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
06f4017c29c9cef6c869571fd8744b5c22d4d56d
subsection
31
69
Polling Model and Notation
\sum _{y}\sigma (\pi , y^{(u+1)} = y) V\left(T\left(\pi ,y^{(u+1)}=y\right) \right) \ge \\ \sum _{y}\sum _{r}\Lambda (r)V\left(T(\pi ,y^{(u)} = r)\right)\sigma (\pi , y^{(u+1)} = y)\\ = \sum _{r}V\left(T\left(\pi ,y^{(u)}=r\right)\right) {\sigma \left(\pi ,y^{(u)} = r\right)} \therefore C(\pi ,1) \le C(\pi ,u) ~\foral...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/ciss.2012.6310920", "end": 693, "openalex_id": "https://openalex.org/W2128428699", "raw": "M. Naghshvar and T. Javidi, “Active hypothesis testing: Sequentiality and adaptivity gains,” in 46th Annual Conference on Information Science...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.025851670652627945, 0.046636778861284256, -0.04794920235872269, -0.0476439893245697, 0.027194615453481674, 0.01279611885547638, 0.0007720973808318377, 0.013604938052594662, 0.03973893076181412, 0.037083566188812256, -0.029117466881871223, 0.013337874785065651, 0.012941095978021622, 0.02...
7cb0150e5388df633ba9773b439986e7ce708c8a
subsection
32
69
Polling Model and Notation
Therefore Q^{j/K} \succeq _B Q^{(j+J)/K}. For Theorem REF c, we have Q^{j/K} = Q^{j/K+1} \times Q^{j/K(K+1)} . Therefore Q^{j/K} \succeq _B Q^{j/K+1}. EM Algorithm with Ultrametric Constraints The parameters of the POMDP are computed using a sequence of observations obtained from level {N} in Fig.REF . Specifically, a...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 524, "openalex_id": "", "raw": "A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via the em algorithm,” Journal of the Royal Statistical Society. Series B (methodological), pp. 1–38, 1977.", ...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03863154724240303, 0.012373690493404865, -0.024808410555124283, -0.011778654530644417, -0.0126178078353405, 0.007258672267198563, 0.019590405747294426, 0.00296945683658123, 0.0014923571143299341, 0.03857051581144333, -0.03350508585572243, 0.0352749340236187, 0.008772961795330048, 0.0475...
377a218c0aeffcf01946f2c4240bfb4b45151421
subsection
33
69
Adaptive Polling and Blackwell Dominance
Armed with the POMDP formulation of the previous section, in this section we give sufficient conditions for Blackwell dominance of Theorem REF to hold for three polling mechanisms: (i) Adaptive Intent Polling (Sec.REF ), (ii) Adaptive Expectation Polling (Sec.REF ), (iii) Adaptive Friendship Polling (Sec.REF ).The form...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.2139/ssrn.1884644", "end": 621, "openalex_id": "https://openalex.org/W3124505561", "raw": "D. M. Rothschild and J. Wolfers, “Forecasting elections: Voter intentions versus expectations,” 2011.", "source_ref_id": "25ffb1d8db11f1e492...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.008874243125319481, 0.008691111579537392, 0.007889911532402039, -0.009614399634301662, -0.016634436324238777, -0.02510426566004753, 0.01239189226180315, -0.011483865790069103, 0.023684997111558914, 0.02766810543835163, -0.049842268228530884, 0.026233574375510216, -0.01777900755405426, -...
a79b732cf706b34a9b8cef6550695d4002cee888
subsection
34
69
Adaptive Intent Polling
In adaptive intent polling, a node at level l is sampled with a probability \beta _l and is asked the following question:``\textit {What does it (a node at level) think the state is?}" This polling mechanism is a more sophisticated version of standard intent polling, for multiple states and hierarchical social networks...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03744722530245781, -0.029222263023257256, -0.014435645192861557, 0.03418165445327759, 0.0017205275362357497, -0.037264108657836914, 0.008057104423642159, -0.01644992083311081, 0.01489343587309122, 0.03463944420218468, -0.028764473274350166, 0.00879719853401184, -0.0012846732279285789, 0...
