sample Entropy, or SysSampEn, as a complexity measure for the system of temperature anomaly time series in the NiƱo 3.4 region, using the 22 grid points in that region as a coupled system (Meng et al., 2019). The quantity is defined by

SysSampEn(m,p,leff,γ)=āˆ’log⁔ ⁣(AB),SysSampEn(m,p,l_{eff},\gamma)=-\log\!\left(\frac{A}{B}\right),

where AA counts similar subsequences of length m+pm+p, BB counts similar subsequences of length mm, leffl_{eff} is the effective number of data points used per yearly calculation, and γ\gamma sets the similarity tolerance. In the formulation given in the paper, SysSampEn measures how likely two sequences that match for mm days remain matching for the next pp days.

The central empirical result is a strong positive correlation between the previous calendar year’s SysSampEn and the following El NiƱo’s magnitude. Across ERA-Interim, ERA5 air temperature, ERA5 SST, and JRA55-do SST, the average reported correlation is approximately rā‰ˆ0.90r\approx 0.90 for the best parameter choices, while the hindcast/forecast comparison of realized events gives about r=0.84r=0.84. For ERA-Interim, the strongest reported case uses m=60m=60 days, p=15p=15 days, leff=345l_{eff}=345 days, and γ=9\gamma=9, yielding r=0.99r=0.99. Older approaches based on averaged SampEn or Cross-SampEn perform much worse, with average correlation only about r=0.42r=0.42 (Meng et al., 2019).

Forecasting is implemented through a linear map,

Y=aX+b,Y=aX+b,

with YY the El NiƱo magnitude and XX the previous year’s SysSampEn. The paper reports an average hindcast RMSE of 0.23∘C0.23^\circ C for El NiƱo events from 1984 to 2017, and for the ongoing 2018 event it forecasts a weak El NiƱo with magnitude 1.11±0.23∘C1.11\pm 0.23^\circ C, where the uncertainty is taken as 1ƗRMSE1\times RMSE (Meng et al., 2019). In this setting, ā€œpositive magnitudeā€ denotes a positive forecasted ENSO amplitude inferred from a prior complexity state. The authors also suggest a physical interpretation: stronger El NiƱos may be preceded by weaker horizontal synchronization and higher temporal disorder in temperature anomalies.

2. Magnitude-based binarization in neural networks

In binary neural networks, ā€œpositive magnitudeā€ denotes a rule in which magnitude, rather than sign, determines the discrete code. The paper ā€œSiMaN: Sign-to-Magnitude Network Binarizationā€ replaces standard {āˆ’1,+1}\{-1,+1\} sign binarization with a {0,+1}\{0,+1\} representation that keeps high-magnitude weights as +1+1 and maps the rest to $0$ (Lin et al., 2021). The core optimization is

arg⁔max⁔bwˉ(bwˉ)T∣w∣∄bwĖ‰āˆ„2ā€‰āˆ„āˆ£w∣∄2,s.t.bwĖ‰āˆˆ{0,+1}n,\mathop{\arg\max}_{\bar{\mathbf{b}_w}} \frac{(\bar{\mathbf{b}_w})^T |\mathbf{w}|}{\|\bar{\mathbf{b}_w}\|_2 \, \||\mathbf{w}|\|_2}, \qquad \text{s.t.}\quad \bar{\mathbf{b}_w}\in\{0,+1\}^n,

so the alignment target is ∣w∣|\mathbf w|, not w\mathbf w.

The paper states that the discrete problem has an analytical global solution in O(nlog⁔n)\mathcal O(n\log n) time: sort the magnitudes, assign +1+1 to the largest kk entries, and assign $0$ to the rest. The final deployment rule is the half-half version,

bw~={+1,top half of sorted ∣w∣,0,otherwise.\tilde{\mathbf{b}_w}= \begin{cases} +1, & \text{top half of sorted } |\mathbf{w}|,\\ 0, & \text{otherwise}. \end{cases}

This is the paper’s explicit ā€œpositive magnitudeā€ interpretation: binarization retains the strongest weights by absolute value, irrespective of their original sign (Lin et al., 2021).

