chunk_uid
stringlengths
40
40
chunk_type
stringclasses
2 values
chunk_index
int64
0
6.71k
total_chunks
int64
1
6.71k
section_title
stringlengths
1
157
embed_text
stringlengths
1
83.3k
spans
dict
paper_doi
stringlengths
0
63
paper_id_arxiv
stringlengths
9
16
title
stringlengths
7
245
authors
listlengths
1
768
categories
listlengths
1
7
year
int64
2k
2.02k
language
stringclasses
2 values
discipline
stringclasses
8 values
dense_vector
listlengths
1.02k
1.02k
bafca359044939e94287ac03f4235c88b2ae306a
subsection
23
26
Comparison of the estimated averages of subsidies amounts under the two policies
Given that the rules for the determination of benefit amount are well known and that we had informations about employees' wages, we were able to compare the amount of subsidies given under the two laws. To do it, we used informations on wages to estimate approximativelyWhere it is approximatively because the rate of wa...
{ "cite_spans": [] }
1802.03343
Long-Term Unemployed hirings: Should targeted or untargeted policies be preferred?
[ "Alessandra Pasquini", "Marco Centra", "Guido Pellegrini" ]
[ "econ.EM", "stat.AP" ]
2,018
en
Economics
[ -0.07460261136293411, 0.030326390638947487, -0.04545142874121666, -0.013804841786623001, -0.022969912737607956, -0.047038719058036804, -0.005215925630182028, 0.06672721356153488, 0.04902283102273941, 0.03775917738676071, -0.002298898994922638, -0.003680146299302578, -0.017673859372735023, ...
a435e70dd467eab1612e90194ed00ddd8ce283f9
subsection
24
26
Comparison of the distribution of treated and controls among non-CICO-detected categories
In the following graph are represented the distributions, among different non-CICO-detected categories, of individuals with 23 and individuals with 24 months of unemployment. In order to build these distributions we used RTFL data. The last is the result of a quarterly survey conducted by ISTAT (Italian National Instit...
{ "cite_spans": [] }
1802.03343
Long-Term Unemployed hirings: Should targeted or untargeted policies be preferred?
[ "Alessandra Pasquini", "Marco Centra", "Guido Pellegrini" ]
[ "econ.EM", "stat.AP" ]
2,018
en
Economics
[ -0.055173493921756744, 0.007747328840196133, -0.04141063615679741, 0.01574641652405262, -0.020140765234827995, 0.0016726753674447536, 0.023634884506464005, 0.03442240133881569, -0.01924053393304348, 0.013526354916393757, -0.008376727811992168, -0.015960032120347023, -0.01634148508310318, 0...
683534aafa560018b45144e8d834e5b3bff39ce5
subsection
25
26
Bandwidth Changes
Note that the different bandwidths we tested for, are all smaller than the chosen bandwidth. Indeed, given the method we used in bandwidth selection, in a bandwidth bigger than the chosen ones there wouldn't be balance between treated and controls. The estimation of treatment effect is not very sensitive to bandwidth s...
{ "cite_spans": [] }
1802.03343
Long-Term Unemployed hirings: Should targeted or untargeted policies be preferred?
[ "Alessandra Pasquini", "Marco Centra", "Guido Pellegrini" ]
[ "econ.EM", "stat.AP" ]
2,018
en
Economics
[ -0.0483812540769577, 0.027288857847452164, -0.05637865886092186, -0.003317626193165779, -0.021336590871214867, -0.049876950681209564, 0.02409905381500721, 0.04719080030918121, 0.027609365060925484, 0.023839594796299934, -0.0019392564427107573, 0.005479138810187578, -0.0064025032334029675, ...
7b56df4224eec0dfad0f5675dadeec0f33b1308a
abstract
0
6
Abstract
Three-dimensional, compressible, magnetohydrodynamic turbulence of an isothermal, self-gravitating fluid is analyzed using two-point statistics in the asymptotic limit of large Reynolds numbers (both kinetic and magnetic). Following an alternative formulation proposed by S. Banerjee and S. Galtier (Phys. Rev. E,93, 033...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ -0.022241389378905296, 0.03481128066778183, -0.008023984730243683, -0.011204595677554607, 0.002263419795781374, -0.03835037723183632, -0.01195207703858614, -0.02446857839822769, 0.031668808311223984, 0.0051827928982675076, 0.0025227500591427088, -0.008962150663137436, -0.006765470374375582, ...
3e5bf1373fa9e9c240cf652594c1cafed1969f40
subsection
1
6
Introduction
Turbulence is a non-linear phenomenon omnipresent in nature. For a fully developed turbulence in the limit of infinitely large Reynolds numbers, the fluid flow contains fluctuations, populating a wide range of length and time scales. In the inertial range sufficiently decoupled from the large-scale forcing and small-sc...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ 0.002170492894947529, 0.017806433141231537, -0.025130417197942734, 0.009742422960698605, 0.015960179269313812, -0.0376879945397377, -0.020079920068383217, 0.013145023956894875, 0.03335464000701904, 0.00939911138266325, -0.01805056631565094, -0.022200822830200195, -0.0023879236541688442, 0....
a7621c9432f70d078af92815b8df8cbd767240d0
subsection
2
6
Introduction
In contrast to the previous flux-source form, the current formulation casts the invariant flux rate \varepsilon in terms of mixed second order structure functions, associating the fluctuations of the density (\rho ), the fluid velocity ({u}), the vorticity ({\omega }), the magnetic field ({B}), the current (\nabla \tim...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ 0.00906224176287651, 0.02553904429078102, -0.009725891053676605, 0.017727820202708244, -0.010931138880550861, -0.009931851178407669, -0.0014093006029725075, 0.008223145268857479, 0.03997150436043739, 0.021328305825591087, -0.004237437155097723, -0.03374693542718887, -0.008833397179841995, ...
f5666fb22e01b00e81d307703911b406a524de57
subsection
3
6
Basic Equations
In this paper, we are interested in a self-gravitating isothermal magnetohydrodynamic fluid. The basic equations are given as\partial _t \rho + \leavevmode \nabla \cdotj= 0,t j+ \nabla \cdot \nabla \cdot (ju) = - \nabla p\nabla p + jbb+ g + dH + f,t b= ( ub) + dM,g = - 4 G (- 0 ), where {b}= {B}/\sqrt{\mu _0} wi...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ -0.00205859518609941, 0.02785494737327099, 0.0007130675949156284, -0.028862304985523224, 0.017033491283655167, -0.11911233514547348, 0.017109805718064308, -0.023459207266569138, 0.00336930388584733, 0.026633908972144127, -0.03269331529736519, -0.011355660855770111, -0.028236521407961845, -...
0a29fca090edf54b7cb12f669a69ffc2fd101e7e
subsection
4
6
Basic Equations
We have (assuming periodic boundary conditions or zero velocity on the boundary surface)\int _V {\partial _t} \left(\frac{\rho {u}^2}{2}\right) dx &= \int _V \left[ {{j}}\cdot {g}- {u}\cdot \nabla p + {u}\cdot \left( {{j_b}}\times {b}\right) \right] dx, \\ \int _V {\partial _t} \left(\frac{{b}^2}{2}\right) dx &= \int _...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ 0.0026722706388682127, 0.04965349659323692, -0.023041419684886932, -0.003467657370492816, -0.010475452989339828, -0.052033934742212296, 0.01623580791056156, 0.03019799292087555, 0.0170903243124485, -0.026108521968126297, -0.008842716924846172, -0.019455503672361374, -0.03711041808128357, -...
5bcc8a088a57af9c3c341d8deb6aef032d582c18
subsection
5
6
Derivation of the exact relation
Unlike the exact relations for compressible isothermal MHD turbulence which have been derived , , in this paper we propose an alternative form which is an extension of in case of compressible fluids. Under this formalism, the final extact relation is simpler and hence for a physical system, the calculation of the energ...
{ "cite_spans": [] }
10.1103/PhysRevE.97.023107
1802.03491
Energy transfer in compressible magnetohydrodynamic turbulence for isothermal self-gravitating fluids
[ "Supratik Banerjee", "Alexei G. Kritsuk" ]
[ "physics.flu-dyn", "astro-ph.SR" ]
2,018
en
Physics
[ 0.0013819660525768995, 0.0020829634740948677, -0.014718081802129745, 0.020219242200255394, -0.0012875460088253021, -0.03479235619306564, 0.03463975712656975, 0.008682829327881336, 0.04812942445278168, 0.009316110983490944, -0.02142476662993431, -0.016465330496430397, -0.04477226734161377, ...
af7b1879853993359736b36bd996230520eeb3cd
abstract
0
45
Abstract
Every student in statistics or data science learns early on that when the sample size largely exceeds the number of variables, fitting a logistic model produces estimates that are approximately unbiased. Every student also learns that there are formulas to predict the variability of these estimates which are used for t...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.047400619834661484, -0.011087105609476566, -0.027637643739581108, 0.018984666094183922, 0.008660608902573586, -0.012086699716746807, -0.0057152388617396355, 0.01684812642633915, 0.011834893375635147, 0.004852993413805962, -0.04309702292084694, 0.016603952273726463, -0.02827860414981842, ...
c8a2ccd7996feafb811f81149db544fd5284474f
subsection
1
45
Logistic regression: classical theory and practice
Logistic regression , , is by and large the most frequently used model to estimate the probability of a binary response from the value of multiple features/predictor variables. It is used all the time in the social sciences, the finance industry, the medical sciences, and so on. As an example, a typical application of...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1002/9781119243830.ch12", "end": 177, "openalex_id": "https://openalex.org/W2045638068", "raw": "Erich L Lehmann and Joseph P Romano. Testing statistical hypotheses. Springer Science & Business Media, 2006.", "source_ref_id": "12e4...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.10356822609901428, -0.01309097558259964, -0.03341402858495712, -0.004035620950162411, 0.013289323076605797, 0.029508095234632492, 0.000424112135078758, 0.007289292756468058, 0.030515093356370926, 0.003194548888131976, -0.048305392265319824, 0.032895270735025406, -0.009146898984909058, 0...
b3981ed50350d5634204b294034c47e233b9769b
subsection
2
45
Logistic regression: classical theory and practice
The coefficient estimates\hat{\beta }_j are obtained by maximum likelihood, and for eachvariable, R provides an estimate of the standard deviation of\hat{\beta _j} as well as a p-value for testing whether\beta _j = 0 or not.]Another well-known result in logistic regression is Wilks' theorem , which gives the asymptotic...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1214/aoms/1177732360", "end": 459, "openalex_id": "https://openalex.org/W2019841176", "raw": "Samuel S Wilks. The large-sample distribution of the likelihood ratio for testing composite hypotheses. The Annals of Mathematical Statistics, ...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.08086387068033218, 0.00462297210469842, -0.03573267534375191, 0.018842807039618492, -0.01698141358792782, -0.0023782369680702686, 0.02085677534341812, -0.015654025599360466, 0.0011519286781549454, 0.011259912513196468, -0.04388009384274483, 0.023755669593811035, -0.026807134971022606, 0...
