The Landscape of the Spiked Tensor Model

We consider the problem of estimating a large rank‐one tensor u⊗k ∈ (ℝn)⊗k, k ≥ 3, in Gaussian noise. Earlier work characterized a critical signal‐to‐noise ratio λ  Bayes = O(1) above which an ideal estimator achieves strictly positive correlation with the unknown vector of interest. Remarkably, no...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Communications on pure and applied mathematics Ročník 72; číslo 11; s. 2282 - 2330
Hlavní autoři: Arous, Gérard Ben, Mei, Song, Montanari, Andrea, Nica, Mihai
Médium: Journal Article
Jazyk:angličtina
Vydáno: Melbourne John Wiley & Sons Australia, Ltd 01.11.2019
John Wiley and Sons, Limited
Témata:
ISSN:0010-3640, 1097-0312
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract We consider the problem of estimating a large rank‐one tensor u⊗k ∈ (ℝn)⊗k, k ≥ 3, in Gaussian noise. Earlier work characterized a critical signal‐to‐noise ratio λ  Bayes = O(1) above which an ideal estimator achieves strictly positive correlation with the unknown vector of interest. Remarkably, no polynomial‐time algorithm is known that achieved this goal unless λ ≥ Cn(k − 2)/4, and even powerful semidefinite programming relaxations appear to fail for 1 ≪ λ ≪ n(k − 2)/4. In order to elucidate this behavior, we consider the maximum likelihood estimator, which requires maximizing a degree‐k homogeneous polynomial over the unit sphere in n dimensions. We compute the expected number of critical points and local maxima of this objective function and show that it is exponential in the dimensions n, and give exact formulas for the exponential growth rate. We show that (for λ larger than a constant) critical points are either very close to the unknown vector u or are confined in a band of width Θ(λ−1/(k − 1)) around the maximum circle that is orthogonal to u. For local maxima, this band shrinks to be of size Θ(λ−1/(k − 2)). These “uninformative” local maxima are likely to cause the failure of optimization algorithms. © 2019 Wiley Periodicals, Inc.
AbstractList We consider the problem of estimating a large rank‐one tensor u⊗k ∈ (ℝn)⊗k, k ≥ 3, in Gaussian noise. Earlier work characterized a critical signal‐to‐noise ratio λ  Bayes = O(1) above which an ideal estimator achieves strictly positive correlation with the unknown vector of interest. Remarkably, no polynomial‐time algorithm is known that achieved this goal unless λ ≥ Cn(k − 2)/4, and even powerful semidefinite programming relaxations appear to fail for 1 ≪ λ ≪ n(k − 2)/4.In order to elucidate this behavior, we consider the maximum likelihood estimator, which requires maximizing a degree‐k homogeneous polynomial over the unit sphere in n dimensions. We compute the expected number of critical points and local maxima of this objective function and show that it is exponential in the dimensions n, and give exact formulas for the exponential growth rate. We show that (for λ larger than a constant) critical points are either very close to the unknown vector u or are confined in a band of width Θ(λ−1/(k − 1)) around the maximum circle that is orthogonal to u. For local maxima, this band shrinks to be of size Θ(λ−1/(k − 2)). These “uninformative” local maxima are likely to cause the failure of optimization algorithms. © 2019 Wiley Periodicals, Inc.
We consider the problem of estimating a large rank‐one tensor u ⊗ k  ∈ ( ℝ n ) ⊗ k , k  ≥ 3 , in Gaussian noise. Earlier work characterized a critical signal‐to‐noise ratio λ    Bayes  =  O (1) above which an ideal estimator achieves strictly positive correlation with the unknown vector of interest. Remarkably, no polynomial‐time algorithm is known that achieved this goal unless λ  ≥  Cn ( k  − 2)/4 , and even powerful semidefinite programming relaxations appear to fail for 1 ≪  λ  ≪  n ( k  − 2)/4 . In order to elucidate this behavior, we consider the maximum likelihood estimator, which requires maximizing a degree‐ k homogeneous polynomial over the unit sphere in n dimensions. We compute the expected number of critical points and local maxima of this objective function and show that it is exponential in the dimensions n , and give exact formulas for the exponential growth rate. We show that (for λ larger than a constant) critical points are either very close to the unknown vector u or are confined in a band of width Θ( λ −1/( k  − 1) ) around the maximum circle that is orthogonal to u . For local maxima, this band shrinks to be of size Θ( λ −1/( k  − 2) ) . These “uninformative” local maxima are likely to cause the failure of optimization algorithms. © 2019 Wiley Periodicals, Inc.