A second contribution concerns the distribution of learned weights. The paper argues that BNN weights roughly follow a Laplacian distribution under standard training with ā„“2\ell_2 regularization, leading to a fraction of selected +1+1 weights

p+1ā‰ˆ0.37.p_{+1}\approx 0.37.

The reported empirical range is about 36%∼38%36\%\sim 38\%, contrasted with a Gaussian-based expectation of about $0.54$. Removing ā„“2\ell_2 regularization shifts the +1+1 fraction to around 0.50∼0.520.50\sim 0.52, reduces the angle gap from about 30.66∘30.66^\circ–36.35∘36.35^\circ to about 7.97∘7.97^\circ–8.05∘8.05^\circ, and lowers the complexity from O(nlog⁔n)\mathcal O(n\log n) to O(n)\mathcal O(n) because only the median of ∣w∣|\mathbf w| is needed (Lin et al., 2021).

The reported experiments place this rule in a performance-oriented context. On CIFAR-10, SiMaN reaches 92.5%92.5\% top-1 on ResNet-18 and 92.5%92.5\% on VGG-small. On ImageNet, it reports 60.1%60.1\% top-1 and 82.3%82.3\% top-5 on ResNet-18, and 63.9%63.9\% top-1 and 84.8%84.8\% top-5 on ResNet-34 (Lin et al., 2021). A plausible implication is that, in this line of work, positivity is not a sign constraint but the occupancy of the +1+1 state by large-magnitude coefficients.

3. Positive-instance recognition by feature magnitude

In weakly supervised video anomaly detection, ā€œpositive magnitudeā€ refers to the use of large feature norms as evidence for positive instances, namely anomalous snippets inside abnormal videos. The paper ā€œWeakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learningā€ formulates the task as MIL, with each video represented as a bag of snippets and only the video-level label observed (Tian et al., 2021). Because abnormal videos contain many normal snippets, the authors replace pure score-based instance selection with feature-magnitude selection.

The top-kk score of a video is defined as

gĪø,k(X)=max⁔Ωk(X)āŠ†{xt}t=1T1kāˆ‘xt∈Ωk(X)∄xt∄2,g_{\theta,k}(\mathbf{X}) = \max_{\Omega_k(\mathbf{X}) \subseteq \{ \mathbf{x}_t\}_{t=1}^T} \frac{1}{k}\sum_{\mathbf{x}_t \in \Omega_k(\mathbf{X})} \|\mathbf{x}_t\|_2,

where Ωk(X)\Omega_k(\mathbf X) is the subset of kk snippets with the largest ℓ2\ell_2-norms. Separability between abnormal and normal videos is

dĪø,k(X+,Xāˆ’)=gĪø,k(X+)āˆ’gĪø,k(Xāˆ’),d_{\theta,k}(\mathbf{X}^+,\mathbf{X}^-) = g_{\theta,k}(\mathbf{X}^{+}) - g_{\theta,k}(\mathbf{X}^{-}),

and the key statistical assumption is

E[∄x+∄2]≄E[∄xāˆ’āˆ„2].\mathbb{E}[\|\mathbf{x}^+\|_2] \ge \mathbb{E}[\|\mathbf{x}^-\|_2].

Under this assumption, the paper proves that for 0<k<μ0<k<\mu, expected separability increases with kk, while for finite μ\mu,

lim⁔kā†’āˆžE[DĪø,k(X+,Xāˆ’)]=0,\lim_{k\to\infty}\mathbb{E}[D_{\theta,k}(\mathbf{X}^+,\mathbf{X}^-)]=0,

so overly large kk values dilute the anomaly signal (Tian et al., 2021).

The temporal feature extractor combines Pyramid Dilated Convolutions with dilation rates $1,2,4$ and Temporal Self-Attention, with the stated purpose of capturing both local and long-range dependencies. The classifier is then trained only on the top-kk magnitude snippets. In this framework, ā€œpositive magnitudeā€ means that larger ā„“2\ell_2-norm features are treated as more likely to be positive instances.

The reported experiments are consistent with that interpretation. The paper gives best results of 97.21%97.21\% AUC on ShanghaiTech, 84.30%84.30\% AUC on UCF-Crime, 77.81%77.81\% AP on XD-Violence, and 98.6%98.6\% AUC on UCSD-Peds using I3D-RGB features. It also states that adding the feature-magnitude module raises AUC by over 7%7\% on ShanghaiTech and over 4%4\% on UCF-Crime, and reports mean top-kk magnitude $53.4$ for abnormal snippets versus $7.7$ for normal snippets on UCF-Crime, with k=3k=3 in experiments (Tian et al., 2021). This suggests a precise operational meaning: magnitude is used as a surrogate for latent positive-instance membership.