6c9e66ed99d7f3c0231561b0da0de00a6f292402
subsection
3
45
Failures in moderately large dimensions
In modern-day data analysis, new technologies now produce extremely large datasets, often with huge numbers of features on each of a comparatively small number of experimental units. Yet, software packages and practitioners continue to perform calculations as if classical theory applies and, therefore, the main issue i...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.07574830204248428, -0.016800740733742714, -0.05157720670104027, 0.018204618245363235, -0.026246389374136925, -0.008125699125230312, 0.031678780913352966, 0.03537158668041229, 0.019165968522429466, 0.010010250844061375, -0.022812996059656143, 0.026200611144304276, -0.0011721225455403328, ...
0a4904a54875f2e3c3264f477d78d25af17ce477
subsection
4
45
Unbiasedness?
Figure REF plots the true and fitted coefficients in the setting where one quarter of the regression coefficients have a magnitude equal to 10, and the rest are zero. Half of the nonzero coefficients are positive and the other half are negative. A striking feature is that the black curve does not pass through the cente...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.08421443402767181, -0.0048858099617064, -0.06865306943655014, -0.01694967970252037, 0.00219308421947062, -0.010259820148348808, 0.010938722640275955, 0.04372437670826912, 0.022518206387758255, 0.005068884696811438, -0.04878944903612137, 0.019283883273601532, -0.030497219413518906, 0.019...
d74aa765a5aefbc030f4fdd6cb8c71b3a54ee321
subsection
5
45
Unbiasedness?
This is illustrated in Figure REF (b). Observe that when f({X}_*) < 1/2, lots of predictions tend to be close to zero, even when f({X}_*) is nowhere near zero. A similar behavior is obtained by symmetry; when f({X}_*) > 1/2, we see a shrinkage toward the other end point, namely, one. Hence, we see a massive shrinkage t...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.05071193724870682, -0.007971416227519512, -0.03933073952794075, 0.014272267930209637, -0.019924726337194443, -0.038079723715782166, -0.014279896393418312, 0.05431242659687996, -0.005244352854788303, 0.014501112513244152, -0.05223757028579712, -0.00081621203571558, -0.023113293573260307, ...
434243ab7049eabaf0b76d0f8419c81c025c0cb6
subsection
6
45
Accuracy of classical standard errors?
Consider the same matrix {X} as before and regression coefficients now sampled as follows: half of the \beta _j's are i.i.d. draws from {\mathcal {N}}(7,1), and the other half vanish. Figure REF (a) shows standard errors computed via Monte Carlo of ML estimates \hat{\beta }_j corresponding to null coordinates. This is ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 980, "openalex_id": "", "raw": "Emmanuel Candès, Yingying Fan, Lucas Janson, and Jinchi Lv. Panning for gold: Model-free knockoffs for high-dimensional controlled variable selection. arXiv preprint arXiv:1610.02351, 2016.", ...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.08134810626506805, -0.0192690659314394, -0.05947016924619675, -0.014653949998319149, 0.027858523651957512, 0.0005439926171675324, -0.004714284557849169, 0.058402206748723984, 0.002975034061819315, 0.011366155929863453, -0.025844654068350792, 0.028865458443760872, -0.022076278924942017, ...
ccfe10bea07be0545861970bc23e504940ef6a0c
subsection
7
45
Distribution of the LRT?
By now, the reader should be suspicious that the chi-square approximation for the distribution of the likelihood-ratio test holds in higher dimensions. Indeed, it does not and this actually is not a new observation. In , the authors established that for a class of logistic regression models, the LRT converges weakly to...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 474, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescale...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.1004357784986496, -0.01147140096873045, -0.005773836746811867, 0.018046099692583084, 0.000613517826423049, 0.0056251054629683495, -0.004008125979453325, 0.034414201974868774, 0.0008594969986006618, 0.009251868352293968, -0.05778413265943527, 0.02259194850921631, -0.03734307363629341, 0....
b5d0ed845fa4a908af43f3f4a76ce4b990717d07
subsection
8
45
Summary.
We have hopefully made the case that classical results, which software packages continue to rely upon, are downright erroneous in higher dimensions.Estimates seem systematically biased in the sense that effect magnitudes are overestimated. Estimates are far more variable than classically predicted. Inference measures...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0728076696395874, 0.01911735348403454, -0.022473951801657677, 0.01080977264791727, -0.004546665120869875, -0.02421327866613865, 0.026273010298609734, 0.028058109804987907, 0.0018537576543167233, 0.01345690805464983, -0.03738030046224594, 0.02229086309671402, -0.041621819138526917, -0.00...
896da0353778d69ac43f50aaba2bd1355ea14bb2
subsection
9
45
Our contribution
Our contribution is to develop a brand new theory, which applies to high-dimensional logistic regression models with independent variables, and is capable of accurately describing all the phenomena we have discussed. Taking them one by one, the theory from this paper predicts:the bias of the MLE; the variability of th...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1073/pnas.1810420116", "end": 1585, "openalex_id": "https://openalex.org/W2963060833", "raw": "Pragya Sur and Emmanuel J Candès. Additional supplementary materials for: A modern maximum-likelihood theory for high-dimensional logistic reg...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.09111166000366211, -0.020016493275761604, -0.014905574731528759, 0.016705838963389397, -0.008970806375145912, -0.02187778428196907, -0.014196149073541164, -0.0028834736440330744, -0.00470662210136652, 0.010519338771700859, -0.05318406596779823, 0.013364670798182487, -0.008528368547558784,...
aff74e4ade9a18cc1046965f9806f083ffe256f4
subsection
10
45
Prior work
Asymptotic properties of M-estimators in the context of linear regression have been extensively studied in diverging dimensions starting from , followed by and . These papers investigated the consistency and asymptotic normality properties of M-estimators in a regime where p = o(n^{\alpha }), for some \alpha < 1. Late...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1214/aos/1176342503", "end": 162, "openalex_id": "https://openalex.org/W2065742895", "raw": "Peter J Huber. Robust regression: asymptotics, conjectures and Monte Carlo. The Annals of Statistics, pages 799–821, 1973.", "source_ref_i...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0689081996679306, -0.035308584570884705, -0.03628513962030411, 0.00848382618278265, 0.024780096486210823, -0.03108193166553974, 0.027786055579781532, 0.021804653108119965, 0.01329031027853489, 0.013046171516180038, -0.03747531771659851, 0.02008042298257351, -0.024551216512918472, 0.0110...
6c9cf2317ec04752c7f68a2c86077401a6abbb89
subsection
11
45
Prior work
The rule of thumb is usually 10 events per variable (EPV) or more as mentioned in , , while a later study suggested that it could be even less. As we clearly see in this paper, such a rule is not at all valid when the number of features is large. contested the previously established 10 EPV rule.To the best of our kno...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1016/s0895-4356(96)00236-3", "end": 144, "openalex_id": "https://openalex.org/W2037668591", "raw": "Peter Peduzzi, John Concato, Elizabeth Kemper, Theodore R Holford, and Alvan R Feinstein. A simulation study of the number of events per ...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0881386324763298, -0.04254337027668953, -0.002153491135686636, 0.022263549268245697, -0.023453786969184875, -0.00915567483752966, 0.0021115276031196117, 0.0017290874384343624, -0.013336766511201859, 0.0005750908167101443, -0.01490849070250988, 0.034547414630651474, -0.0331130251288414, ...
d05176f7da7cacb4c07f63c19281cae9dc7f9ae6
subsection
12
45
Setting.
We describe the asymptotic properties of the MLE and the LRT in a high-dimensional regime, where n and p both go to infinity in such a way that p/n \rightarrow \kappa . We work with independent observations \lbrace {X}_i, y_i\rbrace from a logistic model such that \mathbb {P}(y_i = 1 \, | \, {X}_i) = \rho ^{\prime }({X...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0819072350859642, -0.015288944356143475, 0.0005068658501841128, -0.0025538792833685875, 0.0021571600809693336, 0.0067976294085383415, 0.017577707767486572, 0.023040225729346275, 0.007480443920940161, 0.019134066998958588, -0.04891851544380188, 0.04681285470724106, -0.023147033527493477, ...
cc0ec8dc6d9bce82fe8882796d92d43e072d269d
subsection
13
45
When does the MLE exist?
The MLE \hat{{\beta }} is the minimizer of the negative log-likelihood \ell defined via (observe that the sigmoid is the first derivative of \rho )\ell ({b})=\sum _{i=1}^n \lbrace \rho ({X}_i^{\prime }{b}) - y_i \, ({X}_i^{\prime }{b}) \rbrace , \qquad \rho (t) = \log (1+e^t).A first important remark is that in high di...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 787, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescale...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.08089543879032135, -0.01482065487653017, -0.007379801012575626, 0.007375985383987427, 0.029992366209626198, -0.05085728317499161, -0.020666494965553284, 0.04508776217699051, 0.0011476086219772696, 0.009554820135235786, -0.03629610314965248, 0.012844058685004711, -0.011592469178140163, -...
510b6b19035e3d04b428bbecd25c67d843df873e
subsection
14
45
A system of nonlinear equations
As we shall soon see, the asymptotic behavior of both the MLE and the LRT is characterized by a system of equations in three variables (\alpha ,\sigma ,\lambda ):\left\lbrace \begin{aligned}\sigma ^2 & = \frac{1}{\kappa ^2} \operatorname{\mathbb {E}}\left[2\rho ^{\prime }(Q_1)\left(\lambda \rho ^{\prime }(\mathsf {prox...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0517704039812088, -0.006379725877195597, -0.009302495047450066, -0.029395578429102898, 0.009081188589334488, -0.042246606200933456, -0.011561345309019089, 0.02617518976330757, -0.0007850649417378008, 0.01498777698725462, -0.02219167724251747, 0.039987754076719284, -0.024862615391612053, ...
f735e5bc223653da51dd0f4edc5fe33f59d20e58
subsection
15
45
A system of nonlinear equations
It turns out that the system admits a unique solution if and only if (\kappa ,\gamma ) is in the region where the MLE asymptotically exists!It is instructive to note that in the case where the signal strength vanishes, \gamma =0, the system of equations (REF ) reduces to the following two-dimensional system:\left\lbrac...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 867, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescale...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.08629646897315979, 0.006343309301882982, -0.003122067777439952, -0.022153902798891068, 0.018675191327929497, -0.031552527099847794, 0.027616091072559357, 0.0635932981967926, 0.015059160999953747, 0.012190748937427998, -0.009040073491632938, 0.041408877819776535, -0.007514323107898235, 0...