We consider the problem of estimating a large rank‐one tensor u⊗k ∈ (ℝn)⊗k, k ≥ 3, in Gaussian noise. Earlier work characterized a critical signal‐to‐noise ratio λ  Bayes = O(1) above which an ideal estimator achieves strictly positive correlation with the unknown vector of interest. Remarkably, no polynomial‐time algorithm is known that achieved this goal unless λ ≥ Cn(k − 2)/4, and even powerful semidefinite programming relaxations appear to fail for 1 ≪ λ ≪ n(k − 2)/4. In order to elucidate this behavior, we consider the maximum likelihood estimator, which requires maximizing a degree‐k homogeneous polynomial over the unit sphere in n dimensions. We compute the expected number of critical points and local maxima of this objective function and show that it is exponential in the dimensions n, and give exact formulas for the exponential growth rate. We show that (for λ larger than a constant) critical points are either very close to the unknown vector u or are confined in a band of width Θ(λ−1/(k − 1)) around the maximum circle that is orthogonal to u. For local maxima, this band shrinks to be of size Θ(λ−1/(k − 2)). These “uninformative” local maxima are likely to cause the failure of optimization algorithms. © 2019 Wiley Periodicals, Inc.
Author Arous, Gérard Ben
Mei, Song
Nica, Mihai
Montanari, Andrea
Author_xml – sequence: 1
  givenname: Gérard Ben
  surname: Arous
  fullname: Arous, Gérard Ben
  email: benarous@cims.nyu.edu
  organization: Courant Institute, 251 Mercer St
– sequence: 2
  givenname: Song
  surname: Mei
  fullname: Mei, Song
  email: songmei@stanford.edu
  organization: Stanford University, 475 Via Ortega
– sequence: 3
  givenname: Andrea
  surname: Montanari
  fullname: Montanari, Andrea
  email: montanari@stanford.edu
  organization: Stanford University, 350, Serra Mall
– sequence: 4
  givenname: Mihai
  surname: Nica
  fullname: Nica, Mihai
  email: mnica@math.utoronto.ca
  organization: University of Toronto, 40 St. George St. Toronto
BookMark eNp1kE1Lw0AQhhepYFs9-A8CXvSQdr-yyR5L8QsqCsbzstnMYmrMxt0U6b93NT2JnoYZnvcdeGZo0rkOEDoneEEwpkvT6wUlhSBHaEqwzFPMCJ2gKcYEp0xwfIJmIWzjSnjBpuiyfIVko7s6GN1D4mwyxMNz37xBnZTQBeeTB1dDe4qOrW4DnB3mHL3cXJfru3TzeHu_Xm1SwwQjKQUuREEtAWG05EbmObaGayOkgbwSVS6txZrp2maVxRnLawmWsyoyvK4yNkcXY2_v3ccOwqC2bue7-FJRKgllhSxopJYjZbwLwYNVphn00Lhu8LppFcHqW4eKOtSPjpi4-pXoffOu_f5P9tD-2bSw_x9U66fVmPgCvD5vMg
CitedBy_id crossref_primary_10_1214_25_AAP2152
crossref_primary_10_1109_TSP_2019_2937282
crossref_primary_10_1073_pnas_2015665118
crossref_primary_10_1088_1742_5468_adbe40
crossref_primary_10_1146_annurev_statistics_112723_034548
crossref_primary_10_1214_20_AAP1636
crossref_primary_10_1109_TIT_2022_3152733
crossref_primary_10_1002_cpa_22146
crossref_primary_10_1214_23_AAP1958
crossref_primary_10_1002_cpa_22169
crossref_primary_10_1088_1751_8121_ac6d8e
crossref_primary_10_1109_TIT_2022_3191883
crossref_primary_10_1073_pnas_2100482118