3e6f8db6fc520aa2dfefcbfbbd2279dbecd16771
subsection
35
69
Adaptive Intent Polling
The cost (REF ) expressed in terms of the belief state \pi captures the fact that a control with higher measurement cost should result in a smaller entropy (more reduction in uncertainty) cost and vice versa. Main Result. Myopic Policies for Adaptive Intent Polling Our main result on adaptive intent polling is Theorem ...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.021051522344350815, -0.033255305141210556, -0.03685541823506355, 0.01459877286106348, 0.01983114518225193, -0.02247021161019802, -0.019571814686059952, -0.01754293590784073, 0.03831987455487251, 0.059890057891607285, -0.05802898108959198, 0.012165644206106663, 0.012249545194208622, 0.01...
2bdae4bf0e83efa859467e6f09e1d740d87dfcc5
subsection
36
69
Adaptive Intent Polling
Proposition REF below provides a justification for the polynomial f_{U}(z) to be Hurwitz. If f_{U}(z) is Hurwitz, then a way to compute f_{g}(z) for g \in \lbrace U-1,\cdots ,2,1 \rbrace is by successive long-division of f_{U}(z) by linear or quadratic factors of f_{U}(z ). Matrix polynomials and Blackwell Dominance ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1090/s0002-9939-1991-1072329-2", "end": 1230, "openalex_id": "https://openalex.org/W2001120034", "raw": "R. Barnard, W. Dayawansa, K. Pearce, and D. Weinberg, “Polynomials with nonnegative coefficients,” Proceedings of the American Mathe...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.01667948067188263, 0.024065453559160233, -0.022310523316264153, -0.019533151760697365, 0.01715254969894886, -0.013642686419188976, 0.033420003950595856, 0.0002804075484164059, 0.02166959084570408, 0.03802860528230667, -0.018846439197659492, 0.03012378327548504, -0.017076248303055763, 0....
2351a446e4d98c63b45713ae9eb909c84ec8ebd9
subsection
37
69
Adaptive Intent Polling
Corollary 1.1 If the channel error probabilities (likelihoods) for the pollster satisfy f_u(B) \succeq _B f_{u+1}(B) ~\forall u \in \mathcal {U}, then i.) Shannon Capacity Ordering: \mathcal {C}^{(u)} \ge \mathcal {C}^{(u+1)}~\forall u \in \mathcal {U}. ii.) Rényi Divergence Ordering: D_\alpha (f^i_u(B)|| f^j_u(B)) \g...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.2139/ssrn.1884644", "end": 1523, "openalex_id": "https://openalex.org/W3124505561", "raw": "D. M. Rothschild and J. Wolfers, “Forecasting elections: Voter intentions versus expectations,” 2011.", "source_ref_id": "25ffb1d8db11f1e49...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.010867573320865631, -0.010508882813155651, -0.027748944237828255, 0.014156846329569817, 0.028603699058294296, 0.01674400083720684, 0.0007006875821389258, -0.03635752946138382, -0.00976097397506237, 0.027581047266721725, -0.022406738251447678, -0.009035959839820862, 0.0030049909837543964, ...
682edef90e28865b4591854062a375c0a7ccae28
subsection
38
69
Adaptive Intent Polling
\end{aligned} In (REF ), \pi denotes the posterior distribution updated according to (\ref {eq:FIL}) and \bar{x} \in \lbrace e_1,e_2,\cdots , e_X\rbrace , where e_i is the unit indicator vector. In (REF ) using the law of iterated expectation, \eta _2(\pi ,u) can be expressed in terms of the belief \pi as follows : \...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/bf02055574", "end": 521, "openalex_id": "https://openalex.org/W2065087844", "raw": "V. Krishnamurthy, Partially Observed Markov Decision Processes. Cambridge University Press, 2016.", "source_ref_id": "11e6b835558c8897f3331eda...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.040373630821704865, -0.018508395180106163, -0.028136422857642174, 0.04598870873451233, 0.0007533816969953477, -0.01991216465830803, -0.019500188529491425, -0.00045632044202648103, 0.021178608760237694, 0.04778919368982315, -0.0609724186360836, 0.018676236271858215, 0.009742464870214462, ...