737049059c80cda18446873ffe286895a005edd7
subsection
16
45
The average behavior of the MLE
Our first main result characterizes the `average' behavior of the MLE.Theorem 2 Assume the dimensionality and signal strength parameters \kappa and \gamma are such that \gamma < g_{\text{MLE}}(\kappa ) (the region where the MLE exists asymptotically and shown in Figure REF ). Assume the logistic model described above...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.044072434306144714, 0.004246244672685862, -0.030276907607913017, -0.025530880317091942, -0.012612834572792053, -0.037327274680137634, -0.015733614563941956, 0.018938327208161354, -0.0033554108813405037, 0.038364991545677185, -0.02156314067542553, 0.04355357587337494, -0.029910655692219734...
c835e8defb9c7798eeeb08cee034544bad3ed74a
subsection
17
45
The average behavior of the MLE
This can be seen by taking \psi (t,u) = t^2, which leads to \frac{1}{p} \sum _{j=1}^p (\hat{\beta }_j - \alpha _{\star }\beta _{j})^2 \,\, {\stackrel{\text{a.s.}}{\longrightarrow }} \, \, \sigma _{\star }^2. As before, this can also be seen from the empirical results from the previous section. When \kappa = 0.2 and \...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.03554443269968033, -0.009587843902409077, -0.03243238851428032, -0.03276799991726875, 0.020090993493795395, -0.013638078235089779, -0.047504447400569916, 0.054948944598436356, 0.007734346203505993, 0.021204618737101555, -0.004889266099780798, 0.06102048233151436, -0.043141480535268784, ...
a1c5bd301d8f15c4f2db20b91423fdf3462722ef
subsection
18
45
The average behavior of the MLE
Note that there are several ways of determining \alpha _{\star } empirically. For instance, the limit (REF ) directly suggests taking the ratio \sum _j \hat{\beta }_j/\sum _j \beta _j. An alternative is to consider taking the ratio when restricting the summation to nonzero indices. Empirically, we find there is not muc...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.026678165420889854, -0.03220304846763611, -0.019214995205402374, 0.001926841214299202, -0.014262440614402294, -0.03910152241587639, -0.00546001922339201, 0.03174518421292305, 0.010317184962332249, -0.004757962189614773, -0.0189402773976326, 0.036903779953718185, -0.03623224422335625, 0....
1d8a197a311d0f11c2db362645e434023b4529dd
subsection
19
45
The distribution of the null MLE coordinates
Whereas Theorem REF describes the average or bulk behavior of the MLE across all of its entries, our next result provides the explicit distribution of \hat{\beta }_j whenever \beta _j = 0, i.e. whenever the j-th variable is independent from the response y.Theorem 3 Let j be any variable such that \beta _j = 0. Then in...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.05091564357280731, -0.02251216396689415, -0.03315011411905289, -0.010035082697868347, -0.001697951229289174, -0.020192265510559082, -0.024496285244822502, 0.008478309959173203, 0.014415415935218334, 0.008081484586000443, -0.03333326429128647, 0.03464583680033684, -0.04532957449555397, 0...
59326984f27e62cd45ab5dbcad9f6bf04db793ff
subsection
20
45
The distribution of the LRT
We finally turn our attention to the distribution of the likelihood ratio statistic for testing \beta _j = 0.Theorem 4 Consider the LLR \Lambda _j = \min _{{b}\, : \, b_j = 0} \ell ({b}) - \min _{{b}} \ell ({b}) for testing \beta _j = 0. In the setting of Theorem REF , twice the LLR is asymptotically distributed as a ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 1301, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescal...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.09387185424566269, 0.009620645083487034, -0.028137143701314926, -0.011001561768352985, 0.01161191239953041, 0.03506461903452873, -0.006210313178598881, 0.03436271473765373, -0.0015840494306758046, 0.005886064376682043, -0.047302138060331345, 0.04513539746403694, -0.027435241267085075, 0...
2af5fb1dcabae5536a84a547f9cbada6662e003a
subsection
21
45
The distribution of the LRT
We plot the empirical cdf of the p-values, zooming in the tail, in Figure REF (d). We find that the rescaled chi-squared approximation works extremely well even in the tails of the distribution. [Table: P-value probabilities estimated over 500,000replicates with standard errors in parentheses. Here, \kappa =0.1 and the...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 1501, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescal...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.07701331377029419, -0.007189485710114241, -0.027690008282661438, 0.011953234672546387, 0.018337955698370934, 0.0034116683527827263, 0.0025954614393413067, 0.018032832071185112, -0.024882866069674492, 0.020458567887544632, -0.058919455856084824, 0.03746923431754112, -0.04058149829506874, ...
70bc9b7bc774476e55eeb5774c85fce0216d5f6a
subsection
22
45
Other scalings
Throughout this section, we worked under the assumption that \lim _{n \rightarrow \infty } \text{Var}({X}_i^{\prime }{\beta }) = \gamma ^2, which does not depend on n, and we explained that this is the only scaling that makes sense to avoid a trivial problem. We set the variables to have variance 1/n but this is of cou...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.060204099863767624, -0.028988288715481758, -0.03979023918509483, 0.00016520463395863771, 0.0007814441341906786, -0.00897111278027296, -0.006613906938582659, 0.006060841027647257, 0.012335279956459999, -0.016752179712057114, -0.04711359739303589, 0.017042061313986778, -0.029964735731482506...
be7ca00e8f2dac044721fed50dd39ccbf8c13379
subsection
23
45
Non-Gaussian covariates
Our model assumes that the features are Gaussian, yet, we expect that the same results hold under other distributions with the proviso that they have sufficiently light tails. In this section, we empirically study the applicability of our results for certain non-Gaussian features. [Figure: Simulation for a non-Gaussian...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.07920031249523163, -0.016505150124430656, -0.053664617240428925, 0.00011750576231861487, -0.028403501957654953, -0.0028010704554617405, 0.01644413359463215, 0.03340691328048706, 0.0052398513071238995, 0.018503462895751, -0.059827353805303574, 0.04039337858557701, -0.01925092376768589, 0...
7fbcfdd9fd91e9afa53dc9b6150c9d2e7e728cd4
subsection
24
45
Adjusting Inference by Estimating the Signal Strength
All of our asymptotic results, namely, the average behavior of the MLE, the asymptotic distribution of a null coordinate, and the LLR, depend on the unknown signal strength \gamma . In this section, we describe a simple procedure for estimating this single parameter from an idea proposed by Boaz Nadler and Rina Barber ...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.06511213630437851, 0.011800430715084076, -0.022456670179963112, -0.020732754841446877, 0.0013940070057287812, -0.03280016407370567, 0.03392909839749336, -0.0068765925243496895, 0.021205687895417213, -0.0021091268863528967, -0.044821806252002716, 0.04604227468371391, 0.005381515249609947, ...
93dcb07d4065b0d965336c82de79bc4b7d85b771
subsection
25
45
ProbeFrontier: estimating
We estimate the signal strength by actually using the predictions from our theory, namely, the fact that we have asymptotically characterized in Section 2 the region where the MLE exists. We know from Theorem REF that for each \gamma , there is a maximum dimensionality g_{\text{MLE}}^{-1}(\gamma ) at which the MLE ceas...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.04611726850271225, -0.017915841192007065, -0.006405599880963564, -0.023546969518065453, 0.02276868373155594, -0.03384781628847122, 0.03903638944029808, 0.02166992612183094, 0.007740894798189402, 0.013352948240935802, -0.037022002041339874, -0.000859834486618638, -0.021975137293338776, 0...
cb733ef7495851ffe68c0545088abbfc53ec6306
subsection
26
45
Empirical performance of adjusted inference
We demonstrate the accuracy of ProbeFrontier via some empirical results. We begin by generating 4000 i.i.d. observations (y_i,{X}_i) using the same setup as in Figure REF (\kappa =0.1 and half of the regression coefficients are null). We work with a sequence \lbrace \kappa _j\rbrace of points spaced apart by 10^{-3} a...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.049042392522096634, 0.000445133337052539, -0.0545966662466526, -0.02391083724796772, 0.015243731439113617, -0.02389557845890522, 0.017807241529226303, 0.019821427762508392, 0.023392030969262123, 0.028595346957445145, -0.04623473808169365, 0.03115885704755783, -0.025131557136774063, 0.01...
31d41cea9104d0981a2246c7577d95d6d093bc80
subsection
27
45
Empirical performance of adjusted inference
We report an average over 6000 replicateswith the std. dev. between parentheses.]Finally, we focus on the estimation accuracy for a particular (\kappa ,\gamma ) pair across several replicates. In the setting of Figure REF , we generate 6000 samples and obtain estimates of bias (\hat{\alpha }), std. dev. (\hat{\sigma })...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.055862508714199066, 0.001776212709955871, -0.03291952237486839, -0.013248658739030361, -0.0029346123337745667, 0.009328215382993221, 0.012775764800608158, 0.005300988908857107, -0.002442650031298399, 0.004728939849883318, -0.03407887741923332, 0.06620515882968903, -0.04896740987896919, ...
9d91a0c0d8bfa95413e37825b83e018d288bb38c
subsection
28
45
De-biasing the MLE and its predictions
We have seen that maximum likelihood produces biased coefficient estimates and predictions. The question is how precisely can our proposed theory and methods correct this. Recall the example from Figure REF , where the theoretical prediction for the bias is \alpha _{\star }= 1.499. For this dataset, ProbeFrontier yield...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.05117117241024971, -0.028636937960982323, -0.04354279115796089, 0.008124223910272121, -0.0029922318644821644, -0.019910071045160294, -0.0023857757914811373, 0.049981143325567245, 0.03896576166152954, 0.026684071868658066, -0.05873852223157883, 0.010702616535127163, -0.005427591968327761, ...
c016e9021dad3acce48b046c8e24a238485b43d3
subsection
29
45
Main Mathematical Ideas
As we mentioned earlier, we do not provide detailed proofs in this paper. The reader will find them in and the first author's Ph. D. thesis. However, in this section we give some of the key mathematical ideas and explain some of the main steps in the arguments, relying as much as possible on published results from .
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1073/pnas.1810420116", "end": 130, "openalex_id": "https://openalex.org/W2963060833", "raw": "Pragya Sur and Emmanuel J Candès. Additional supplementary materials for: A modern maximum-likelihood theory for high-dimensional logistic regr...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ 0.006297281477600336, 0.03246666118502617, -0.03225306421518326, -0.016019735485315323, -0.0067893159575760365, -0.015295032411813736, 0.018582891672849655, 0.007430105470120907, 0.0107179656624794, 0.004763964097946882, 0.014471160247921944, -0.02196992188692093, -0.010649309493601322, 0....