crossref_primary_10_1214_23_AAP1962
crossref_primary_10_1093_ptep_ptad138
crossref_primary_10_1214_23_AOS2331
crossref_primary_10_1214_23_AOS2257
crossref_primary_10_1103_PhysRevX_10_011057
crossref_primary_10_1140_epjs_s11734_023_00844_2
crossref_primary_10_1016_j_spa_2022_04_013
crossref_primary_10_1088_1751_8121_ac564a
crossref_primary_10_1214_19_AAP1547
crossref_primary_10_5802_crmath_240
crossref_primary_10_1214_21_AOS2079
crossref_primary_10_1007_s00220_023_04739_0
crossref_primary_10_1007_s10955_019_02480_4
crossref_primary_10_1088_1742_5468_ad3a5b
crossref_primary_10_1007_s10107_022_01896_3
crossref_primary_10_1007_s00220_021_04027_9
crossref_primary_10_1214_25_AAP2187
crossref_primary_10_1063_5_0086919
crossref_primary_10_1214_20_AOP1443
crossref_primary_10_1088_1742_5468_ad2929
crossref_primary_10_1007_s11005_025_01983_4
crossref_primary_10_1214_19_AOP1415
crossref_primary_10_1137_24M1680489
crossref_primary_10_1088_1742_5468_abfa1e
crossref_primary_10_1007_JHEP01_2025_071
crossref_primary_10_1073_pnas_2023719118
crossref_primary_10_1080_01621459_2021_2002157
crossref_primary_10_1007_s10955_021_02855_6
crossref_primary_10_1214_24_AOS2484
crossref_primary_10_1214_21_AOS2146
crossref_primary_10_1007_s10955_022_02875_w
Cites_doi 10.1103/PhysRevLett.92.240601
10.1090/surv/089
10.1214/17-AOS1637
10.1016/j.jfa.2004.09.015
10.1109/TPAMI.2012.39
10.29007/964b
10.1007/s10208-017-9365-9
10.1109/TIT.2016.2637959
10.1002/cpa.21748
10.1007/s00440-005-0433-8
10.1145/2897518.2897529
10.1051/jp1:1995164
10.1007/s00440-016-0724-2
10.1214/12-AOS1018
10.1007/BF01312184
10.1007/s10208-015-9269-5
10.1214/16-AOP1139
10.1109/TIT.2010.2046205
10.1002/cpa.21638
10.1007/PL00008774
10.1214/EJP.v12-438
10.1002/cpa.21422
10.1007/s00222-017-0726-4
10.1002/widm.1
10.1214/13-AOP862
10.1190/geo2013-0022.1
10.1088/0266-5611/27/2/025010
ContentType Journal Article
Copyright 2019 Wiley Periodicals, Inc.
Copyright_xml – notice: 2019 Wiley Periodicals, Inc.
DBID AAYXX
CITATION
JQ2
DOI 10.1002/cpa.21861
DatabaseName CrossRef
ProQuest Computer Science Collection
DatabaseTitle CrossRef
ProQuest Computer Science Collection
DatabaseTitleList ProQuest Computer Science Collection
CrossRef

DeliveryMethod fulltext_linktorsrc
Discipline Mathematics
EISSN 1097-0312
EndPage 2330
ExternalDocumentID 10_1002_cpa_21861
CPA21861
Genre article
GroupedDBID --Z
.3N
.GA
.Y3
05W
0R~
10A
1L6
1OB
1OC
1ZS
31~
33P
3SF
3WU
4.4
4ZD
50Y
50Z
51W
51X
52M
52N
52O
52P
52S
52T
52U
52W
52X
5GY
5VS
66C
6J9
6OB
6TJ
702
7PT
8-0
8-1
8-3
8-4
8-5
8UM
930
A03
AAESR
AAEVG
AAHHS
AAHQN
AAMNL
AANHP
AANLZ
AAONW
AASGY
AAXRX
AAYCA
AAZKR
ABCQN
ABCUV
ABDBF
ABEFU
ABEML
ABIJN
ABLJU
ABTAH
ACAHQ
ACBEA
ACBWZ
ACCFJ
ACCZN
ACGFO
ACGFS
ACIWK
ACPOU
ACRPL
ACSCC
ACXBN
ACXQS
ACYXJ
ADBBV
ADEOM
ADIZJ
ADKYN
ADMGS
ADNMO
ADOZA
ADXAS
ADZMN
ADZOD
AEEZP
AEIGN
AEIMD
AENEX
AEQDE
AEUQT
AEUYR
AFBPY
AFFPM
AFGKR
AFPWT
AFWVQ
AFZJQ
AHBTC
AI.