acbc00969d8864871c9e06612b5e9e128969ee8d
subsection
39
69
Adaptive Intent Polling
The proof of Theorem REF follows from Proposition REF below and Theorem REF . The expectation polling mechanism employed by the pollster determines how the opinions are gathered, and the opinions are distributed as O(u) for the pollster. Proposition REF below provides a justification for the opinion distribution B to b...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1016/j.laa.2010.04.007", "end": 986, "openalex_id": "https://openalex.org/W2118918399", "raw": "N. J. Higham and L. Lin, “On pth roots of stochastic matrices,” Linear Algebra and its Applications, vol. 435, no. 3, pp. 448–463, 2011.", ...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.01397506333887577, -0.012708763591945171, -0.04235237464308739, -0.0029063860420137644, -0.004096402321010828, -0.027385631576180458, -0.0008829768630675972, 0.004927888046950102, 0.013578391633927822, 0.046044476330280304, -0.027507685124874115, 0.03338148444890976, -0.010191421955823898...
bbbd2705dc726d4b199ec13baf39d96abdbd1ae2
subsection
40
69
Adaptive Intent Polling
Corollary 2.1 If the channel error probabilities (likelihoods) for the pollster satisfy Q^{l_{u} / K} \succeq _B Q^{l_{v} / K} for any K>0, we have i.) Shannon Capacity Ordering: \mathcal {C}^{(l_u)} \ge \mathcal {C}^{(l_v)} for l_u>l_v. ii.) Rényi Divergence Ordering: D_\alpha (Q^{l_{u} / K}_i|| Q^{l_{u} / K}_j) \ge ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1541, "openalex_id": "", "raw": "B. Nettasinghe and V. Krishnamurthy, “What Do Your Friends Think? Efficient Polling Methods for Networks Using Friendship Paradox,” arXiv preprint arXiv:1802.06505, 2018.", "source_ref_id": "...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03782309591770172, -0.0069754458963871, -0.019644198939204216, -0.008929181843996048, -0.0033923257142305374, 0.004174583125859499, 0.02248322032392025, -0.043836940079927444, -0.011844521388411522, 0.03431247919797897, -0.033976681530475616, -0.0037014128174632788, -0.0000590270428801886...
774b47da376a00c9524c5340dc9109b8470320f1
subsection
41
69
Adaptive Intent Polling
Since more informative opinion fractions are costlier to obtain, from (REF ) below, S(u) \ge S(u+1)~\forall u \in \mathcal {U}. ii.) State-Estimation error: The state-estimation error incurred in choosing action u is modelled as in (REF ) \begin{aligned}\eta _2(\pi ,u) = w_u\left(1 - \pi ^\prime \pi \right) \end{align...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.029988236725330353, -0.01568850316107273, -0.032903123646974564, 0.03030872344970703, 0.004631771240383387, -0.0292709618806839, 0.032903123646974564, -0.004524942487478256, 0.031804319471120834, 0.04700446501374245, -0.014459976926445961, 0.014215798117220402, -0.023929545655846596, 0....
443eac616aec68045433d1ef47ee15a5787b3633
subsection
42
69
Adaptive Intent Polling
\times \cdots \times {n}^{(j)}_X!} & \prod _{h=1}^{X} (B_l)_{ih}^{{n}^{(j)}_h}. Here \mathcal {N}_j and {n}^{(j)}_{i} indicate the total and the number in favor of x = i reported and B_l denotes the opinion distribution (REF ) at level l. The likelihood in (REF ) is the well known multinomial distribution. Remark: In ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 709, "openalex_id": "", "raw": "C. R. Blyth, “On simpson's paradox and the sure-thing principle,” Journal of the American Statistical Association, vol. 67, no. 338, pp. 364–366, 1972.", "source_ref_id": "05fd63b088c8d1243005...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03689757362008095, -0.017838405445218086, -0.02737561985850334, 0.004127706401050091, -0.011093688197433949, -0.020829277113080025, -0.00452826963737607, -0.011177615262567997, 0.03390670567750931, 0.022843535989522934, -0.06116024777293205, 0.009819515980780125, -0.04129232466220856, 0...