655b23333cef48372ee51b88fdec62d8296aae37
subsection
30
45
The bulk distribution of the MLE
To analyze the MLE, we introduce an approximate message passing (AMP) algorithm that tracks the MLE in the limit of large n and p. Our purpose is a little different from the work in which, in the context of generalized linear models, proposed AMP algorithms for Bayesian posterior inference, and whose properties have l...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/isit.2011.6033942", "end": 347, "openalex_id": "https://openalex.org/W2166670884", "raw": "Sundeep Rangan. Generalized approximate message passing for estimation with random linear mixing. In IEEE International Symposium on Informat...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.04287608340382576, -0.02520686388015747, -0.02610711008310318, 0.008666766807436943, -0.010825829580426216, -0.05114613100886345, 0.010596953332424164, -0.014602282084524632, -0.012397443875670433, 0.03137125447392464, -0.023818351328372955, 0.011413278058171272, -0.03796287998557091, 0...
a02ad46e9f28fdfd579bd9ef1f41f56324437e1a
subsection
31
45
The bulk distribution of the MLE
Then starting from an initial pair \alpha _0, \sigma _0, for t = 0, 1, \ldots , we inductively define \lambda _t as the solution to\operatorname{\mathbb {E}}\left[\frac{2\rho ^{\prime }(Q_1^t)}{1+\lambda \rho ^{\prime \prime }(\mathsf {prox}_{\lambda \rho }(Q_2^t))} \right] = 1- \kappaand the extra parameters \alpha _{...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ 0.00473451241850853, 0.005520418751984835, -0.011048467829823494, -0.00002283086178067606, 0.007637787144631147, -0.04868040233850479, 0.037662457674741745, 0.043491896241903305, 0.010804302990436554, 0.003805540967732668, 0.005047348793596029, 0.020708246156573296, -0.037906620651483536, ...
3b386914bd304242b92e4203e23b72d462e90b97
subsection
32
45
The bulk distribution of the MLE
From now on, we use \alpha _0 = \alpha _{\star }, \sigma _0 = \sigma _{\star } so that the sequence \lbrace \alpha _t, \sigma _t, \lambda _t\rbrace is stationary; i. e. for all t \ge 0,\alpha _t = \alpha _{\star }, \quad \sigma _t = \sigma _{\star }, \quad \lambda _t = \lambda _{\star }.With this stationary sequence of...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1093/imaiai/iat004", "end": 1819, "openalex_id": "https://openalex.org/W2963072996", "raw": "Adel Javanmard and Andrea Montanari. State evolution for general approximate message passing algorithms, with applications to spatial coupling. ...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.015614408999681473, -0.005713332910090685, -0.025202739983797073, 0.027033446356654167, 0.0075059006921947, -0.02131248638033867, 0.0024352222681045532, 0.021800674498081207, 0.02285333164036274, 0.030938956886529922, -0.01520249992609024, 0.017284929752349854, -0.039268676191568375, 0....
cd3e521ef6359abe51ccb493578d4f57912ff310
subsection
33
45
The bulk distribution of the MLE
From here on, an analysis similar to that in establishes that in the limit of large iteration counts, the AMP iterates converge to the MLE, that is,\lim _{t \rightarrow \infty } \lim _{n \rightarrow \infty }\frac{1}{p} \sum _{j=1}^{p}\psi (\hat{\beta }_{j}^{t } - \alpha _{\star }\beta _{j}, \beta _j) = \lim _{n \right...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 461, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescale...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.0066957250237464905, 0.014818346127867699, -0.020983753725886345, -0.0030273834709078074, -0.02296767197549343, -0.007638086564838886, -0.012315556406974792, 0.019625533372163773, -0.007928043603897095, 0.012246881611645222, -0.03494748845696449, 0.013986625708639622, -0.00309033459052443...
25c15e70564e2f37bb12c5c791a28ec7b7c242cf
subsection
34
45
The distribution of a null coordinate
We sketch the proof of Theorem REF in the case where the empirical limiting distribution \Pi has a point mass at zero. The analysis in the general case, where the number of vanishing coefficients is arbitrary, and in particular, o(n), is very different and may be found in Appendix .Now consider Theorem REF with \psi (t...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.039027441293001175, 0.02412133850157261, -0.041712675243616104, -0.03164304792881012, 0.008124359883368015, -0.017240425571799278, 0.0049852291122078896, 0.015287527814507484, 0.02703542821109295, 0.022626152262091637, -0.03829510509967804, 0.02064274065196514, -0.015180729329586029, 0....
318d99676d9d632bae1c9514d74235284a970a18
subsection
35
45
The distribution of a null coordinate
From (REF ) and the weak law of large numbers, we have \Vert \hat{{\beta }}_{[k]} \Vert /\Vert {Z}\Vert \stackrel{\mathbb {P}}{\rightarrow }\sigma _{\star }, leading to \hat{\beta }_j \stackrel{\mathrm {d}}{\rightarrow }{\mathcal {N}}(0, \sigma _{\star }^2).
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.029064413160085678, 0.03499935194849968, 0.007002921774983406, -0.03518243506550789, -0.015653589740395546, -0.006156163290143013, 0.008406557142734528, 0.0432380847632885, -0.0004145588609389961, 0.024731451645493507, 0.003375591477379203, 0.02889658883213997, -0.02972046099603176, 0.0...
113ec7616b927f76713c342c4e0dbbe19c70b050
subsection
36
45
Broader Implications and Future Directions
This paper shows that in high-dimensions, classical ML theory is unacceptable. Among other things, classical theory predicts that the MLE is approximately unbiased when in reality it seriously overestimates effect magnitudes. Since the purpose of logistic modeling is to estimate the risk of a specific disease given a p...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.06398127973079681, -0.00001728978895698674, -0.02516820840537548, -0.005574704613536596, 0.0009949379600584507, -0.019169963896274567, -0.006517176982015371, 0.004147641360759735, 0.0066240159794688225, 0.014965087175369263, -0.04459763690829277, 0.03858413174748421, -0.01079836767166853,...
d4ac2d041fd525f34c46dedbcdb3f3d0344d21cb
subsection
37
45
Fisher information
We work with the model from Section and introduce the Fisher information matrix defined asI({\beta }) = \operatorname{\mathbb {E}}\left[ \sum _i \rho ^{\prime \prime }({X}_i^{\prime }{\beta }) {X}_i {X}_i^{\prime }\right] = n \operatorname{\mathbb {E}}\left[ \rho ^{\prime \prime }({X}_i^{\prime }{\beta }) {X}_i {X}_i^{...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.06519968807697296, -0.03180621564388275, -0.05872855708003044, 0.018833430483937263, 0.020298592746257782, -0.023167867213487625, 0.01854345016181469, 0.060590535402297974, 0.030676819384098053, 0.040597185492515564, -0.011866282671689987, 0.010332440957427025, -0.047281984239816666, 0....
9259f34d69a3c46036a7b35802e514c793e06f8e
subsection
38
45
Properties of fixed points of the AMP algorithm
In this section, we elaborate on the connection between the fixed points of (REF ) and the MLE \hat{{\beta }}. From (REF ), we immediately see that if (\hat{{\beta }}_{\star },{S}_{\star }) is a fixed point, the pair satisfies{X}^{\prime }\lbrace {y}- \rho ^{\prime }(\mathsf {prox}_{\lambda _{\star }\rho }(\lambda _{\s...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ 0.007790914736688137, -0.005375807639211416, -0.026402121409773827, -0.016314374282956123, -0.005265162792056799, -0.009523076936602592, 0.024937031790614128, 0.02347194403409958, 0.054208286106586456, 0.005143071990460157, -0.015887057408690453, 0.014063326641917229, 0.006577637977898121, ...
b8ae8bbb2c82bbe79851040e606fa717c324d7d5
subsection
39
45
Refined analysis of the distribution of a null coordinate
The AMP analysis is useful to analyze the bulk behavior of the MLE; i.e. the expected behavior when averaging over all coordinates. It also helps in characterizing the distribution of a null coordinate when the limiting empirical cdf does not have a point mass at zero, as we have seen in Section REF . However, the stud...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-016-0754-9", "end": 503, "openalex_id": "https://openalex.org/W2581138301", "raw": "Noureddine El Karoui. On the impact of predictor geometry on the performance on high-dimensional ridge-regularized generalized robust regress...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.031866416335105896, -0.02976030297577381, -0.05549151822924614, 0.016009517014026642, -0.015292217023670673, -0.03906993567943573, 0.027715235948562622, -0.004742571152746677, -0.014582548290491104, 0.03726905584335327, -0.04801328852772713, -0.006932624150067568, -0.04938684031367302, ...
4812d04ea4f786c6a2e5a38028fb2ff3160d7d17
subsection
40
45
Refined analysis of the distribution of a null coordinate
On the other hand, we can subtract the two score equations \nabla \ell (\hat{{\beta }})=0 and \nabla {\ell _{[-j]}} (\hat{{\beta }}_{[-j]})=0 (\ell _{[-j]} is the negative log-likelihood for the reduced model), which upon separating the components corresponding to the j-th coordinate from the others, yields\sum _{i=1}...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.053511787205934525, 0.005052025429904461, -0.02590116858482361, 0.03144160658121109, -0.013652678579092026, -0.008096977137029171, 0.019139697775244713, 0.061387449502944946, -0.012584274634718895, 0.009470639750361443, -0.03125845268368721, 0.03144160658121109, -0.03598995879292488, 0....
a04dbaf3fadfb122cb42226525ea048301ddc23d
subsection
41
45
Refined analysis of the distribution of a null coordinate
It was established in that the denominator above is equal to \kappa /\lambda _{[-j]} + o_P(1), where, we have see in Section REF that\lambda _{[-j]} := \frac{1}{n} \mathrm {Tr}[\nabla ^2(\ell _{[-j]}(\hat{{\beta }}_{[-j]}))^{-1}].Note that since \beta _j = 0, {y} and {X}_{\bullet -j}, \hat{{\beta }}_{[-j]} are independ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00440-018-00896-9", "end": 232, "openalex_id": "https://openalex.org/W2963669167", "raw": "Pragya Sur, Yuxin Chen, and Emmanuel J Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescale...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.01495679747313261, 0.019855912774801254, -0.0478007048368454, 0.030554601922631264, -0.007619572803378105, -0.057843126356601715, 0.0014432164607569575, 0.03507216274738312, -0.008134666830301285, 0.008127035573124886, -0.045206159353256226, 0.02081742137670517, -0.07032747566699982, 0....
4102805becd2ce480867275df5ea61e602edf2a7
subsection
42
45
Refined analysis of the distribution of a null coordinate
Denote by \nabla {\ell }_{[-i],[-j]}(\tilde{{\beta }}_{[-i],[-j]})=0 the reduced score equation and subtract it from the score equation for \hat{{\beta }} to obtain{X}_{i,-j}\left(y_i - \rho ^{\prime }({X}_{i,-j}^{\prime } \hat{{\beta }}_{[-j]} ) \right) + \sum _{k \ne i} {X}_{k,-j}\left( \rho ^{\prime }({X}_{k,-j}^{\p...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1214/ecp.v18-2865", "end": 1382, "openalex_id": "https://openalex.org/W2032291279", "raw": "Mark Rudelson, Roman Vershynin, et al. Hanson-Wright inequality and sub-gaussian concentration. Electron. Commun. Probab, 18(82):1–9, 2013.", ...
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.04394644871354103, 0.020920341834425926, -0.0469982847571373, 0.02233944460749626, -0.011528313159942627, -0.00245863595046103, 0.014686964452266693, 0.04095564782619476, 0.005024086218327284, -0.002945022424682975, -0.006336376070976257, -0.005500935949385166, -0.0470898412168026, 0.02...