AIAGR
AITYG
AIURR
AIWBW
AJBDE
AJXKR
ALAGY
ALMA_UNASSIGNED_HOLDINGS
ALUQN
ALVPJ
AMBMR
AMYDB
ASPBG
ATUGU
AUFTA
AVWKF
AZBYB
AZFZN
AZVAB
BAFTC
BDRZF
BFHJK
BHBCM
BMNLL
BNHUX
BROTX
BRXPI
BY8
CS3
D-E
D-F
DCZOG
DPXWK
DR2
DRFUL
DRSTM
DU5
EBS
EJD
F00
F01
F04
FEDTE
FSPIC
G-S
G.N
GNP
GODZA
H.T
H.X
HBH
HF~
HGLYW
HHY
HVGLF
HZ~
H~9
IX1
J0M
JPC
KQQ
LATKE
LAW
LC2
LC3
LEEKS
LH4
LITHE
LOXES
LP6
LP7
LUTES
LW6
LYRES
M6L
MEWTI
MK4
MRFUL
MRSTM
MSFUL
MSSTM
MXFUL
MXSTM
N04
N05
N9A
NF~
NNB
O66
O9-
OHT
P2P
P2W
P2X
P4D
PALCI
PQQKQ
Q.N
Q11
QB0
QRW
R.K
RIWAO
RJQFR
ROL
RWI
RX1
RYL
S10
SAMSI
SUPJJ
TN5
TWZ
UB1
UHB
V2E
VH1
W8V
W99
WBKPD
WH7
WIB
WIH
WIK
WOHZO
WQJ
WRC
WWM
WXSBR
WYISQ
XBAML
XG1
XPP
XV2
YZZ
ZY4
ZZTAW
~IA
~WT
AAMMB
AAYXX
ABUFD
ADXHL
AEFGJ
AETEA
AEYWJ
AGHNM
AGQPQ
AGXDD
AGYGG
AIDQK
AIDYY
AIQQE
AMVHM
CITATION
O8X
JQ2
ID FETCH-LOGICAL-c3631-2e46682f1e6ca94c9770fc4ac69ce7b6b79ff0a3adf5bf0537d9ef43bc4a4db53
IEDL.DBID DRFUL
ISICitedReferencesCount 67
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000486251500002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 0010-3640
IngestDate Fri Jul 25 19:25:22 EDT 2025
Sat Nov 29 04:07:13 EST 2025
Tue Nov 18 22:04:19 EST 2025
Wed Jan 22 16:38:18 EST 2025
IsPeerReviewed true
IsScholarly true
Issue 11
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c3631-2e46682f1e6ca94c9770fc4ac69ce7b6b79ff0a3adf5bf0537d9ef43bc4a4db53
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
PQID 2291238982
PQPubID 48818
PageCount 49
ParticipantIDs proquest_journals_2291238982
crossref_citationtrail_10_1002_cpa_21861
crossref_primary_10_1002_cpa_21861
wiley_primary_10_1002_cpa_21861_CPA21861
PublicationCentury 2000
PublicationDate November 2019
PublicationDateYYYYMMDD 2019-11-01
PublicationDate_xml – month: 11
  year: 2019
  text: November 2019
PublicationDecade 2010
PublicationPlace Melbourne
PublicationPlace_xml – name: Melbourne
– name: New York
PublicationTitle Communications on pure and applied mathematics
PublicationYear 2019
Publisher John Wiley & Sons Australia, Ltd
John Wiley and Sons, Limited
Publisher_xml – name: John Wiley & Sons Australia, Ltd
– name: John Wiley and Sons, Limited
References 2010; 56
2017; 63
2001; 120
2011; 1
2013; 66
2010
2013; 41
2017; 45
2009
1998
2005
2017; 210
2016; 16
2007; 12
2006; 134
1995; 5
2018; 46
2018; 18
2004; 92
2017; 70
2005; 222
2013; 35
1993; 92
2013; 78
2000; 33
2018
2017
2016
2015
2014
2018; 71
2011; 27
2017; 168
2012; 40
Bhattiprolu V. (e_1_2_1_10_1) 2016
Bhattiprolu V. (e_1_2_1_11_1) 2016
Gillin P. (e_1_2_1_19_1) 2000; 33
e_1_2_1_42_1
e_1_2_1_20_1
e_1_2_1_41_1
Hopkins S. B. (e_1_2_1_22_1) 2015
e_1_2_1_40_1
Janzamin M. (e_1_2_1_23_1) 2015
e_1_2_1_24_1
Montanari A. (e_1_2_1_34_1) 2014
e_1_2_1_21_1
e_1_2_1_44_1
e_1_2_1_43_1
Bouchaud J.‐P. (e_1_2_1_12_1) 1998
e_1_2_1_25_1
e_1_2_1_26_1
Lesieur T. (e_1_2_1_27_1) 2017
Li N. (e_1_2_1_28_1) 2010
e_1_2_1_29_1
Adler R. J. (e_1_2_1_2_1) 2009
Barak B. (e_1_2_1_7_1) 2016
Barbier J. (e_1_2_1_8_1) 2017
Zhong K. (e_1_2_1_45_1) 2017
Ge R. (e_1_2_1_18_1) 2016
e_1_2_1_31_1
e_1_2_1_30_1
e_1_2_1_5_1
e_1_2_1_6_1
e_1_2_1_35_1
e_1_2_1_13_1
e_1_2_1_33_1
e_1_2_1_32_1
Anderson G. W. (e_1_2_1_4_1) 2010
e_1_2_1_16_1
e_1_2_1_39_1
e_1_2_1_17_1
e_1_2_1_38_1
e_1_2_1_14_1
e_1_2_1_15_1
e_1_2_1_36_1
Anandkumar A. (e_1_2_1_3_1) 2016
Perry A. (e_1_2_1_37_1) 2016
e_1_2_1_9_1
References_xml – volume: 18
  start-page: 1131
  issue: 5
  year: 2018
  end-page: 1198
  article-title: A geometric analysis of phase retrieval
  publication-title: Found. Comput. Math.