fc2706d9990f25ddf142584e5206d03d19ca98e3
subsection
43
69
Adaptive Intent Polling
Proposition 3 For the channel error probabilities (likelihoods) O(u) \succeq _B O(u+1), we have i.) Shannon Capacity Ordering: \mathcal {C}^{(u)} \ge \mathcal {C}^{(u+1)} for u \in \mathcal {U}. ii.) Rényi Divergence Ordering: D_\alpha ((O(u))_{i}|| (O(u))_{j}) \ge ~D_\alpha ((O(u+1))_{i}|| (O(u+1)_{j}) for all u \in ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 2284, "openalex_id": "", "raw": "V. Spokoiny and A. Shiryayev, Statistical Experiments And Decision, Asymptotic Theory. World Scientific, 2000, vol. 8.", "source_ref_id": "16c5a03729162f8f91aa33acee15fd9bc0661f83", "st...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.022668693214654922, -0.015285350382328033, -0.02050250582396984, -0.013790376484394073, 0.0038346857763826847, 0.022027989849448204, 0.014248021878302097, 0.007310882676392794, 0.015209076926112175, 0.04808326065540314, -0.04033380001783371, 0.009793607518076897, -0.006754080764949322, ...
2c56b4eb8a4ff9991ab323a7c87bc36b7651ad81
subsection
44
69
Adaptive Intent Polling
Initialize: O(1) = \hat{O}(1) For u \in \lbrace 1,2,\cdots ,U-1\rbrace , do: R^*_{u+1} = \operatornamewithlimits{arg\,min}_{R \in \mathcal {M}}\Vert O(u+1) - \hat{O}(u) R \Vert \hat{O}(u+1) = \hat{O}(u)R^*_{u+1} end Output: \hat{O}(u) for u \in \mathcal {U}. Consider a POMDP model \theta = (P,O(u),C,\rho ), where O(u)...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.04448392614722252, -0.016689101234078407, -0.04973169043660164, -0.0048282453790307045, 0.011471849866211414, -0.03960229083895683, -0.007642814889550209, 0.016872161999344826, 0.046345051378011703, 0.0372529998421669, -0.03984637185931206, 0.004786293487995863, 0.00012811916531063616, ...
5d6239abb2bf33a021c526ed78aaf3e63600ec03
subsection
45
69
Adaptive Intent Polling
\\ {\hspace{-17.07182pt}} \text{Mis-specified Policy:}&~J_{\mu ^*(\gamma )}(\pi ;\theta ) \le J_{\mu ^*(\theta )}(\pi ;\theta ) + 2G \Vert \gamma - \theta \Vert . Here G = \max _{i \in \mathcal {X},u} \frac{C(e_i,u)}{1-\rho } and e_i denotes the indicator vector with a `1' in the i^{th} position, and \Vert \gamma - ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/bf02055574", "end": 686, "openalex_id": "https://openalex.org/W2065087844", "raw": "V. Krishnamurthy, Partially Observed Markov Decision Processes. Cambridge University Press, 2016.", "source_ref_id": "11e6b835558c8897f3331eda...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.042634833604097366, -0.03090033493936062, -0.04321468994021416, 0.006355550605803728, 0.0017071481561288238, -0.03888101503252983, -0.009186173789203167, -0.01693796180188656, 0.025422202423214912, 0.05865722894668579, -0.051973599940538406, 0.019333690404891968, 0.0007829992100596428, ...
