Temporal Difficulty Diagnostic \(D_T(\omega)\) ni kiashiria kisicho na vipimo kilichopendekezwa ambacho, katika mifumo teule ya Hamilton iliyo karibu na integrable, hulinganisha ukubwa wa nonlinear coupling uliotatuliwa kwa frequency katika dirisha fulani la uchunguzi lenye ukomo dhidi ya kikadiriaji cha finite-time redistribution au decorrelation kinachohitaji kukalibishwa tofauti. Kazi ya Tamás Nagy inawasilisha ufafanuzi wa kihisabati wa muundo huu, log-frequency derivative yake, na jinsi inavyopaswa kupimwa baadaye kwa namna inayoweza kufalsifiwa dhidi ya chaos diagnostic huru. Hata hivyo, jaribio la uchunguzi la trajectories 36 lililomo katika kazi halionyeshi kuwa pendekezo limethibitishwa. Kwa D4, iliyoripotiwa kama proxy bora, AUC ilikuwa 0,51029, bootstrap %95 confidence interval 0,30864–0,71193, na one-sided permutation test p=0,474. Tofauti ya AUC dhidi ya baseline rahisi ya initial-coordinate \(-\rho\) ilikuwa 0,05350 tu, na paired score-swap test ilitoa p=0,401. Kwa hiyo kazi inaweka kikomo cha wazi kwa hitimisho lake kama “hakuna ushahidi wa predictive value”. Diagnostic iliyopendekezwa si theorem ya KAM au Arnold diffusion, si criterion ya ulimwengu wote ya chaos, wala classifier iliyothibitishwa.

Katika mienendo ya Hamilton, tabia za mpangilio na za kikhaotiki zinaweza kujitokeza katika vipimo vya muda vilivyo tofauti sana. Trajectory inaweza kuonekana ya mpangilio kwa muda mfupi lakini ikaonyesha resonance, mabadiliko ya polepole ya frequency au transport katika muda mrefu zaidi. Kwa sababu hiyo, mbinu zilizopo kama frequency-map analysis, SALI, mbinu za Lyapunov, na time-frequency analysis hutumiwa kwa chaos na tabia ya muda mrefu.

Kazi hii haipendekezi kiwango kipya cha kuchukua nafasi ya mbinu hizo. Swali la utafiti ni finyu zaidi: je, diagnostic inayounganisha ukubwa wa frequency wa nonlinear coupling na finite-time redistribution/decorrelation scale katika kiasi kimoja kilichonormalishwa inaweza kutoa taarifa ya ziada ya utabiri juu ya chaos indicators huru na zilizoimarika?

Jibu la sasa la chanzo bado si “ndiyo”. Pilot ya trajectories 36 inaonyesha tu kwamba mbinu hiyo ni kitu cha utafiti kinachoweza kuhesabiwa na kuzalishwa upya. Thamani ya kisayansi ya pendekezo inategemea kujaribiwa baadaye kwa protocol iliyofungwa dhidi ya chaos labels huru katika holdout trajectories.

Temporal Difficulty Diagnostic ni nini?

Temporal Difficulty Diagnostic ni pendekezo la finite-window diagnostic lisilo na vipimo, linaloundwa kwa kugawanya ukubwa wa frequency-domain wa nonlinear coupling information, unaotokana na derivatives za pili za perturbation Hamiltonian katika trajectory, kwa finite-time redistribution/decorrelation scale iliyofafanuliwa kando.

Chanzo kwanza kinahitaji kuwekwa kwa nondimensional canonical coordinates kabla ya calibration:

\[ \tilde z=(\tilde q,\tilde p) \]

Coordinate scales \(Q_i\) na \(P_i\) huchaguliwa kwa sharti la common action scale:

\[ Q_iP_i=A_\* \]

Perturbation Hamiltonian pia hunormalishwa kwa energy scale:

\[ \tilde H_1=\frac{H_1}{E_\*} \]

Kitengo cha muda huchukuliwa kuwa \(A_\*/E_\*\).

Hatua hii si undani mdogo wa kiufundi. Chanzo kinasema wazi kwamba diagnostic inategemea uchaguzi wa coordinate na scale. Nondimensionalization tofauti inaweza kutoa thamani tofauti za nambari kwa mfumo uleule, na recalibration huhitajika.

Diagnostic ya msingi inafafanuliwaje?

Chini ya finite observation window \(T\), kwa kuchagua positive reference scales \(C_\*\) na \(\lambda_\*\):

\[ D_T(\omega) = \frac{ \|C_T(\omega)\|_F/C_\* }{ \lambda_{\mathrm{eff},T}(\omega)/\lambda_\* }, \qquad \lambda_{\mathrm{eff},T}(\omega)>0 \]

hufafanuliwa.