– year: 2009
– year: 2016
  article-title: Statistical limits of spiked tensor models
  publication-title: Preprint
– volume: 1
  start-page: 24
  issue: 1
  year: 2011
  end-page: 40
  article-title: Applications of tensor (multiway array) factorizations and decompositions in data mining
  publication-title: Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery
– volume: 71
  start-page: 2381
  issue: 11
  year: 2018
  end-page: 2425
  article-title: Spectral algorithms for tensor completion
  publication-title: Comm. Pure Appl. Math.
– volume: 78
  start-page: V273
  issue: 6
  year: 2013
  end-page: V284
  article-title: Tensor completion based on nuclear norm minimization for 5D seismic data reconstruction
  publication-title: Geophysics
– volume: 63
  start-page: 1572
  issue: 3
  year: 2017
  end-page: 1579
  article-title: On the limitation of spectral methods: from the Gaussian hidden clique problem to rank one perturbations of Gaussian tensors
  publication-title: IEEE Trans. Inform. Theory
– year: 2005
– volume: 35
  start-page: 208
  issue: 1
  year: 2013
  end-page: 220
  article-title: Tensor completion for estimating missing values in visual data
  publication-title: IEEE Transactions on Pattern Analysis and Machine Intelligence
– volume: 222
  start-page: 435
  issue: 2
  year: 2005
  end-page: 490
  article-title: A Fourier view on the R‐transform and related asymptotics of spherical integrals
  publication-title: J. Funct. Anal.
– year: 2017
  article-title: Recovery guarantees for one‐hidden‐layer neural networks
  publication-title: Preprint
– start-page: 161
  year: 1998
  end-page: 223
– volume: 168
  start-page: 773
  issue: 3‐4
  year: 2017
  end-page: 820
  article-title: The extremal process of critical points of the pure ‐spin spherical spin glass model
  publication-title: Probab. Theory Related Fields
– volume: 70
  start-page: 822
  issue: 5
  year: 2017
  end-page: 883
  article-title: Solving random quadratic systems of equations is nearly as easy as solving linear systems
  publication-title: Comm. Pure Appl. Math.
– start-page: 2897
  year: 2014
  end-page: 2905
  article-title: A statistical model for tensor PCA
  publication-title: Advances in Neural Information Processing Systems
– volume: 16
  start-page: 1031
  issue: 4
  year: 2016
  end-page: 1068
  article-title: On tensor completion via nuclear norm minimization
  publication-title: Found. Comput. Math.
– year: 2015
  article-title: Beating the perils of non‐convexity: Guaranteed training of neural networks using tensor methods
  publication-title: Preprint
– start-page: 956
  year: 2015
  end-page: 1006
  article-title: Tensor principal component analysis via sum‐of‐square proofs
  publication-title: Conference on Learning Theory
– year: 2017
  article-title: The layered structure of tensor estimation and its mutual information
  publication-title: Preprint
– year: 2010
– volume: 56
  start-page: 2980
  issue: 6
  year: 2010
  end-page: 2998
  article-title: Matrix completion from a few entries
  publication-title: IEEE Trans. Inform. Theory
– volume: 210
  start-page: 135
  issue: 1
  year: 2017
  end-page: 209
  article-title: The geometry of the Gibbs measure of pure spherical spin glasses
  publication-title: Invent. Math.
– volume: 12
  start-page: 1131
  year: 2007
  end-page: 1150
  article-title: Large deviations for the largest eigenvalue of rank one deformations of Gaussian ensembles
  publication-title: Electron. J. Probab.
– volume: 120
  start-page: 1
  issue: 1
  year: 2001
  end-page: 67
  article-title: Aging of spherical spin glasses
  publication-title: Probab. Theory Related Fields
– volume: 92
  start-page: 257
  issue: 2
  year: 1993
  end-page: 271
  article-title: The spherical ‐spin interaction spin‐glass model
  publication-title: Zeitschrift für Physik B Condensed Matter
– start-page: 2973
  year: 2016
  end-page: 2981
  article-title: Matrix completion has no spurious local minimum
  publication-title: Advances in Neural Information Processing Systems
– year: 2016
  article-title: Multiplicative approximations for polynomial optimization over the unit sphere
  publication-title: Preprint
– start-page: 178
  year: 2016
  end-page: 191
– year: 2017
  article-title: Statistical and computational phase transitions in spiked tensor estimation
  publication-title: Preprint
– year: 2010
  article-title: Tensor completion for on‐board compression of hyperspectral images. 517–520
  publication-title: IEEE
– volume: 40
  start-page: 1637
  issue: 3
  year: 2012
  end-page: 1664
  article-title: High‐dimensional regression with noisy and missing data: Provable guarantees with nonconvexity
  publication-title: Ann. Statist.