e05f2bb79a0b47b33bd99eff36aea5f1c5ff41a3
subsection
46
69
Adaptive Intent Polling
This implies that expectation polling is more informative than intent polling. For l_1>1, there is no apparent comparison of ultrametric and polynomial channels. However, Algorithm REF can be used to design polling POMDPs for arbitrary l_1~\text{and}~f_2. Proposition REF and hence Theorem REF provides the performance b...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/bf02055574", "end": 1774, "openalex_id": "https://openalex.org/W2065087844", "raw": "V. Krishnamurthy, Partially Observed Markov Decision Processes. Cambridge University Press, 2016.", "source_ref_id": "11e6b835558c8897f3331ed...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.033835191279649734, -0.04042527452111244, -0.013881887309253216, 0.035421695560216904, 0.011997916735708714, -0.03221818059682846, -0.022790702059864998, -0.01009869109839201, 0.04823574423789978, 0.04341522231698036, -0.053208816796541214, 0.006712121423333883, -0.01057159062474966, 0....
cca7897c61cd3c92a87e9fc9e8a0d4604f684237
subsection
47
69
Adaptive Intent Polling
We consider the following two measures for measuring the effectiveness of the myopic polling policy: (i) The percentage loss in optimality due to using the myopic policy \bar{\mu } instead of optimal policy \mu ^* is \begin{aligned}\mathcal {L}_1 = \cfrac{J_{\bar{\mu }}(\pi _0) - {J}_{\mu ^*}(\pi _0)}{{J}_{\mu ^*}(\pi...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1287/opre.39.1.162", "end": 896, "openalex_id": "https://openalex.org/W2149476049", "raw": "W. S. Lovejoy, “Computationally feasible bounds for partially observed Markov decision processes,” Operations research, vol. 39, no. 1, pp. 162–1...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.06531983613967896, -0.017444059252738953, -0.028325609862804413, 0.01655888371169567, -0.009660620242357254, 0.015528722666203976, -0.01488773338496685, 0.02901238389313221, 0.022633017972111702, 0.05063813179731369, -0.043984055519104004, -0.011476756073534489, -0.003380454145371914, 0...
70885d78bdc91e24e5c134219968825a583e590b
subsection
48
69
Adaptive Intent Polling
Also, the critics and those who see the movie before its release will influence the future movie goers by sharing opinions on social media platforms. So production, media house, and critics are in \text{Level}~0 and the common movie goer is \text{Level}~1. We will use adaptive expectation polling (Sec.REF ) to poll the...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03900359570980072, -0.02029529958963394, -0.02475110813975334, 0.031648457050323486, 0.03204520791769028, 0.022858915850520134, -0.004238360095769167, 0.00655400799587369, 0.0025922285858541727, 0.005794841796159744, -0.02919165976345539, 0.008301234804093838, -0.013909146189689636, 0.0...
6a8b39e50947325ac418b2c14b20f5d8285f71e1
subsection
49
69
Adaptive Intent Polling
\end{aligned} Note that the costs associated with the actions u=1 and u=2 in (REF ) assume the following structure: w_1 \le w_2 model the accuracy of the observations and S(1) \ge S(2) model the additional cost in expectation polling – nodes need to be compensated for exhausting their resources gathering information f...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.045176565647125244, -0.018849005922675133, -0.022130412980914116, -0.003290944965556264, -0.003689673962071538, -0.007345007732510567, 0.0063720326870679855, 0.012904319912195206, 0.04514604061841965, 0.03989579156041145, -0.061537809669971466, 0.027670646086335182, -0.007692226208746433,...
e52f07e266c869a7923a4ca1042a251d08b54f0c
subsection
50
69
Adaptive Intent Polling
The notion of Blackwell dominance was extended to the case of polynomial observation likelihoods (channels) described by matrix polynomials. Second, we presented an adaptive generalization of expectation polling to hierarchical social networks. The notion of Blackwell dominance was extended to the case of ultrametric o...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.027284633368253708, 0.008598932065069675, -0.005630888976156712, -0.013833065517246723, -0.010285147465765476, 0.02140958607196808, 0.006878382991999388, 0.021485885605216026, 0.013588907197117805, 0.05035754665732384, -0.02966517210006714, 0.00706913135945797, -0.03555547818541527, 0.0...