Hapa \(C_T(\omega)\) ni frequency-resolved statistic ya nonlinear coupling structure iliyopimwa kando ya trajectory.

\(\lambda_{\mathrm{eff},T}\) haijafafanuliwa kwenye chanzo kama physical damping coefficient. Kiasi hiki kinaachwa kama estimator inayopaswa kupendekezwa kwa finite-time redistribution au decorrelation.

Kwa hiyo ukubwa wa nambari wa \(D_T\) unaweza kufasiriwa tu baada ya ufafanuzi wa \(C_\*\), \(\lambda_\*\), na \(\lambda_{\mathrm{eff},T}\) kufungwa.

Je, \(D_T=1\) ni kizingiti cha chaos cha ulimwengu wote?

Hapana. Chanzo kinasema wazi kwamba thamani ya \(D_T=1\) haina invariant physical meaning inayojitegemea na calibration. Kizingiti kinaweza kufitishwa kwenye training data tu chini ya coordinate system, trajectory ensemble, observation window, normalization, na independent validation diagnostic maalum; hakiwezi kuhamishwa kiotomatiki kwenda mfumo mwingine.

Tofauti hii inaifanya pendekezo kuwa finyu na la tahadhari zaidi kisayansi kuliko wazo la “candidate unit threshold” katika unpublished research note ya awali.

Nonlinear Coupling inapimwaje katika frequency domain?

Near-integrable representation huandikwa kama:

\[ \dot{\tilde z} = J\nabla_{\tilde z}\tilde H(\tilde z,t) \]

na

\[ \tilde H(\tilde z,t) = \tilde H_0(\tilde z) + \epsilon\tilde H_1(\tilde z,t) \]

.

Kwa trajectory fulani ya \(\tilde z(t)\), kila element ya coupling tensor inafafanuliwa kama:

\[ C_{T,jk}(\omega) = \frac{1}{T} \int_0^T \frac{ \partial^2\tilde H_1 }{ \partial\tilde z_j\partial\tilde z_k } (\tilde z(t),t) e^{-i\omega t}dt \]

.

Kwa maneno mengine, model hugawa mabadiliko ya Hessian ya perturbation Hamiltonian kando ya trajectory katika Fourier frequencies.

Kwa sababu elements zote za Hessian hazina vipimo katika nondimensionalized coordinates, Frobenius norm:

\[ \|C_T(\omega)\|_F \]

inaweza kutumika kama scalar summary ya ukubwa wa coupling.

Normalized components

Ili kurahisisha ufafanuzi, chanzo kinafafanua:

\[ A(\omega) = \frac{\|C_T(\omega)\|_F}{C_\*} \]

na

\[ L(\omega) = \frac{\lambda_{\mathrm{eff},T}(\omega)}{\lambda_\*} \]

.

Hivyo:

\[ D_T(\omega)=\frac{A(\omega)}{L(\omega)} \]

.

Log-frequency derivative inaonyesha nini?

Mojawapo ya michango ya kihisabati ya chanzo ni kutoa wazi correct quotient derivative expression ya diagnostic katika logarithmic frequency axis:

\[ \beta_{D_T}(\omega) = \frac{dD_T}{d\log\omega} = \omega \left[ \frac{A'(\omega)}{L(\omega)} - \frac{A(\omega)L'(\omega)}{L(\omega)^2} \right] \]

Hapa:

\[ A'(\omega) = \frac{ \operatorname{Re} \langle C_T(\omega), C'_T(\omega) \rangle_F }{ C_\*\|C_T(\omega)\|_F } \]

hutolewa.

Expression hii hutumika tu chini ya masharti:

• \(\omega>0\),
• \(A\) na \(L\) differentiable,
• \(C_T(\omega)\neq0\),
• \(L(\omega)>0\)

.

Je, \(\beta_{D_T}=0\) inamaanisha chaotic fixed point?

Hapana. \(\beta_{D_T}(\omega)=0\) inaonyesha tu kwamba calibrated diagnostic ni stationary kwenye frequency husika. Chanzo kinasema wazi kwamba terminology ya “ultraviolet” au “infrared fixed point” inaweza kutumiwa kama analogy tu, na stability, transport, au diffusion haiwezi kuhitimishwa kutokana na sifuri hii peke yake.

Kwa nini Coordinate dependence ni muhimu?