4. Positive earthquake magnitude differences

In seismology, the expression has a narrower and more literal meaning. The paper ā€œA positive answer on the existence of correlations between positive earthquake magnitude differencesā€ studies

Ī“mi=mi+1āˆ’mi,\delta m_i=m_{i+1}-m_i,

and then restricts attention to the positive part,

qi=mi+1āˆ’mi>0,q_i=m_{i+1}-m_i>0,

that is, cases in which the later earthquake is larger than the previous one (Lippiello et al., 2024). The motivation comes from the observation by Van der Elst that, under appropriate conditions, positive magnitude differences are much less affected by missed small earthquakes than raw magnitudes or signed differences.

The paper analyzes relocated Southern California and Northern California catalogs. The Southern California catalog covers January 1981 to March 2022 and contains N=800499N=800499 earthquakes with mā‰„āˆ’1m\ge -1; the Northern California catalog covers January 1984 to December 2021 and contains N=879547N=879547 earthquakes with mā‰„āˆ’0.8m\ge -0.8. The completeness magnitudes are estimated as Mc=2.6M_c=2.6 for Southern California and Mc=2.7M_c=2.7 for Northern California (Lippiello et al., 2024).

To test correlation structure, the paper compares real catalogs with reshuffled catalogs and defines

Ī”P+(Q,Y)=P(Ī“qi>Qāˆ£Ī”yi<Y)āˆ’P(Ī“qiāˆ—>Qāˆ£Ī”yi<Y).\Delta P_+(Q,Y)= P(\delta q_i>Q \vert \Delta y_i<Y)-P(\delta q^*_i> Q \vert \Delta y_i<Y).

Under the null hypothesis of i.i.d. magnitudes following a Gutenberg-Richter law, the authors argue that Ī”P+\Delta P_+ should be small and should vanish when positive differences are well above the detection threshold. Empirically, however, they find the opposite trend: as the threshold on positive differences increases, E[Ī”P+(Q,Y)]E[\Delta P_+(Q,Y)] increases rather than decreases, and for Ī“Mth≄0.5\delta M_{\text{th}}\ge 0.5, above the estimated completeness threshold for differences Ī“Mcā‰ƒ0.4\delta M_c\simeq 0.4, the correlations remain strong (Lippiello et al., 2024).

The paper further reports similar results for both California catalogs, stronger correlations when events are closer in time, and only weak dependence on spatial separation once proximity is imposed. It interprets the findings as consistent with a time-dependent bb-value in the Gutenberg-Richter law and suggests that ETAS-type models may need explicit magnitude-memory effects (Lippiello et al., 2024). Here, ā€œpositive magnitudeā€ is not a forecast label or a representation rule, but a catalog statistic chosen for robustness to incompleteness.

5. Magnitude as an invariant of positive definite metric spaces

A different tradition uses ā€œmagnitudeā€ in Leinster’s sense: a numerical invariant of metric spaces. The paper ā€œPositive definite metric spacesā€ develops the theory for spaces whose similarity kernel

ζA(x,y)=eāˆ’d(x,y)\zeta_A(x,y)=e^{-d(x,y)}

is positive definite (Meckes, 2010). For a finite metric space AA, a weighting is a vector ww satisfying ζAw=1\zeta_A w=\mathbf 1, and the magnitude is

∣A∣=āˆ‘x∈Aw(x).|A|=\sum_{x\in A} w(x).

If AA is positive definite, then ζA\zeta_A is invertible, the weighting is unique,

w=ζAāˆ’11,w=\zeta_A^{-1}\mathbf 1,

and the magnitude is positive (Meckes, 2010).