a6d4cb347ec19dd469aca6c54b65b682837ba45b
subsection
51
69
Adaptive Intent Polling
Observe that, \begin{aligned}T\left(\pi ,y^{(u+1)}=y\right) =& \cfrac{O_{u+1}(y)P^\prime \pi }{\sigma \left(\pi ,y^{(u+1)} = y\right)} =\sum _{r}\Lambda (r)T(\pi ,y^{(u)} = r) \end{aligned} where \Lambda (r) is a probability mass function w.r.t r and defined as \Lambda (r) = \mathbb {P}\left(y^{(u+1)}= y|y^{(u)} = ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/ciss.2012.6310920", "end": 1487, "openalex_id": "https://openalex.org/W2128428699", "raw": "M. Naghshvar and T. Javidi, “Active hypothesis testing: Sequentiality and adaptivity gains,” in 46th Annual Conference on Information Scienc...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.04809156805276871, 0.023542288690805435, -0.01995678059756756, -0.01945328339934349, 0.03347490727901459, 0.02425939030945301, -0.01769867353141308, 0.012030521407723427, 0.05721554160118103, 0.053126536309719086, -0.02848571166396141, 0.020933259278535843, -0.010603946633636951, 0.0332...
35ded951f31fef7fcbe2626fcffb647b4e7bd9db
subsection
52
69
Adaptive Intent Polling
We know that : {\vspace{0.0pt}}O(u) \succeq _B O(u+1) \Rightarrow \\ D(O_i(u)|| O_j(u)) \ge D(O_i(u+1)|| O_j(u+1)), for all i,j \in \mathcal {X}. From (REF ) and (REF ), the result follows. Proof of Theorem  REF: It is given that p(z) \in \mathcal {P}_n and q(z) \in \mathcal {P}_m, with n>m. Clearly, f(Q) and g(Q) ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 148, "openalex_id": "", "raw": "M. Sakaguchi, Information theory and decision making. Statistics Dept., George Washington University, 1964.", "source_ref_id": "c76a550a7d1e1e3c710a321e6956593ebf16c6a2", "start": 0 ...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.020506050437688828, 0.03213224932551384, -0.03585507348179817, -0.03854038938879967, 0.012373814359307289, 0.014334400184452534, 0.029355386272072792, 0.0026605227030813694, 0.001985379261896014, 0.03313923999667168, -0.032559458166360855, 0.026364920660853386, 0.03765545412898064, 0.01...
5310c38e6d94d80a9be65eb891455c4a8c1e7ebb
subsection
53
69
Adaptive Intent Polling
For all i,j,k \in \mathcal {X}, i \ne j \ne k: B_{N+1}(i,j) &\ge B_{N+1}(i,k) + M (1 - \kappa ), \\ B_{N+1}(i,j) &\ge B_{N+1}(k,j) + M \kappa , \\ B_{N+1}(k,j) & \ge B_{N+1}(i,k) + M (1 - \kappa ), \\ B_{N+1}(i,k) & \ge B_{N+1}(k,j) +M \kappa , \\ \kappa & \ge 0, \\ -\kappa & \ge -1, for some large positive value M. ...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ 0.013791929930448532, -0.002139732474461198, -0.02866707555949688, 0.0037302286364138126, 0.005961500573903322, -0.010511768981814384, 0.005156717263162136, 0.025569994002580643, -0.0007518623024225235, 0.005175787955522537, -0.03136748448014259, 0.01217473391443491, 0.012182362377643585, ...
3d42ee2513b95c1190fe610f0a97bb18f5a75e36
subsection
54
69
Approximate Blackwell Dominance, Performance Bounds, and Ordinal Sensitivity
So far we have discussed sufficient conditions for Blackwell dominance; when these conditions hold, the optimal adaptive polling policy is provably upper bounded by a myopic policy. This section discusses approximate Blackwell dominance and the performance loss due to this approximation.