– volume: 46
  start-page: 2747
  issue: 6A
  year: 2018
  end-page: 2774
  article-title: The landscape of empirical risk for non‐convex losses
  publication-title: Ann. Statist.
– year: 2018
  article-title: Complex energy landscapes in spiked‐tensor and simple glassy models: ruggedness arrangements of local minima and phase transitions
  publication-title: Preprint
– volume: 27
  start-page: 19
  issue: 2
  year: 2011
  article-title: Tensor completion and low‐ ‐rank tensor recovery via convex optimization
  publication-title: Inverse Problems
– volume: 41
  start-page: 4214
  issue: 6
  year: 2013
  end-page: 4247
  article-title: Complexity of random smooth functions on the high‐dimensional sphere
  publication-title: Ann. Probab.
– volume: 92
  start-page: 4
  issue: 24
  year: 2004
  article-title: Complexity of random energy landscapes, glass transition, and absolute value of the spectral determinant of random matrices
  publication-title: Phys. Rev. Lett
– volume: 33
  start-page: 3081
  issue: 16
  year: 2000
  end-page: 3091
  article-title: spin glasses with first‐order ferromagnetic transitions
  publication-title: J. Phys
– volume: 66
  start-page: 165
  issue: 2
  year: 2013
  end-page: 201
  article-title: Random matrices and complexity of spin glasses
  publication-title: Comm. Pure Appl. Math
– volume: 134
  start-page: 339
  issue: 3
  year: 2006
  end-page: 382
  article-title: Free energy of the spherical mean field model
  publication-title: Probab. Theory Related Fields
– volume: 5
  start-page: 805
  issue: 7
  year: 1995
  end-page: 813
  article-title: Thouless‐Anderson‐Palmer approach to the spherical ‐spin spin glass model
  publication-title: Journal de Physique I
– year: 2016
  article-title: Certifying random polynomials over the unit sphere via sum of squares hierarchy
  publication-title: Preprint
– start-page: 417
  year: 2016
  end-page: 445
  article-title: Noisy tensor completion via the sum‐of‐squares hierarchy
  publication-title: Conference on Learning Theory
– volume: 45
  start-page: 3385
  issue: 5
  year: 2017
  end-page: 3450
  article-title: The complexity of spherical ‐spin models—a second moment approach
  publication-title: Ann. Probab.
– year: 2016
  article-title: Homotopy analysis for tensor PCA
  publication-title: Preprint
– start-page: 161
  volume-title: Spin glasses and random fields
  year: 1998
  ident: e_1_2_1_12_1
– year: 2010
  ident: e_1_2_1_28_1
  article-title: Tensor completion for on‐board compression of hyperspectral images. 2010 IEEE International Conference on Image Processing 517–520
  publication-title: IEEE
– ident: e_1_2_1_16_1
  doi: 10.1103/PhysRevLett.92.240601
– ident: e_1_2_1_26_1
  doi: 10.1090/surv/089
– volume-title: An introduction to random matrices
  year: 2010
  ident: e_1_2_1_4_1
– year: 2017
  ident: e_1_2_1_27_1
  article-title: Statistical and computational phase transitions in spiked tensor estimation
  publication-title: Preprint
– ident: e_1_2_1_32_1
  doi: 10.1214/17-AOS1637
– ident: e_1_2_1_20_1
  doi: 10.1016/j.jfa.2004.09.015
– ident: e_1_2_1_29_1
  doi: 10.1109/TPAMI.2012.39
– ident: e_1_2_1_38_1
  doi: 10.29007/964b
– ident: e_1_2_1_42_1
  doi: 10.1007/s10208-017-9365-9
– ident: e_1_2_1_33_1
  doi: 10.1109/TIT.2016.2637959
– year: 2017
  ident: e_1_2_1_45_1