The paper emphasizes the structural advantages of positive definiteness. For finite spaces, magnitude is always defined and monotone under inclusion: if āˆ…ā‰ BāŠ†A\emptyset\neq B\subseteq A, then 0<∣Bāˆ£ā‰¤āˆ£A∣0<|B|\le |A|. For compact positive definite spaces, the proposed definitions of magnitude coincide, including the supremum over finite subspaces, the approximation-by-finite-subsets definition, the weight-measure definition, and the variational Rayleigh-quotient formula

∣A∣=sup⁔{μ(A)2ZA(μ,μ) |ā€‰Ī¼āˆˆM(A),Ā ZA(μ,μ)≠0},|A|=\sup\left\{\frac{\mu(A)^2}{Z_A(\mu,\mu)} \,\middle|\, \mu\in M(A),\ Z_A(\mu,\mu)\neq 0\right\},

where

ZA(μ,ν)=∫A∫Aeāˆ’d(x,y) dμ(x) dν(y).Z_A(\mu,\nu)=\int_A\int_A e^{-d(x,y)}\,d\mu(x)\,d\nu(y).

Magnitude is lower semicontinuous with respect to Gromov–Hausdorff distance on compact positive definite metric spaces, while maximum diversity is continuous (Meckes, 2010).

This framework connects magnitude to negative type. The same paper proves that stable positive definiteness is equivalent to negative type, and lists many classes of examples, including LpL_p for 0<p≤20<p\le 2, ultrametric spaces, round spheres, real and complex hyperbolic spaces, weighted trees, and spaces with at most four points (Meckes, 2010). In this domain, positivity is a property of the similarity kernel, and positive magnitude means that the invariant itself is forced to be well defined and positive by that kernel structure.

6. Geometric and continuity refinements of magnitude

Recent work gives geometric formulas for magnitude and sharp continuity conditions. The paper ā€œGeometric interpretation of magnitudeā€ studies a positive definite symmetric matrix ZZ with Zii=1Z_{ii}=1 and realizes it as a Gram matrix Z=VtVZ=V^tV of unit vectors v1,…,vn∈Rnv_1,\dots,v_n\in\mathbb R^n. If RR is the radius of the circumsphere of VZ={v1,…,vn}V_Z=\{v_1,\dots,v_n\}, then

Mag⁔(Z)=11āˆ’R2.\operatorname{Mag}(Z)=\frac{1}{1-R^2}.

The same paper states that a positive weighting exists if and only if the center of the circumsphere lies in the interior of the convex hull of the viv_i. For an nn-point metric space XX of negative type, it proves

lim⁔t→0Mag⁔(ZtX)=1\lim_{t\to 0}\operatorname{Mag}(Z_{tX})=1

and

Mag⁔(ZX)<n(n>1),\operatorname{Mag}(Z_X)<n \qquad (n>1),

thereby giving a negative answer to the Gomi–Meckes problem (Asao et al., 30 Oct 2025).

The paper ā€œConvergence of Magnitude of Finite Positive Definite Metric Spacesā€ reformulates the same invariant through a similarity embedding Y={y1,…,yk}āŠ‚RkY=\{y_1,\dots,y_k\}\subset\mathbb R^k with Gram matrix ζX\zeta_X, yielding

∣X∣=11āˆ’ĻY2,|X|=\frac{1}{1-\rho_Y^2},

where ρY\rho_Y is the circumradius of YY. It then classifies when magnitude converges under Gromov–Hausdorff convergence by a clustering type r=⟨r1,…,rm⟩r=\langle r_1,\dots,r_m\rangle. The main theorem states that lim⁔nā†’āˆžāˆ£Xn∣=∣X∣\lim_{n\to\infty}|X_n|=|X| holds if and only if

∄r∄1≤2\|r\|_1\le 2

for any kk, or

∄r∄1=3 and k=1.\|r\|_1=3 \text{ and } k=1.

Outside these cases, the paper constructs counterexamples in which Hausdorff convergence of similarity embeddings does not imply convergence of circumradii (So, 13 Nov 2025).

A complementary continuity program is developed in ā€œTractable Metric Spaces and the Continuity of Magnitude.ā€ A tractable metric space is defined by three conditions: positive definiteness, the Heine–Borel property, and finite magnitude for every closed ball. In this setting, continuity of magnitude on compact subspaces is equivalent to uniform continuity on bounded subspaces, and on intervals [a,b]āŠ‚R[a,b]\subset\mathbb R the magnitude map is Lipschitz with constant

1+bāˆ’a2.1+\frac{b-a}{2}.