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.0411750003695488, 0.0301105547696352, -0.019931264221668243, 0.0010425378568470478, -0.048988789319992065, -0.03056839480996132, 0.0024379934184253216, 0.024357043206691742, 0.04242642968893051, 0.03571145236492157, -0.034460026770830154, 0.013765697367489338, -0.004822572227567434, 0.0...
cee74efa42d99992a300951b3e2c7ad8173ef46d
subsection
55
69
Le Cam Deficiency
Given a collection of matrices, it is important to check whether there exists a Blackwell dominance relation, as Theorem REF can used to compute inexpensive policies. What if the pollster would like to choose between different polling mechanisms at each polling epoch to estimate the state?Now the control inputs u \in \...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 942, "openalex_id": "", "raw": "V. Spokoiny and A. Shiryayev, Statistical Experiments And Decision, Asymptotic Theory. World Scientific, 2000, vol. 8.", "source_ref_id": "16c5a03729162f8f91aa33acee15fd9bc0661f83", "sta...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03564019873738289, -0.006186686921864748, -0.025341767817735672, -0.00896344892680645, -0.017667530104517937, 0.0129607655107975, 0.017865870147943497, 0.007380541879683733, 0.01650800183415413, 0.05654219537973404, -0.035945337265729904, 0.009497441351413727, 0.013571042567491531, 0.03...
b7bd70b2852d4e7c83ae18a164db6fd87900f7e7
subsection
56
69
Performance Bounds on Comparison of Polling POMDPs
Let \theta = (P,O(u),C,\rho ) denote the given adaptive polling POMDP model and \gamma = (P,\hat{O}(u),C,\rho ) denote the adaptive polling POMDP model having a Blackwell dominance relation between the observation distributions. Let J_{\mu ^*(\gamma )}(\pi ;\theta ) and J_{\mu ^*(\gamma )}(\pi ;\gamma ) be defined as i...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/bf02055574", "end": 1957, "openalex_id": "https://openalex.org/W2065087844", "raw": "V. Krishnamurthy, Partially Observed Markov Decision Processes. Cambridge University Press, 2016.", "source_ref_id": "11e6b835558c8897f3331ed...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.045804813504219055, -0.03001268208026886, -0.06176478788256645, -0.0011128876358270645, -0.004005250986665487, -0.07152996957302094, -0.013648370280861855, 0.00804864801466465, 0.05291508883237839, 0.04391280934214592, -0.0427531935274601, 0.03001268208026886, 0.0021971664391458035, 0.0...
2b28c5797f58c9bf0ef7abe77560c29691a9d13d
subsection
57
69
Performance Bounds on Comparison of Polling POMDPs
Algorithm REF and Theorem REF can be used to design polling POMDPs that have observation distributions that are not Blackwell comparable, for example, when the polling distributions in case of adaptive intent polling are not Hurwitz; in case of adaptive friendship polling; to name a few.Proposition REF below highlights...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03518570587038994, -0.03506363928318024, -0.02973848208785057, 0.006873877253383398, -0.015014195814728737, -0.03918338939547539, -0.009994205087423325, -0.02078184485435486, 0.03503312170505524, 0.04583602026104927, -0.05804268270730972, 0.015685562044382095, -0.0010308908531442285, 0....
04a28caa77550908e56f945951b752276cc8196b
subsection
58
69
Ordering of Hierarchical Social Networks
So far we have discussed three types of polling mechanisms on a single hierarchical social network. In this section, we briefly discuss how to order hierarchical networks that differ in the opinion distributions B (defined in (REF )), according to the expected polling cost. Theorem REF below shows that some networks ar...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/bf02055574", "end": 1332, "openalex_id": "https://openalex.org/W2065087844", "raw": "V. Krishnamurthy, Partially Observed Markov Decision Processes. Cambridge University Press, 2016.", "source_ref_id": "11e6b835558c8897f3331ed...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.014259226620197296, -0.032012686133384705, -0.01358021516352892, 0.02496318332850933, 0.01802048832178116, -0.020950153470039368, -0.0072173504158854485, -0.01817307434976101, 0.03970305621623993, 0.04193082079291344, -0.04614221677184105, 0.0157774630934, -0.0056533366441726685, 0.0142...