  article-title: Recovery guarantees for one‐hidden‐layer neural networks
  publication-title: Preprint
– year: 2016
  ident: e_1_2_1_37_1
  article-title: Statistical limits of spiked tensor models
  publication-title: Preprint
– ident: e_1_2_1_35_1
  doi: 10.1002/cpa.21748
– volume-title: Random fields and geometry
  year: 2009
  ident: e_1_2_1_2_1
– ident: e_1_2_1_43_1
  doi: 10.1007/s00440-005-0433-8
– ident: e_1_2_1_21_1
  doi: 10.1145/2897518.2897529
– start-page: 956
  year: 2015
  ident: e_1_2_1_22_1
  article-title: Tensor principal component analysis via sum‐of‐square proofs
  publication-title: Conference on Learning Theory
– ident: e_1_2_1_15_1
  doi: 10.1051/jp1:1995164
– ident: e_1_2_1_41_1
  doi: 10.1007/s00440-016-0724-2
– ident: e_1_2_1_30_1
  doi: 10.1214/12-AOS1018
– ident: e_1_2_1_14_1
  doi: 10.1007/BF01312184
– ident: e_1_2_1_44_1
  doi: 10.1007/s10208-015-9269-5
– ident: e_1_2_1_39_1
  doi: 10.1214/16-AOP1139
– ident: e_1_2_1_24_1
  doi: 10.1109/TIT.2010.2046205
– year: 2016
  ident: e_1_2_1_11_1
  article-title: Certifying random polynomials over the unit sphere via sum of squares hierarchy
  publication-title: Preprint
– ident: e_1_2_1_13_1
  doi: 10.1002/cpa.21638
– start-page: 2897
  year: 2014
  ident: e_1_2_1_34_1
  article-title: A statistical model for tensor PCA
  publication-title: Advances in Neural Information Processing Systems
– ident: e_1_2_1_9_1
  doi: 10.1007/PL00008774
– year: 2016
  ident: e_1_2_1_10_1
  article-title: Multiplicative approximations for polynomial optimization over the unit sphere
  publication-title: Preprint
– volume: 33
  start-page: 3081
  issue: 16
  year: 2000
  ident: e_1_2_1_19_1
  article-title: p > 2 spin glasses with first‐order ferromagnetic transitions
  publication-title: J. Phys
– ident: e_1_2_1_31_1
  doi: 10.1214/EJP.v12-438
– ident: e_1_2_1_6_1
  doi: 10.1002/cpa.21422
– year: 2017
  ident: e_1_2_1_8_1
  article-title: The layered structure of tensor estimation and its mutual information
  publication-title: Preprint
– ident: e_1_2_1_40_1
  doi: 10.1007/s00222-017-0726-4
– ident: e_1_2_1_36_1
  doi: 10.1002/widm.1
– ident: e_1_2_1_5_1
  doi: 10.1214/13-AOP862
– start-page: 417
  year: 2016
  ident: e_1_2_1_7_1
  article-title: Noisy tensor completion via the sum‐of‐squares hierarchy
  publication-title: Conference on Learning Theory
– year: 2015
  ident: e_1_2_1_23_1
  article-title: Beating the perils of non‐convexity: Guaranteed training of neural networks using tensor methods
  publication-title: Preprint
– ident: e_1_2_1_25_1
  doi: 10.1190/geo2013-0022.1
– year: 2016
  ident: e_1_2_1_3_1
  article-title: Homotopy analysis for tensor PCA
  publication-title: Preprint
– start-page: 2973
  year: 2016
  ident: e_1_2_1_18_1
  article-title: Matrix completion has no spurious local minimum
  publication-title: Advances in Neural Information Processing Systems
– ident: e_1_2_1_17_1
  doi: 10.1088/0266-5611/27/2/025010
SSID ssj0011483
Score 2.60387
Snippet We consider the problem of estimating a large rank‐one tensor u⊗k ∈ (ℝn)⊗k, k ≥ 3, in Gaussian noise. Earlier work characterized a critical signal‐to‐noise...
We consider the problem of estimating a large rank‐one tensor u ⊗ k  ∈ ( ℝ n ) ⊗ k , k  ≥ 3 , in Gaussian noise. Earlier work characterized a critical...