Diagnostic inategemea uchaguzi wa canonical coordinate. Mfumo uleule wa kimwili unaweza kutoa \(C_T\), normalization na hivyo \(D_T\) tofauti chini ya canonical nondimensionalization tofauti.

Kwa hiyo chanzo hakidai kuwa ni coordinate-invariant chaos diagnostic.

Application lazima ifunge kwa uwazi tangu mwanzo:

• canonical coordinate system,
• scales,
• trajectory ensemble,
• observation window \(T\),
• \(C_\*\),
• \(\lambda_\*\),
• ufafanuzi wa \(\lambda_{\mathrm{eff},T}\).

Mbinu na Matokeo ya Utafiti

Jaribio la uchunguzi la trajectories 36

Katika numerical demonstration:

• 12 tofauti za initial separation \(\rho=r_{12}\),
• 3 tofauti za initial angle \(\theta\),
• jumla ya trajectories 36

zilitumika.

Trajectories zote zilikokotolewa kwa:

• 300 leapfrog steps,
• \(\Delta t=5\times10^{-4}\),
• physical window ileile \(T=0,15\)

.

Modal structure iliundwaje?

Katika initial point, kwa kutumia diagonal mass matrix \(M\) husika, generalized Hessian:

\[ U^\top M^{-1/2} V'' M^{-1/2} U = \operatorname{diag}(\kappa_j) \]

ili-diagonalize.

Spectrum ilikuwa indefinite katika initial conditions zote 36. Chanzo kinaripoti:

• negative eigenvalues mbili katika raw-coordinate Hessian,
• negative eigenvalue moja katika Levi–Civita Hessian

.

Kwa hiyo, positive spectral subspace pekee ilitumika katika action calculation.

Frequency inafafanuliwa kama:

\[ \omega_j=\sqrt{\kappa_j} \]

; implementation pia hutumia:

\[ \omega_j>0,01 \]

retention rule.

Negative, zero, na modes zilizo chini ya kikomo hiki hazifasiriwi kama real oscillatory frequency.

Canonical modal action

Chanzo kinafafanua modal coordinates kama:

\[ Q= U^\top M^{1/2}(q-q_0) \]

\[ P= U^\top M^{-1/2}p \]

.

Kwa kila retained mode, action hukokotolewa kama:

\[ I_j(t) = \frac{ P_j(t)^2+\omega_j^2Q_j(t)^2 }{ 2\omega_j } \]

.

Je, label iliyotumiwa katika trajectories 36 ilikuwa chaos label halisi?

Hapana. Positive/negative classification iliyotumiwa katika jaribio haikutoka kwenye independent KAM, SALI, Lyapunov, au invariant-torus validation. Chanzo huiita tu “spectator-advantage label” na kinasema wazi kwamba si chaos ground truth.

Lowest temporal standard deviation ya raw-coordinate positive modes:

\[ \sigma_{\mathrm{raw,best}} \]

na temporal standard deviation ya lowest-positive-frequency Levi–Civita spectator mode:

\[ \sigma_{\mathrm{spectator}} \]

zilitumiwa kufafanua:

\[ W= \log \left( \frac{ \sigma_{\mathrm{raw,best}} }{ \sigma_{\mathrm{spectator}} } \right) \]

.

Positive class ilichukuliwa kama:

\[ W>0 \]

.

Kati ya trajectories 36:

• 9 zilipata label positive,
• 27 zilipata label non-positive

.

Exploratory proxy results

Je, D4 ilikuwa predictor yenye mafanikio katika pilot?

Hapana. Ingawa D4 ilikuwa na AUC ya juu zaidi kwenye jedwali, AUC ilikuwa 0,51029 tu. Stratified 10.000-draw bootstrap %95 confidence interval ilikuwa 0,30864–0,71193 na one-sided 10.000-permutation test ilitoa p=0,474. Chanzo hakikubali matokeo haya kama ushahidi wa predictive value.

AUC ya baseline rahisi ya \(-\rho\) ilikuwa 0,45679.

Tofauti ya AUC ya D4 dhidi ya baseline ilikuwa:

\[ 0,51029-0,45679=0,05350 \]

lakini paired score-swap test ilitoa:

\[ p=0,401 \]

.

Kwa hiyo, katika sampuli ya sasa haiwezi kuhitimishwa kwamba D4 ina predictive value zaidi ya baseline rahisi inayotokana na initial coordinate.

D4 inafafanuliwaje?

Chanzo hutumia ufafanuzi:

\[ g=\max(1-\eta,0,01) \]

.

Sweep-rate proxy ni:

\[ s= sd \left( \frac{ \eta(t_{i+1})-\eta(t_i) }{ t_{i+1}-t_i } \right) \]

.