8f329321c397818a2e6bf882a21b11fb0f798ed9
subsection
43
45
Refined analysis of the distribution of a null coordinate
Hence, the fitted values can be approximated as{X}_{i,-j}^{\prime }\hat{{\beta }}_{[-j]} \approx {X}_{i,-j}^{\prime } \hat{{\beta }}_{[-i],[-j]} + \lambda _{[-j]} \left(y_i - \rho ^{\prime }({X}_{i,-j}^{\prime }\hat{{\beta }}_{[-j]}) \right).Recalling the definition of the proximal mapping operator, \mathsf {prox}_{\la...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.04913298040628433, 0.0034885178320109844, -0.053222306072711945, -0.012275615707039833, -0.006648973561823368, -0.03216531500220299, 0.0361936092376709, 0.04165622219443321, 0.017669562250375748, 0.025024250149726868, -0.038299310952425, 0.030685223639011383, -0.030364790931344032, 0.03...
f54855125b4b6e08778e420da40bb5e4bda1fc7c
subsection
44
45
Refined analysis of the distribution of a null coordinate
Instead, we find the joint distribution of ({X}_{i,-j}^{\prime }{\beta }_{-j},{X}_{i,-j}^{\prime }(\hat{{\beta }}_{[-i],[-j]} - \alpha _{\star }{\beta }_{-j}) ) and denote this pair as (Q_{1}^{\star }, Q_{2}^{\star }). The asymptotic variance of Q_{1}^{\star } is given by \gamma ^2, that of Q_{2}^{\star } by \kappa \si...
{ "cite_spans": [] }
10.1073/pnas.1810420116
1803.06964
A modern maximum-likelihood theory for high-dimensional logistic regression
[ "Pragya Sur", "Emmanuel J. Candes" ]
[ "math.ST", "stat.ME", "stat.TH" ]
2,018
en
Mathematics
[ -0.05711529403924942, 0.03150346502661705, -0.029000282287597656, -0.004857547581195831, -0.010661419481039047, -0.023673389106988907, -0.017568066716194153, 0.051437344402074814, 0.016377529129385948, -0.006044269539415836, -0.010424838401377201, 0.033915068954229355, -0.04838468134403229, ...
495b3e97ceba37a90794f46502bf4d1a63a0732e
abstract
0
92
Abstract
For many nonlinear physical systems, approximate solutions are pursued by conventional perturbation theory in powers of the non-linear terms. Unfortunately, this often produces divergent asymptotic series, collectively dismissed by Abel as "an invention of the devil." An alternative method, the self-consistent expansio...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04256337881088257, 0.008207562379539013, -0.03780360519886017, 0.020122257992625237, 0.013806400820612907, -0.018672967329621315, 0.05858185887336731, -0.0068536195904016495, 0.028329819440841675, 0.04781133681535721, -0.012013857252895832, 0.002069053240120411, -0.02994692325592041, 0....
8d0bdfc9bc66076309355578791d716c69a8cd5a
abstract
1
92
Abstract
These results allow us to treat the Airy function, and to see the fingerprints of Stokes lines in the SCE.
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ 0.005611927714198828, 0.038089387118816376, -0.05707303807139397, 0.003254231298342347, 0.02299707755446434, -0.017182951793074608, 0.02432471141219139, 0.02728518471121788, 0.005718749016523361, 0.04224015399813652, 0.016496244817972183, 0.009804659523069859, 0.001426825998350978, 0.03283...
6eab5c9b8de3287e0af3641dcc2b7f131c29beaa
subsection
2
92
Introduction
A recurring theme in physics is the necessity of approximations. Exactly solvable systems are rare in the landscape of modern science, and the majority of physical problems cannot be assailed directly without resorting to any estimates or simplifications. First-principle calculations must often be replaced by phenomeno...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 863, "openalex_id": "", "raw": "J. P. Boyd, Acta Applicandae Mathematica 56, 1 (1999).", "source_ref_id": "a1f0dc9b3e8b3359250931a8cd2526252333c171", "start": 797 }, { "arxiv_id": "", "doi": "", ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.05490342527627945, 0.01844865083694458, -0.0460224375128746, 0.019425254315137863, 0.015495950356125832, -0.01606817916035652, 0.04437442123889923, -0.012344878166913986, -0.007252043578773737, 0.043519891798496246, -0.03939984738826752, 0.003975081257522106, -0.03576810285449028, 0.003...
1ef7d78a05d7e27bd2c18c7b11cdb9c005923224
subsection
3
92
Introduction
This method was subsequently applied to more complicated variants of the KPZ equation , , , , , and to other problems such as fracture, turbulence, and the XY model , , , .The SCE’s central idea is to smartly choose the zeroth-order system around which one expands, so that the zeroth-order system is as close as possibl...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 172, "openalex_id": "", "raw": "E. Katzav and M. Schwartz, Phys. Rev. E 60, 5677 (1999).", "source_ref_id": "41ac279c838b05618c0fc6fc7399515368ef3348", "start": 0 }, { "arxiv_id": "", "doi": "", ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.011003776453435421, 0.009462332352995872, -0.07136429846286774, 0.042580496519804, 0.021458128467202187, -0.01802421733736992, 0.04200054705142975, -0.016589606180787086, 0.0003042826720047742, 0.060772594064474106, -0.019641971215605736, 0.007237921003252268, -0.015765467658638954, 0.0...
db60a38eb401ccad841e53ca1f8e1c1c70c78933
subsection
4
92
Introduction
Moreover, our simplified approach will allow us to then extend our treatment to the double well case, higher anharmonicities, many coupled oscillators, and, most importantly, the complex coupling case vis-à-vis the Stokes phenomenon.Our main result is a proof that by imposing self-consistency (i.e., determining the spl...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.00545972166582942, 0.03075147047638893, -0.05619201809167862, 0.03659654036164284, 0.01855771243572235, -0.02896590158343315, 0.009706171229481697, 0.0010301361326128244, 0.024463824927806854, 0.06122823804616928, -0.017458898946642876, 0.020129622891545296, 0.014017482288181782, 0.0195...
d23502b607c5892c2a13e19047d6e048487aad46
subsection
5
92
Divergence of PT
A canonical example of the limitation of PT is found in the anharmonic oscillator. The simplest anharmonicity that keeps global stability is a quartic perturbation, which we writeV(x)=\frac{1}{2}\gamma x^{2}+g_{0}x^{4}\,,with g_{0}>0. This potential refines models of ideal binding interactions and also lends itself to ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1119/1.1972842", "end": 1053, "openalex_id": "https://openalex.org/W2120062331", "raw": "M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, Vol. 55 (Dover, New York, 1964).", "source_ref_id": "d88a70f14cea1e8d188fa...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.042532000690698624, 0.038748666644096375, -0.024942547082901, 0.012227187864482403, 0.0195116326212883, -0.06431668996810913, 0.020335422828793526, 0.0016189772868528962, 0.014492612332105637, 0.05333281308412552, -0.047749344259500504, 0.01800134964287281, -0.009015929885208607, 0.0036...
850cebbf692bc8649e5750a7d739fa45c890c194
subsection
6
92
Divergence of PT
Second, the application of the resummation procedure presents its own challenges, such as knowledge of the late PT terms, which may not be available. Therefore, we will use the anharmonic oscillator as a test-bed and show that a convergent series may instead be obtained by applying the SCE.
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.019292686134576797, 0.05146751180291176, -0.04875066131353378, 0.013080562464892864, 0.0075629157945513725, -0.04230959340929985, -0.0007445579976774752, 0.02690902166068554, 0.0070019932463765144, 0.05210856720805168, -0.008646602742373943, 0.01793426088988781, -0.03412851691246033, -0...
3fce5083130f12a953aa82df19e3c45f4840e73d
subsection
7
92
SCE Around a Modified Oscillator
Ref. offers a treatment of the anharmonic oscillator by expanding its Fokker-Planck equation of motion. Here we pursue another approach: In the language of the SCE, instead of expanding the system around the harmonic term, we expand around a modified harmonic potential, whose strength is consistently varied to obtain a...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 104, "openalex_id": "", "raw": "M. Schwartz and E. Katzav, J. Stat. Mech: Theory Exp. 2008, P04023 (2008).", "source_ref_id": "514a3da569fdcb97e9e7ce4be7023d41e674bf71", "start": 0 }, { "arxiv_id": "", ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04123443737626076, 0.011086905375123024, -0.01980840042233467, 0.028842739760875702, 0.01730564422905445, -0.0403798371553421, 0.044164493680000305, 0.0011140317656099796, 0.0015470543876290321, 0.04157017171382904, -0.021914377808570862, 0.009835527278482914, -0.03720561042428017, 0.00...
692fdf559d02e725a5c606073c29b864525e6583
subsection
8
92
SCE Around a Modified Oscillator
Binomial expansion then gives:\mathcal {Z}^{\left(N\right)} & =\int _{-\infty }^{\infty }e^{-\frac{1}{2}G\left(N\right)x^{2}}\sum _{n=0}^{N}\frac{\left(-1\right)^{n}}{n!}\sum _{l=0}^{n}\binom{n}{l}\left[\frac{1}{2}(1-G\left(N)\right)x^{2}\right]^{n-l}\left[gx^{4}\right]^{l}dx\\ & =\sum _{n=0}^{N}\frac{\left(-1\right)^{...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0623958483338356, 0.03364003077149391, -0.0024917128030210733, -0.021322650834918022, -0.0058190845884382725, 0.0023867785930633545, 0.05775584280490875, 0.0028732921928167343, -0.026115287095308304, -0.0064944797195494175, -0.04917794093489647, 0.004288951400667429, -0.05870215967297554,...
8c4123d281dc856fac447b3750b4f7cf37d6a00f
subsection
9
92
The Self-Consistent Criteria for
Lastly, we require a choice of the function G\left(N\right). In the ODM it is determined solely based on mathematical convergence properties, while for the OPT and the LDE this is usually done by one of two common criteria: the principle of minimal sensitivity (PMS), or the principle of fastest apparent convergence (FA...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 324, "openalex_id": "", "raw": "P. M. Stevenson, Phys. Rev. D 23, 2916 (1981).", "source_ref_id": "310cc416c25d060fe5ded668878abb76ef44a130", "start": 61 }, { "arxiv_id": "", "doi": "", "end":...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0387607142329216, 0.001280896714888513, -0.029223138466477394, 0.013230527751147747, 0.01104832999408245, -0.05063309147953987, 0.014405556954443455, -0.01718289963901043, 0.03680742159485817, 0.07221090793609619, -0.04895447939634323, -0.021120011806488037, 0.004207979422062635, 0.0181...