The paper concludes that magnitude is continuous on +(R)+(\mathbb R), with bounded restrictions Lipschitz (KaliŔnik et al., 26 Jun 2025).

For odd-dimensional Euclidean balls, potential-function methods give explicit determinant formulas. ā€œOn the magnitude of odd balls via potential functionsā€ uses reverse Bessel polynomials and a Hankel linear system to derive a determinant expression for ∣BR2p+1∣|B_R^{2p+1}|, and reports conjectured Hankel determinant formulas for both the magnitude and its derivative (Willerton, 2018). This suggests that, within positive definite settings, magnitude often admits rigid geometric or algebraic normal forms.

7. Positivity programs adjacent to ā€œpositive magnitudeā€

Several adjacent literatures study positivity of magnitude-bearing quantities even when the phrase itself is not used in the same way. In operator theory, ā€œPositivity of ∣p∣a∣q∣b+∣q∣b∣p∣a|p|^a|q|^b+|q|^b|p|^aā€ proves that

Ja,b,n:=12(∣p∣a∣q∣b+∣q∣b∣p∣a)≄0J_{a,b,n}:=\frac12\bigl(|p|^a|q|^b+|q|^b|p|^a\bigr)\ge 0

as a quadratic form on C0āˆž(Rn)C_0^\infty(\mathbb R^n) under the conditions n>a+bn>a+b and min⁔{a,b}∈[0,2]\min\{a,b\}\in[0,2]. The same paper derives the sharper bound

Ja,b,n≄La,b,n∣q∣bāˆ’a,J_{a,b,n}\ge L_{a,b,n}|q|^{b-a},

connecting the result to generalized Hardy inequalities for fractional Laplacians (Chen et al., 2013).

In EFT, ā€œHigher-Point Positivityā€ studies shift-symmetric P(X)P(X) theories with

L=āˆ’12X+āˆ‘i=2āˆžĪ»iXi,X=(āˆ‚Ļ•)2,\mathcal L=-\frac12 X+\sum_{i=2}^\infty \lambda_i X^i, \qquad X=(\partial\phi)^2,

and proves a parity-dependent sign rule for the first relevant higher operator: λn>0 for even n,λn<0 for odd n\lambda_n>0 \text{ for even } n,\qquad \lambda_n<0 \text{ for odd } n in mostly-plus signature. The result is derived from causality, analyticity of scattering amplitudes, and unitarity of the spectral representation (Chandrasekaran et al., 2018).

In perturbative QCD, ā€œOn the positivity of MSbar distributionsā€ shows that MS‾\overline{\mathrm{MS}} PDFs inherit positivity from physical PDFs above the perturbative bound

Q2≳5Ā GeV2.Q^2\gtrsim 5~\mathrm{GeV}^2.

The argument uses the perturbative scheme transformation

f(x,Q2)=[C(αs(Q2))]āˆ’1āŠ—fPHYS(Q2),f(x,Q^2)=\left[C(\alpha_s(Q^2))\right]^{-1}\otimes f^{\rm PHYS}(Q^2),

and the same lower bound is stated to apply to longitudinally polarized PDFs (Hekhorn, 2024).

In amplitude theory, ā€œPositivity properties of scattering amplitudesā€ promotes ordinary non-negativity to complete monotonicity,

(āˆ’āˆ‚x)nf(x)≄0,(-\partial_x)^n f(x)\ge 0,

and its multivariable generalizations. The paper argues that many planar and non-planar Feynman integrals with a Euclidean region, certain Euler integrals, and several observables in planar maximally supersymmetric Yang–Mills theory satisfy this stronger positivity hierarchy. It also states that the QCD and QED cusp anomalous dimensions exhibit the same property to three and four loops, respectively (Henn et al., 2024).

Taken together, these works indicate that the wider research ecosystem around ā€œpositive magnitudeā€ is organized less by a single definition than by recurring positivity mechanisms: kernel positivity, magnitude ordering, positive differences, positive weightings, and complete monotonicity. A plausible implication is that ā€œmagnitudeā€ functions as a bridge term between statistical amplitude, discrete salience, metric size, and formal positivity constraints, with each field fixing its own operational meaning.

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Positive Magnitude.