D4 inafafanuliwa kama:

\[ D_4=\frac{s}{g^2} \]

, huku implementation ikitumia denominator:

\[ \max(g^2,10^{-4}) \]

.

Kwa nini Static Quartic Coupling Proxy haikukubaliwa kuwa ya kuaminika?

Chanzo kinaonyesha kwamba static quartic coupling proxy iliyofafanuliwa kama \(\sigma_k^\*\) haikukonverge vya kutosha dhidi ya finite-difference steps zilizotumiwa. Kwa \(\rho=0,40\) na aligned initial angle, steps mbili za differentiation zilipopunguzwa nusu, thamani ilipanda kutoka 1,97191 hadi 8,21449, ikiwa na relative change 3,16576.

Kwa sababu ya sensitivity hii kubwa, matokeo yanayotumia \(\sigma_k^\*\) yanachukuliwa kuwa exploratory tu.

Chanzo pia kinabainisha hasa kwamba kiashiria kikuu cha D4 hakitumii \(\sigma_k^\*\).

Je, \(\rho=0,45\) ni dynamic transition point?

Hapana. Thamani 0,25, 0,45, na 0,65 katika code ni mipaka iliyo-hard-coded mapema kwa madhumuni ya kuunda descriptive regions tu. \(\rho=0,45\) si mixing boundary au dynamic transition iliyogunduliwa kutoka kwenye jaribio, na hakuna hitimisho la kazi linalotegemea uainishaji huu wa maeneo.

Verification manifest inathibitisha nini?

Supplementary manifest inaorodhesha implication checks kumi za msingi kuhusu nonnegativity, monotonicity, na baadhi ya conditional quotient/AUC relationships.

Lakini manifest:

• haihesabu wala kuthibitisha AUC results,
• haithibitishi Hamiltonian regularity,
• haithibitishi KAM-torus persistence,
• haionyeshi Arnold diffusion,
• haithibitishi proxy performance,
• haitoi ushahidi wa universality.

Kwa hiyo machine-readable receipt inapaswa kutathminiwa kama numerical reproducibility artifact tu.

Diagnostic iliyopendekezwa inapaswa kuthibitishwaje baadaye?

Chanzo kinapendekeza protocol ya hatua sita, na ikiwezekana iliyosajiliwa mapema, kwa validation halisi:

1. Funga mapema canonical nondimensionalization, trajectory ensemble, observation window, \(C_\*\), \(\lambda_\*\), na ufafanuzi wa \(\lambda_{\mathrm{eff},T}\).
2. Freeze proxy selection na regularization constants kabla ya evaluation.
3. Fit threshold yoyote kwenye training ensemble pekee.
4. Linganisha katika held-out trajectories dhidi ya independent diagnostics kama frequency-map, SALI, Lyapunov, au invariant-torus.
5. Ripoti uncertainties, permutation baselines, na coordinate/window sensitivity.
6. Jaribu protocol hiyo hiyo iliyofungwa katika Hamiltonian model nyingine.

Kigezo kikuu cha falsification cha chanzo pia kiko wazi: ikiwa mbinu iliyopendekezwa haina kuongeza held-out performance juu ya established independent diagnostics au haiwezi kuhamishwa kwenye Hamiltonian models nyingine, dai la universality linapaswa kukataliwa.

Hitimisho zinazoungwa mkono na utafiti

• Chini ya uchaguzi maalum wa canonical na calibration, \(D_T(\omega)\) inaweza kufafanuliwa kihisabati.
• Log-frequency derivative ya diagnostic inaweza kutolewa kwa explicit quotient rule.
• Numerical run ya trajectories 36 inatoa exploratory receipt inayoweza kuzalishwa upya.
• D4 ilitoa AUC ya juu zaidi kati ya proxies zilizojaribiwa.
• D4 AUC ni 0,51029 na uncertainty interval inajumuisha chance performance.
• Ubora wa D4 juu ya \(-\rho\) baseline haukuungwa mkono kitakwimu.
• Static quartic coupling proxy haikukonverge numerically chini ya finite-difference scales zilizotumiwa.
• Validation halisi ya baadaye inahitaji independent chaos diagnostics na held-out trajectories.