61ea928bf4ec78564d1cff3e1beb199a03f9fcc9
subsection
10
92
The Self-Consistent Criteria for
To first order in the SCE perturbation, the partition function is\mathcal {Z}^{\left(1\right)} & =\sqrt{\frac{2}{G}}\Gamma \left(\frac{1}{2}\right)\left(1+\frac{1}{2}\left(\left(1-\frac{1}{G}\right)-\frac{4g}{G}\frac{3}{2}\right)\right)\,,while the corresponding moment x^{2M} would be\left[\mathcal {Z}\cdot \left\langl...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ 0.014866609126329422, -0.015126087702810764, -0.029656900092959404, -0.01005861908197403, 0.021948853507637978, -0.04804936796426773, 0.017232445999979973, -0.0042508733458817005, 0.0027874892111867666, 0.02280360832810402, 0.008219372481107712, -0.027840549126267433, -0.00536510581150651, ...
114cea3f49e10d4f1574ef2130418dd5fc48867e
subsection
11
92
The Self-Consistent Criteria for
A useful inequality for G is then:\max \left(1,\,\frac{1}{2}+2\sqrt{gK}\right)<G\left(K\right)<1+2\sqrt{gK}\,.A particular convenience of this choice is that\frac{(1-G)G}{4g}=\frac{1}{16g}\left(1-\sqrt{1+16Kg}\right)\left(1+\sqrt{1+16Kg}\right)=-K\,,so now the expansion takes the form\mathcal {Z}^{\left(N\right)}=\sqrt...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 716, "openalex_id": "https://openalex.org/W3089116201", "raw": "DLMF, “NIST Digital Library of Mathematical Functions,” http://dlmf.nist.gov/.", "source_ref_id": "1bad14f3312569fc1768be8ccfc8645c155416c3", "start": 548...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.05850151926279068, 0.01907326653599739, -0.026000676676630974, -0.027877485379576683, -0.008689779788255692, -0.002992595313116908, 0.03204308822751045, -0.030257828533649445, -0.0064925397746264935, -0.0015477954875677824, -0.02946438081562519, 0.03295860439538956, -0.01275620050728321, ...
f01e87292b9fff954b7e8324f44ae49ec08cfb51
subsection
12
92
Convergence Properties of the SCE
Let us state and prove our main result for the convergence properties of the SCE; its relation to the convergence properties and their proofs for the related schemes mentioned above will be discussed at the end of this Section. We will show that the following proposition holds:Proposition 1 Let the self-consistently co...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.007044944446533918, 0.03948525711894035, -0.04928028583526611, 0.013754233717918396, -0.01045108214020729, -0.028835831210017204, 0.009451745077967644, 0.01621824875473976, 0.020139310508966446, 0.045923732221126556, -0.0012072144309058785, 0.010847765021026134, -0.011030850000679493, 0...
f7355b8b314531eef42ff67d95df91e7b225b812
subsection
13
92
Convergence to
For our proof we will denote explicitly the limiting operations involved in the definitions of the summations of infinite series, in order to emphasize that we never stumble into the same pitfall as regular perturbation theory. Our proof begins by examining:\lim _{N\rightarrow \infty }\mathcal {Z}^{\left(N\right)} & =\...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.057771697640419006, 0.009437874890863895, -0.044068846851587296, -0.008987725712358952, 0.009918542578816414, -0.005443754140287638, -0.0002768131671473384, -0.010109283961355686, -0.02323990873992443, 0.03195296600461006, -0.04004039242863655, -0.02266005612909794, -0.007103202398866415,...
f3c83441a9912ffa9d18189a512d8f871862436c
subsection
14
92
Convergence to
Thus, the error associated with the expansion is due to the remainder of the Taylor series, which is truncated before integration.We note that the integrand has three distinct regions where its behavior is qualitatively different:\mathcal {D}_{1}=\left[0,1\right],\qquad \mathcal {D}_{2}=\left[1,\infty \right)=\mathcal ...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.054366424679756165, 0.04011888429522514, -0.021005209535360336, 0.0089085279032588, -0.008511914871633053, -0.016962813213467598, -0.008153438568115234, 0.008176320232450962, -0.019418759271502495, 0.029852719977498055, -0.019830627366900444, 0.011501763947308064, -0.04085109010338783, ...
ec3a6242c187e67ad32c77eb3d567f9d0f7891a0
subsection
15
92
The Domain
In this domain, the remainder can be bounded explicitly. The error is negative, and is the sum of the truncated terms in the exponent's Taylor series:-R_{1}^{\left(N\right)} & =\sqrt{\frac{2K\left(N\right)}{G\left(K\left(N\right)\right)}}\int _{0}^{1}e^{-Kv}\left[\lim _{L\rightarrow \infty }\sum _{n=N+1}^{L}\frac{K^{n}...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.03756161406636238, 0.003886604681611061, -0.016431298106908798, -0.011396639049053192, -0.005202481988817453, -0.009337005205452442, 0.0004467307007871568, 0.007162947673350573, -0.0059195393696427345, 0.007464264519512653, -0.04625784233212471, 0.021526984870433807, -0.024440985172986984...
f0559fcf7f5bab0ed1348923fd36f6662ca88a66
subsection
16
92
The Domain
We then find-R_{1}^{\left(N\right)} & =\sqrt{\frac{2K\left(N\right)}{G\left(K\left(N\right)\right)}}\sum _{n=N+1}^{\infty }\left(1-\frac{1}{G\left(K\left(N\right)\right)}\right)^{n}\frac{K^{n}\left(N\right)}{n!}\int _{0}^{1}e^{-Kv}v^{n}\left(1-v\right)^{n}\frac{dv}{\sqrt{v}}\\ & <\sqrt{\frac{2K\left(N\right)}{G\left(K\...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 862, "openalex_id": "https://openalex.org/W3089116201", "raw": "DLMF, “NIST Digital Library of Mathematical Functions,” http://dlmf.nist.gov/.", "source_ref_id": "1bad14f3312569fc1768be8ccfc8645c155416c3", "start": 754...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.050438057631254196, 0.025005437433719635, -0.011915342882275581, -0.008993718773126602, -0.029780728742480278, 0.008772499859333038, 0.040673885494470596, -0.01920795999467373, 0.003781710285693407, 0.021252334117889404, -0.04192491993308067, 0.007754127029329538, -0.019818220287561417, ...
6e4f428bb90cf00f6f78af4e460c926aeff66eb7
subsection
17
92
The Domain
With x=n and s=\frac{1}{2} it reads \frac{\Gamma \left(n+\frac{1}{2}\right)}{\Gamma \left(n+1\right)}<n^{-\frac{1}{2}}, leaving us with-R_{1}^{\left(N\right)} & <\sqrt{\frac{2}{G\left(K\left(N\right)\right)}}\sum _{n=N+1}^{\infty }\left(1-\frac{1}{G\left(K\left(N\right)\right)}\right)^{n}\left[\frac{K\left(N\right)}{K\...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0515316016972065, 0.028115080669522285, -0.02549120970070362, -0.0059685432352125645, -0.009030996821820736, -0.01015987154096365, 0.021936779841780663, 0.0005119980196468532, -0.003287466010078788, 0.03438491001725197, -0.03429337963461876, 0.011898948810994625, -0.02192152477800846, -...
dbcb2e2714755d47dd1b6dbfcbd8aba6f719e117
subsection
18
92
The Domain
We thus have\left(-1\right)^{N}R_{2}^{\left(N\right)} & <\sqrt{\frac{2K\left(N\right)}{G\left(K\left(N\right)\right)}}\int _{1}^{\infty }e^{-Kv}\frac{K^{N^{\prime }}\left(N\right)}{N^{\prime }!}\left(1-\frac{1}{G\left(K\left(N\right)\right)}\right)^{N^{\prime }}v^{N^{\prime }-\frac{1}{2}}\left(v-1\right)^{N^{\prime }}d...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0211348719894886, 0.018555959686636925, -0.013031898997724056, -0.00660750875249505, -0.034822944551706314, 0.03259500861167908, 0.03241188824176788, -0.018494920805096626, 0.006229827646166086, 0.020005643367767334, -0.015839708968997, 0.018922194838523865, -0.02397320047020912, 0.0145...
ca0a257e098ce2f61833a8b3a312a7c6c289ae5e
subsection
19
92
The Domain
However, it turns out that this bound is a bit looser, and only shows convergence for M/N>1.04 . We go the extra mile so we can show that M=N, as used in Ref. , leads to convergence as well., which occurs atFor K>2N, this maximum lies outside the domain of integration, as 2N/K<1. However, this maximum still bounds the ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 280, "openalex_id": "", "raw": "M. Schwartz and E. Katzav, J. Stat. Mech: Theory Exp. 2008, P04023 (2008).", "source_ref_id": "514a3da569fdcb97e9e7ce4be7023d41e674bf71", "start": 97 } ] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.06494872272014618, 0.05252666771411896, -0.043950263410806656, 0.00747764902189374, -0.025500308722257614, 0.03114669770002365, -0.030795706436038017, -0.004578152671456337, 0.005184757523238659, 0.028018293902277946, -0.01764114759862423, 0.01366578508168459, -0.004803244955837727, 0.0...
1cb1e4e28d100bf2f3fca18aab855b16c2fad0e1
subsection
20
92
The Domain
We so proceed to bound\ln \left(v\right)+\ln \left(v+1\right)\le \ln \left(v_{0}\right)+\ln \left(v_{0}+1\right)+\left(\frac{1}{v_{0}}+\frac{1}{v_{0}+1}\right)\left(v-v_{0}\right)\,.In total, we now get\int _{1}^{\infty }v^{N^{\prime }} & \left(v+1\right)^{N^{\prime }}e^{-Kv}dv\le \int _{1}^{\infty }v_{0}^{N^{\prime }}...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.037681277841329575, 0.03957296907901764, 0.0003859184216707945, -0.010297514498233795, -0.016201423481106758, 0.0022978712804615498, 0.023310521617531776, -0.003958822228014469, 0.0027040510904043913, 0.018108369782567024, -0.04244101792573929, 0.009954264387488365, -0.02530900202691555, ...
e73c1f33a0cbe1cb54f144ba56f317bfca9e943b
subsection
21
92
The Domain
If q<4 then r<1, and v^{r} is a concave function which on the interval \left[0,1\right] is simply bounded by v. This will reproduce the eventual \left(\frac{K}{K+N^{\prime }}\right)^{N} result obtained for q=4. Let us then focus on q>4, so henceforth r>1: As v^{r} is now a convex function, it is bounded by any line dra...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04520493000745773, 0.053659867495298386, -0.008966203778982162, -0.03562062978744507, 0.002720384392887354, 0.04450289532542229, 0.026570485904812813, 0.004280885215848684, 0.003222110215574503, 0.050027601420879364, -0.014849559403955936, -0.011781973764300346, 0.0009052050299942493, 0...
68623fc5876c9277a033f6c74237527437003a2d
subsection
22
92
The Domain
However, since e^{\left(r-1\right)v_{0}^{r}} can be expanded in powers of v_{0}^{r}, the first non-vanishing derivative of the lhs is the \left(r-1\right)-th, giving N^{\prime }\left(-r!\right)<0. Thus, this equation is satisfied at least in the neighborhood of 0^{+}. Indeed, for v_{0}\ll 1, we can expandK\left(r-1\rig...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.008233148604631424, 0.035648997873067856, -0.02650783769786358, 0.02646205574274063, -0.018236538395285606, 0.014131409116089344, 0.046850353479385376, 0.00021126533101778477, 0.014749467372894287, 0.01381856482475996, -0.004555319435894489, 0.004768969025462866, -0.06446120142936707, 0...