SourceID proquest
crossref
wiley
SourceType Aggregation Database
Enrichment Source
Index Database
Publisher
StartPage 2282
SubjectTerms Algorithms
Critical point
Economic models
Mathematical analysis
Maxima
Maximum likelihood estimators
Optimization
Polynomials
Random noise
Semidefinite programming
Tensors
Title The Landscape of the Spiked Tensor Model
URI https://onlinelibrary.wiley.com/doi/abs/10.1002%2Fcpa.21861
https://www.proquest.com/docview/2291238982
Volume 72
WOSCitedRecordID wos000486251500002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVWIB
  databaseName: Wiley Online Library
  customDbUrl:
  eissn: 1097-0312
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0011483
  issn: 0010-3640
  databaseCode: DRFUL
  dateStart: 19960101
  isFulltext: true
  titleUrlDefault: https://onlinelibrary.wiley.com
  providerName: Wiley-Blackwell
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV3dS8MwED_G5oM--C1OpwTxYS91XZKmDT6N6fBhjqEb7K0kaQLDsY1t-veb9GsKCoJvhV7S5rjL_a7p_Q7gViRtriLKPRJR41HFAk_ikHsiUYYkIVOMyrTZRDgYRJMJH1bgvqiFyfghyg9uzjPS_do5uJDr1pY0VC3FnWuoZFOfmiuqsplX7eGlN-6XhwgW6WcHzG6rYdQviIV83CoHfw9HW4z5FammoaZ38K-XPIT9HGGiTmYSR1DR82PYey7pWdcn0LTGgfquyNf9_oQWBtmb6HU5fdMJGtnEdrFCrkna7BTGvcdR98nLWyZ4ijDS9rCmjEXYtDVTglNl0Z1vFBWKcaVDyWTIjfEFEYkJpHFcLgnXhhJpZWgiA3IG1flirs8BRZxyLYVgjnFN00D4kjBBtLIB37GU1aFZaC5WOZ-4a2sxizMmZBzbxcfp4utwU4ouMxKNn4Qahfrj3I_WMcbchtaIR9g-LlX07xPE3WEnvbj4u-gl7FoExLPiwgZUN6t3fQU76mMzXa-uc4P6BFI6zBs
linkProvider Wiley-Blackwell
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV3dS8MwED_GJqgPfovTqUV82Etd16RpA76M6ZjYjaEb7K0kaQLDsY1t-veb9GsKCoJvhV7S5rjL_fJxvwO4ZXGTigBTGwVY2VgQz-auT20WC4VinwiCeVJswu_3g_GYDkpwn-fCpPwQxYab8YxkvjYObjakGxvWULFgd6aikl77VDBBflCGysNLZxQWpwga6qcnzGauIdjJmYUct1E0_h6PNiDzK1RNYk1n_39_eQB7Gca0WqlRHEJJzo5gt1cQtK6Ooa7NwwpNmq-5AGXNlaVfWq-LyZuMraFe2s6XlimTNj2BUedx2O7aWdEEWyCCmrYrMSGBq5qSCEax0PjOUQIzQaiQPifcp0o5DLFYeVwZNpeYSoUR1zI45h46hfJsPpNnYAUUU8kZI4ZzTWKPORwRhqTQId_wlFWhnqsuEhmjuClsMY1SLmQ30oOPksFX4aYQXaQ0Gj8J1XL9R5knrSLXpTq4BjRw9ecSTf_eQdQetJKH87-LXsN2d9gLo_Cp_3wBOxoP0TTVsAbl9fJdXsKW-FhPVsurzLo-AZzA0As
linkToPdf http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1bS8MwFD6MTUQfvIvTqUV82Etd16RpA76MaVGcY-gGeytJmsBwbGWb_n6T3qagIPhW6Oklh3NyvjQ93wdwzeI2FQGmNgqwsrEgns1dn9osFgrFPhEE81Rswu_3g_GYDipwW_TCZPwQ5Qc3kxnpfG0SXCaxaq1ZQ0XCboyikl771LBHPVyF2t1LOOqVuwga6mc7zGauIdgpmIUct1Ve_L0erUHmV6ia1ppw939vuQc7Oca0OllQ7ENFzg5g-7kkaF0eQlOHh9Uzbb7mByhrrix90npNJm8ytoZ6aTtfWEYmbXoEo_B-2H2wc9EEWyCC2rYrMSGBq9qSCEax0PjOUQIzQaiQPifcp0o5DLFYeVwZNpeYSoUR1zY45h46hupsPpMnYAUUU8kZI4ZzTWKPORwRhqTQJd_wlNWhWbguEjmjuBG2mEYZF7Ib6cFH6eDrcFWaJhmNxk9GjcL_UZ5Jy8h1qS6uAQ1c_bjU07_fIOoOOunB6d9NL2FzcBdGvcf-0xlsaThEs07DBlRXi3d5DhviYzVZLi7y4PoEbXPPhg
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=The+Landscape+of+the+Spiked+Tensor+Model&rft.jtitle=Communications+on+pure+and+applied+mathematics&rft.au=G%C3%A9rard+Ben+Arous&rft.au=Song%2C+Mei&rft.au=Montanari%2C+Andrea&rft.au=Nica%2C+Mihai&rft.date=2019-11-01&rft.pub=John+Wiley+and+Sons%2C+Limited&rft.issn=0010-3640&rft.eissn=1097-0312&rft.volume=72&rft.issue=11&rft.spage=2282&rft.epage=2330&rft_id=info:doi/10.1002%2Fcpa.21861&rft.externalDBID=NO_FULL_TEXT
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0010-3640&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0010-3640&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0010-3640&client=summon