Hitimisho ambazo utafiti hauungi mkono

• Haijaonyeshwa kwamba Temporal Difficulty Diagnostic ni chaos classifier iliyothibitishwa.
• Haijaonyeshwa kwamba D4 inatenganisha chaotic trajectories kwa uaminifu.
• \(W>0\) si independent chaos ground truth.
• \(D_T=1\) si universal critical threshold.
• \(\beta_{D_T}=0\) pekee si ushahidi wa physical fixed point.
• \(\rho=0,45\) si mixing transition iliyogunduliwa.
• Haijaonyeshwa kwamba pendekezo ni coordinate-invariant.
• Hakuna theorem ya KAM au Arnold-diffusion iliyotolewa.
• Jaribio la trajectories 36 si ushahidi wa generalizability kwenda Hamiltonian systems nyingine.

Vikwazo vikuu

\(\lambda_{\mathrm{eff},T}\) bado haijafungwa operationally.

Sampuli ina trajectories 36 tu.

Jaribio ni in-sample; hakuna held-out test.

Exploratory label iliyotumiwa si independent chaos ground truth.

Diagnostic inategemea coordinate, trajectory, time window, na regularization choices.

Static quartic proxy inaonyesha sensitivity kubwa kwa finite-difference scale.

Kwa hiyo kazi inaweka kikomo cha matokeo yake kama “testable research proposal”.

Kwa nini matokeo hasi ni muhimu kisayansi?

Jambo la kutia maanani katika kazi ni kwamba haifichi kwamba kiashiria kilichopendekezwa hakikuonyesha mafanikio yaliyotarajiwa katika pilot ndogo. Badala ya kureframe AUC≈0,51 kama kipimo kipya cha mafanikio, chanzo kinasema wazi kwamba hakuna ushahidi wa predictive value.

Mtazamo huu hutoa sehemu muhimu ya kuanzia inayozuia mbinu kurekebishwa mara kwa mara kwenye data ileile na kutoa mafanikio bandia. Kazi inayofuata inaweza freeze proxy, normalization, na validation conditions mapema kisha kuzijaribu kwenye independent holdout.

Maelezo ya Chanzo na Mbinu

Kichwa asili: A Proposed Temporal Difficulty Diagnostic for Hamiltonian Systems

Mwandishi: Tamás Nagy, Ph.D.

ORCID: 0009-0004-8079-4679

Aina ya chanzo: Pendekezo la theoretical/mathematical diagnostic kwa Hamiltonian dynamics na small-scale exploratory numerical study.

Jarida / volume / issue: Haijatajwa kwenye PDF.

DOI: Haijatajwa kwenye PDF kwa kazi hii.

Hali ya peer review: Hakuna taarifa ndani ya PDF inayothibitisha kuwa ni peer-reviewed journal publication.

Pendekezo kuu: \(D_T(\omega)\), inayounganisha frequency-resolved nonlinear coupling magnitude na finite-time redistribution/decorrelation estimator.

Coupling object kuu: Windowed temporal Fourier statistic ya perturbation Hamiltonian Hessian kando ya trajectory, \(C_T(\omega)\).

Log-frequency derivative: \(\beta_{D_T}(\omega)=dD_T/d\log\omega\).

Exploratory sample: trajectories 36; values 12 za \(\rho\) × initial angles 3.

Numerical integration: 300 leapfrog step; \(\Delta t=5\times10^{-4}\); \(T=0,15\).

Exploratory positive label: \(W>0\); samples 9 positive na 27 non-positive. Label hii si independent chaos ground truth.

Proxy kuu: \(D_4=s/g^2\).

D4 AUC: 0,51029.

D4 bootstrap %95 interval: 0,30864–0,71193.

D4 permutation test: p=0,474.

Baseline: \(-\rho\), AUC=0,45679.

AUC difference: 0,05350; paired score-swap p=0,401.

Numerical convergence warning: \(\sigma_k^\*\), katika hali ya \(\rho=0,40\) aligned, ilibadilika kutoka 1,97191 hadi 8,21449 steps za differentiation zilipopunguzwa nusu.

Hard-coded descriptive boundaries: \(\rho=0,25\), 0,45, na 0,65; hizi si physical transitions zilizogunduliwa.

Numerical source SHA-256: 1e213a089bcc54acdcbbb18ae3eb2494a9a6784812226f816958a9ef552ab76e.

Receipt SHA-256: db11fbba9d927a7b5907b5dd04e9827b9d9f4ec290662b4be46f1ab192eb7acd.

Generative AI statement: Mwandishi anasema alitumia AI-based tools kwa manuscript drafting, literature search, symbolic computation verification, na coding assistance; alikagua na kuhariri outputs na anawajibika kwa final content.

Kikomo kikuu cha tafsiri: Kazi haitoi Hamiltonian chaos law wala validated classifier. Corrected exploratory run ya sasa haitoi ushahidi kwamba proxy iliyopendekezwa ina predictive value zaidi ya simple input-coordinate baseline.