5a6ae90c35f19714484f137ea42d5c21c83a96ce
subsection
23
92
The Domain
Inserting it into the error bound, we have-R_{1,q}^{\left(N\right)} & <\sqrt{\frac{2K}{GN^{\prime }\left(K+N^{\prime }rv_{0}^{r-1}\right)}}\left(1-\frac{1}{G}\right)^{N^{\prime }}\sum _{n=N^{\prime }}^{\infty }Q_{1}^{n}=\sqrt{\frac{2Kv_{0}}{G\left(N^{\prime }\right)^{2}}}\left(1-\frac{1}{G}\right)^{N^{\prime }}\frac{Q_...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.030436959117650986, 0.05028582364320755, -0.0007313643000088632, -0.04107082262635231, 0.0020157811231911182, -0.008101265877485275, 0.027400892227888107, -0.017163699492812157, -0.005908126477152109, 0.018140122294425964, -0.006892178673297167, 0.007246895227581263, -0.009672697633504868...
3933b535e612269e51b8a08ebd68a4d124575210
subsection
24
92
The Domain
We may write\left|R_{2,q}^{\left(N\right)}\right| & =\sqrt{\frac{2K}{G}}\frac{K^{N^{\prime }}}{N^{\prime }!}\left(1-\frac{1}{G}\right)^{N^{\prime }}\int _{1}^{\infty }e^{-Kv}v^{N^{\prime }}\left(v^{r}-1\right)^{N^{\prime }}\frac{dv}{\sqrt{v}}\\ & <\sqrt{\frac{2K}{G}}\frac{K^{N^{\prime }}}{N^{\prime }!}\left(1-\frac{1}{...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.028065526857972145, 0.051491618156433105, 0.020572230219841003, -0.030766775831580162, -0.012117470614612103, -0.004139626864343882, 0.0018800011603161693, -0.007462774869054556, 0.021548952907323837, 0.013567293994128704, -0.02252567559480667, 0.01593279466032982, -0.007367391604930162, ...
8ccbd575aff625edccb412d673ddb98aac007268
subsection
25
92
The Domain
Note that we now have to evaluate the same expression for R_{2}^{\left(N\right)} as we did for q=4 in (), only multiplied by r^{N^{\prime }}. Since q<4, we have r=\frac{q-2}{2}<1, and thus r^{N^{\prime }}<1. This means that the requirements imposed on K\left(N\right), and consequently on M\left(N\right), are more relax...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.03722968325018883, 0.06286323815584183, -0.0006441764999181032, 0.007377275265753269, -0.02604551985859871, 0.009101437404751778, 0.018569067120552063, 0.02088829316198826, 0.02001858316361904, 0.021819036453962326, -0.022719262167811394, -0.022200487554073334, -0.01252687256783247, 0.0...
06f70aa5ec03e522d01f94f9c9ea8ab99bca05ea
subsection
26
92
Domain of Convergence
Summing the magnitudes of the remainders from both domains, we get a total error bound ofR^{\left(N\right)} & =\left|\mathcal {Z}^{\left(N\right)}-\mathcal {Z}\right|=\left|R_{1}^{\left(N\right)}+R_{2}^{\left(N\right)}\right|<\left|R_{1}^{\left(N\right)}\right|+\left|R_{2}^{\left(N\right)}\right|\\ & <\sqrt{\frac{2}{G}...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.01863643154501915, 0.03596022352576256, -0.03461705520749092, -0.007421756628900766, 0.010325590148568153, -0.02037644200026989, 0.00320909870788455, -0.031259141862392426, 0.009516637772321701, 0.02385646291077137, -0.05506981536746025, 0.0062121436931192875, -0.019002748653292656, 0.0...
b4f861e2bb9db4b8cf14e5e3c93ca2dd74de4692
subsection
27
92
Domain of Convergence
This would require us to pick a minimal value of \alpha >\ln 2\approx 0.693.(iii) The last term scales as\frac{2^{N}N^{N}}{N!}\left(\frac{2N}{M}+1\right)^{N}e^{-2N-\frac{M}{2}-\frac{M^{2}}{2N+M}} & \sim \left(2\left(\frac{2}{\alpha }+1\right)e^{-1-\frac{\alpha }{2}-\frac{\alpha ^{2}}{\alpha +2}}\right)^{N}\,.The expone...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.03695235401391983, 0.0111070666462183, -0.04397055506706238, -0.003871453460305929, -0.013311697170138359, -0.02163436822593212, -0.01574518159031868, 0.011114695109426975, -0.0005392381572164595, -0.0003532939590513706, -0.03274143487215042, 0.04952409118413925, -0.04915792495012283, -...
85992660db5844583d29c1bef905c38c91599725
subsection
28
92
Domain of Convergence
This case is discussed further in .In total, we expect the large-N error to scale asymptotically asR^{\left(N\right)}=\mathcal {O}\left(10^{-A\left(\alpha \right)N-B\left(\alpha ,g\right)\sqrt{N}}\right)\,,where the bound on A\left(\alpha \right) isA\left(\alpha \right)={\left\lbrace \begin{array}{ll} -\log _{10}\left(...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0077199190855026245, 0.011511223390698433, -0.04964548721909523, -0.010450879111886024, -0.03092544712126255, -0.007704662624746561, -0.003461376763880253, -0.01621793396770954, -0.00800216943025589, -0.009360020980238914, -0.026424704119563103, 0.02018469013273716, -0.02875898778438568, ...
2649538645782baf39aaf7f9c357ab5cf694991e
subsection
29
92
Domain of Convergence
The error will then be dominated by the first term, and have the functional form\sqrt{\frac{2}{G}}\sqrt{\frac{M^{2}}{N^{3}}}\left(1-\frac{1}{G}\right)^{N}\left[\frac{M}{M+N}\right]^{N} & \sim N^{-\frac{p}{4}}N^{p-\frac{3}{2}}\left(1-\frac{1}{N^{p/2}}\right)^{N}\left(1-\frac{N}{N^{p}}\right)^{N}\sim N^{\frac{3}{4}\left(...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.023576315492391586, 0.00880487635731697, -0.013497250154614449, -0.022950666025280952, -0.00682873884215951, 0.006061936262995005, 0.010811533778905869, -0.013550659641623497, -0.02911560609936714, 0.006882147863507271, -0.028718851506710052, 0.005749111529439688, -0.008629389107227325, ...
519bf8792c08c12f6a4f02eee056bdf4dfb2fc46
subsection
30
92
Domain of Convergence
It is bounded by the first term in the sum, l=0 (which is positive), so we have\mathcal {Z}^{\left(N\right)} & <\sqrt{\frac{2}{G\left(K\left(N\right)\right)}}\sum _{n=0}^{N}\left[1-\frac{1}{G\left(K\left(N\right)\right)}\right]^{n}\frac{\Gamma \left(n+\frac{1}{2}\right)}{n!}\\ & <\sqrt{\frac{2}{G\left(K\left(N\right)\r...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0200037881731987, 0.037383127957582474, 0.018310103565454483, -0.04131980240345001, -0.007270636968314648, -0.0032347850501537323, 0.039641372859478, -0.008636265993118286, -0.012801813893020153, 0.01157351117581129, -0.032653018832206726, 0.014854071661829948, -0.02203315868973732, -0....
846090ac89211e59f073a134c2cde736c1d59153
subsection
31
92
Comparison with Results for Related Methods
It is instructive to compare the results we have shown here with those obtained for the ODM/OPT/LDE schemes. Following arguments by Zinn-Justin and Seznec , Buckley, Duncan, and Jones showed that the sequence \mathcal {Z}^{\left(N\right)} converges to \mathcal {Z}\left(g\right) if the modified harmonic coefficient G sc...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 462, "openalex_id": "", "raw": "J. Zinn-Justin and R. Seznec, J. Math. Phys. 20, 1398 (1979).", "source_ref_id": "2595fec7292f1ea07fbdfe7f791c394666cd058c", "start": 109 }, { "arxiv_id": "", "doi": ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.02165156602859497, -0.01687571033835411, -0.062009841203689575, -0.019759533926844597, 0.012061707675457, -0.006835730746388435, 0.003978608641773462, -0.033995553851127625, 0.040434565395116806, 0.048399414867162704, -0.03927493467926979, 0.014014773070812225, 0.014907385222613811, 0.0...
aefdfdc242036052ebd1d1bf02371e7a56c68891
subsection
32
92
Numerical Results
In order to demonstrate the properties of SCE, the expansion in () was evaluated in Mathematica , and was compared against a direct evaluation of (). Mathematica was chosen by virtue of its ability to evaluate both to arbitrary numerical precision . However, this precluded the usage of floating-point values of g and \a...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 149, "openalex_id": "", "raw": "Wolfram Research, Inc., “Mathematica, Version 11.2,” Champaign, IL, 2017.", "source_ref_id": "49ea443e10222148a16bbc236e1acb863227fe24", "start": 0 }, { "arxiv_id": "", ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.03736111894249916, 0.00881374441087246, -0.05185992270708084, 0.016711775213479996, 0.0020069393794983625, -0.0172459427267313, 0.003905137535184622, 0.00809643603861332, 0.020771440118551254, 0.05875829979777336, -0.010400981642305851, 0.03839892894029617, -0.00583004392683506, 0.00706...
376a7b25c0132b91b5e010c0b5a46eb36aa3cce5
subsection
33
92
Numerical Results
The minimal error shown is roughly 1.5\times 10^{-92}.Note that in order to obtain this many accurate digits, the intermediatecalculations had to be performed with over 300 digit precision.]REF depicts the convergence properties of the SCE as a function of \alpha . It shows the error of the expansion for two orders, N=...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1808, "openalex_id": "", "raw": "I. R. C. Buckley, A. Duncan, and H. F. Jones, Phys. Rev. D 47, 2554 (1993).", "source_ref_id": "939c4a588d4b82d0d306dfa9bae7c7c6c042f3ca", "start": 1469 }, { "arxiv_id": "...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04830910637974739, 0.00851436611264944, -0.045348916202783585, -0.008796651847660542, 0.006401033140718937, 0.0023193254601210356, 0.0132827153429389, -0.0011949485633522272, 0.0258024949580431, 0.05813572183251381, -0.028259148821234703, 0.045532021671533585, -0.014488155022263527, 0.0...
8bb31c7c95b458c6c1dcb9df799b0d1198948a14
subsection
34
92
Numerical Results
The fitted values for the parameter A are 0.072, 0.200 and 0.283, for \alpha =1 , 2, and \frac{4}{3}, respectively. These are in agreement with the bounding values 0.018, 0.176, and 0.243, given by ().We continue with a comparison of the SCE with other asymptotic and numerical approximation schemes. These include the m...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 460, "openalex_id": "", "raw": "J. P. Boyd, Acta Applicandae Mathematica 56, 1 (1999).", "source_ref_id": "a1f0dc9b3e8b3359250931a8cd2526252333c171", "start": 301 }, { "arxiv_id": "", "doi": "", ...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.024835431948304176, 0.01120492909103632, -0.0552542582154274, -0.009755686856806278, 0.03028153069317341, -0.06327848136425018, 0.028313612565398216, -0.009290403686463833, 0.013020295649766922, 0.03911427780985832, -0.038748156279325485, 0.023965885862708092, 0.005007512401789427, 0.01...
67a94c836df64df06f945b02bda0f50324817e5c
subsection
35
92
Numerical Results
To make a “fair” comparison, the SCE, Padé, and \tau methods are evaluated at N=N_{0}, though in principle they converge as N\rightarrow \infty .A striking result of REF is the similarity of the errors produced by SCE (at order N_{0}=\frac{1}{16g}) and hyperasymptotics. This can be explained by the error estimates of b...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1570, "openalex_id": "", "raw": "M. V. Berry and C. J. Howls, Proc. R. Soc. A 430, 653 (1990).", "source_ref_id": "dfee5969e4d320f8d915ee156a5c53b119ff41a5", "start": 1346 }, { "arxiv_id": "", "doi"...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.03516208380460739, -0.00942386593669653, -0.05213266983628273, -0.009317035786807537, 0.019152957946062088, -0.01823727786540985, 0.02292250283062458, 0.024845430627465248, 0.005383430980145931, 0.047554273158311844, -0.004414336755871773, 0.023990796878933907, -0.015856511890888214, -0...
992395e2da0863f469dc27142f7ec55486b34969
subsection
36
92
Numerical Results
If compared with hyperasymptotics when carried through to its conclusion, halting at roughly 2N_{0} terms with an error of order e^{-2.386N_{0}}, then at order 2N_{0} the SCE would result in an error of order \left(1-\frac{1}{\frac{1}{2}+\frac{1}{2}\sqrt{1+2\times \alpha }}\right)^{2N_{0}}\left(\frac{\alpha }{\alpha +1...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04071997478604317, -0.006204148754477501, -0.06117153540253639, -0.006040078587830067, 0.025350777432322502, -0.0434977225959301, -0.004826720803976059, 0.007944057695567608, -0.0152699900791049, 0.04938899353146553, -0.03488975390791893, -0.004574892111122608, -0.02184043452143669, 0.0...
8e1cca9317678f4c41c09e8928c9f55a1e9430d6
subsection
37
92
Numerical Results
Both the Padé approximants and the \tau approximations are rational functions of g, where the numerator and denominator are of identical order in g; as such, for a fixed order N, they will tend to a constant as g\rightarrow \infty . \mathcal {Z}\left(g\right) decays to zero for g\rightarrow \infty (the anharmonicity co...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04659232869744301, 0.021007739007472992, -0.055105265229940414, -0.01106071937829256, 0.03734708949923515, -0.036096084862947464, 0.03560788929462433, -0.008482427336275578, 0.003117978572845459, 0.06590662896633148, -0.017819199711084366, -0.02984105795621872, -0.00023372920986730605, ...
bd9f07cbd2769728639ac849eabd5179adeb2066
subsection
38
92
Non-Perturbative SCE: The Double-Well Potential
Now that the convergent nature of the SCE has been observed, we can examine a more intricate case, that of the double-well potential, corresponding to a negative quadratic part of the potential (\gamma <0 in ()). This has the effect of flipping the sign of the quadratic term in (). It is an interesting test case, since...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1119/1.1972842", "end": 909, "openalex_id": "https://openalex.org/W2120062331", "raw": "M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, Vol. 55 (Dover, New York, 1964).", "source_ref_id": "d88a70f14cea1e8d188fa5...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.043985217809677124, 0.013720580376684666, -0.03473641723394394, -0.006089556496590376, 0.024114033207297325, -0.0555843748152256, 0.03052409365773201, 0.019001247361302376, 0.013079574331641197, 0.04575561732053757, -0.016650892794132233, 0.051127854734659195, -0.020451143383979797, -0....
ff489270a87249c80d7670628755b22614e2670e
subsection
39
92
Non-Perturbative SCE: The Double-Well Potential
This is of course unavoidable, since at g=0 the potential is not bound from below and the partition function cannot be defined.We attribute the initial divergence of the SCE at low orders to an incompatibility between two conflicting goals. The first is that the SCE zeroth order potential V_{0}\left(x\right)=\frac{1}{2...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.0254458487033844, 0.04213515296578407, -0.06626904010772705, 0.021662533283233643, 0.006605546921491623, -0.023279594257473946, 0.030586885288357735, 0.017162218689918518, 0.00000494605592393782, 0.04280638322234154, -0.01835213229060173, 0.008573480881750584, -0.011372829787433147, -0....
243db9d9f54163b940b86e57934cdb6c715eb8a8
subsection
40
92
Non-Perturbative SCE: The Double-Well Potential
As \gamma \rightarrow 0^{+}, g\rightarrow \infty and so does G. As \gamma crosses zero, G\rightarrow \infty for \gamma \rightarrow 0^{-}, but as \left|\gamma \right| increases, G descends from infinity along the double-well branch in ().Strikingly, we have shown that the SCE provides a means to write down a perturbat...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 915, "openalex_id": "", "raw": "I. R. C. Buckley, A. Duncan, and H. F. Jones, Phys. Rev. D 47, 2554 (1993).", "source_ref_id": "939c4a588d4b82d0d306dfa9bae7c7c6c042f3ca", "start": 711 }, { "arxiv_id": "",...
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.04498652368783951, 0.030336910858750343, -0.03882147744297981, 0.0271323062479496, 0.0005846493295393884, -0.07727671414613724, 0.05099897086620331, 0.011658651754260063, -0.004463554359972477, 0.04095787927508354, -0.030581070110201836, 0.020707840099930763, -0.005196035373955965, -0.0...
ea4ab117d34fe5db450019dfb7b1fcaf1bfba31d
subsection
41
92
General Power-Law Perturbations
A SCE may be performed in the case of a general perturbation g\left|x\right|^{q}. We assume that q>2, and that g may be complex but has a positive real part. Expanding again around a modified \frac{1}{2}Gx^{2} harmonic oscillator, the partition function is now\mathcal {Z}^{\left(N\right)}=\sqrt{\frac{2}{G}}\sum _{n=0}^...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.02658437192440033, 0.022662337869405746, -0.035740870982408524, -0.026859065517783165, 0.03128470852971077, -0.05206996202468872, 0.007489253766834736, -0.014093379490077496, -0.00914123933762312, 0.04120424762368202, -0.009064935147762299, -0.004597325809299946, -0.005310769658535719, ...
3b73acc8e1485e1ddfba0c7762e1763bc574e513
subsection
42
92
General Power-Law Perturbations
If G-1 is small but M\gg q (i.e., gM^{\frac{q}{2}-1} is small but M is large), then a simpler limit applies,G\approx 1+2^{\frac{q}{2}}gM^{\frac{q}{2}-1}\,.Going back to the expansion for \mathcal {Z}, it now becomes\mathcal {Z}^{\left(N\right)}=\sqrt{\frac{2}{G}}\sum _{n=0}^{N}\left(1-\frac{1}{G}\right)^{n}\sum _{l=0}^...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.06609059870243073, 0.03023812547326088, 0.0004843897477257997, -0.03930040821433067, 0.019940076395869255, 0.0055304341949522495, -0.00847491342574358, 0.0037168331909924746, -0.00862747710198164, -0.0019032321870326996, -0.04811859130859375, -0.0016190822934731841, -0.007159051485359669,...
cf97a06d501bcc59bd22b1abd191b2e76e0a3de7
subsection
43
92
General Power-Law Perturbations
The error can once again be calculated by\lim _{N\rightarrow \infty }\mathcal {Z}^{\left(N\right)} & =\lim _{N\rightarrow \infty }\int _{-\infty }^{\infty }e^{-\frac{1}{2}Gx^{2}}\sum _{n=0}^{N}\frac{1}{n!}\left(-\left[\frac{1}{2}\left(1-G\right)x^{2}+gx^{q}\right]\right)^{n}dx\\ & =\lim _{N\rightarrow \infty }\sqrt{\fr...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.05608803778886795, 0.0310803409665823, -0.01911814883351326, -0.01492985524237156, 0.0064006890170276165, -0.005866662599146366, 0.027601540088653564, -0.03152282163500786, 0.004184478893876076, 0.03117188811302185, -0.020216716453433037, -0.003761072177439928, -0.014266136102378368, 0....
d95227f31d23e2e1188a373cd95dc99fcfa9401b
subsection
44
92
Rates and Domains of Convergence
Collecting the contributions from all domains and cases, we find the total bound on the error to beR_{q}^{\left(N\right)} & <\sqrt{\frac{2}{G}}\left(1-\frac{1}{G}\right)^{N^{\prime }}{\left\lbrace \begin{array}{ll} \sqrt{\frac{Kv_{0}}{\left(N^{\prime }\right)^{2}}}\frac{Q_{1}^{N^{\prime }}}{1-Q_{1}}+\frac{\left(\left(r...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.007011271547526121, 0.023330215364694595, -0.0241083987057209, -0.010314732789993286, 0.002429913030937314, -0.027038026601076126, 0.0017985933227464557, -0.04067910835146904, 0.0005736234597861767, 0.03338555246591568, -0.021941693499684334, -0.003801268758252263, -0.015403435565531254, ...
5e47ed164de88f68dee7fc998600a6d5da7a7abc
subsection
45
92
Rates and Domains of Convergence
For r>1, from \mathcal {D}_{1} we have Q_{1}<1 given by (), while from \mathcal {D}_{2} we haveQ_{2}=\alpha ^{-r}\left(\frac{r+1}{e}\right)^{r+1}e\Rightarrow \alpha _{c}=\frac{1}{e}\left(r+1\right)^{1+\frac{1}{r}}\,.For r\le 1, Q_{1}=\frac{\alpha }{\alpha +1} and \mathcal {D}_{2} has two components, which areQ_{2A}=e^{...
{ "cite_spans": [] }
10.1103/PhysRevD.98.056017
1803.06631
A Deal with the Devil: From Divergent Perturbation Theory to an Exponentially-Convergent Self-Consistent Expansion
[ "Benjamin Remez", "Moshe Goldstein" ]
[ "math-ph", "cond-mat.stat-mech", "hep-th", "math.MP" ]
2,018
en
Physics
[ -0.021805444732308388, 0.04635754972696304, -0.06567573547363281, 0.009834575466811657, 0.007267209701240063, -0.017014045268297195, 0.029694467782974243, 0.023544996976852417, 0.004428992513567209, 0.06079278513789177, -0.011978497728705406, -0.007637246046215296, -0.04394659027457237, 0....