3cac99cbba05d2144ba2f4b9bbfe1854c1fc9efa
subsection
59
69
Numerical Examples
The main results of this paper involve using Blackwell dominance to construct myopic policies that provably upper bound the optimal adaptive polling policy. In this section, the performance of this myopic upper bound is illustrated using numerical examples for adaptive polling. Let \Pi ^s_1 represent the set of belief ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1287/opre.39.1.162", "end": 1514, "openalex_id": "https://openalex.org/W2149476049", "raw": "W. S. Lovejoy, “Computationally feasible bounds for partially observed Markov decision processes,” Operations research, vol. 39, no. 1, pp. 162–...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.0706004798412323, -0.0022463095374405384, -0.03025081194937229, 0.0070859817788004875, -0.026391277089715004, -0.031059330329298973, 0.008199602365493774, 0.02604041062295437, 0.05647428333759308, 0.06050162389874458, -0.06535273790359497, 0.018473895266652107, 0.011181969195604324, 0.0...
653f3491ed54ff1a66465fc381b0453b7504544e
subsection
60
69
Numerical Examples
\end{aligned}In (REF ), the cumulative discounted cost is evaluated using 1000 Monte carlo simulations over a horizon of 100 time units.
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.03900553286075592, -0.012994213961064816, -0.0038093726616352797, -0.011361353099346161, -0.013192599639296532, 0.012177783995866776, 0.024248745292425156, 0.029666179791092873, 0.008500032126903534, 0.05389966443181038, -0.007283016573637724, 0.011948877945542336, -0.0347326286137104, ...
0f7aaaa188f0a1ec568f0f27015c33f0988b509e
subsection
61
69
Example 1: Market Research. Adaptive Expectation Polling with X=3, Y=3, U = 2 and N=1
We describe how to estimate the revenue level a movie generates based on the response received on social media platform YouTube. The popularity in the movie is modeled as a 3 state Markov chain x taking values on the state-space \mathcal {X}=\lbrace \text{High}, \text{Medium}, \text{Low} \rbrace . Prior to a movie's re...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.0412239208817482, -0.022015588358044624, -0.014387179166078568, 0.03954567015171051, 0.00984064768999815, 0.012403792701661587, -0.006423120386898518, 0.005027121398597956, -0.005351328756660223, 0.005095777101814747, -0.026653660461306572, 0.013357344083487988, -0.010626373812556267, 0...
75f65bfc590b359d99881f9237c3536994121a8f
subsection
62
69
Example 1: Market Research. Adaptive Expectation Polling with X=3, Y=3, U = 2 and N=1
The computed parameters for P, O(1) = O^{1/2}(2), and O(2) are as follows:\begin{aligned}\begin{pmatrix} 0.9089 & 0.0281 & 0.0630\\ 0.0346 & 0.9433 & 0.0221\\ 0.0065 & 0.0138 & 0.9797 \end{pmatrix}, \begin{pmatrix} 0.6382 & 0.1809 & 0.1809\\ 0.1809 & 0.6382 & 0.1809\\ 0.1809 & 0.1809 & 0.6382 \end{pmatrix}, \\ \begin{p...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.043799370527267456, -0.008988791145384312, -0.0494765006005764, 0.009210077114403248, 0.002199506852775812, -0.008034971542656422, 0.020770363509655, 0.00941610150039196, 0.030430642887949944, 0.03467322885990143, -0.04865240305662155, -0.005177329760044813, 0.027363160625100136, -0.000...
6cef616fe9b8e4775022b128b132ce37841349de
subsection
63
69
Example 2: Large Dimensional Example. Adaptive Intent Polling with X=20, Y=20, U = 5 and N = 9
The Blackwell dominance structural result is particularly useful for large number of states and observation symbols since solving the POMDP (for the optimal policy) is intractable. RandomThe matrices are generated by stochastic simulation as follows: twenty (1 \times 20) probability vectors were simulated from the Diri...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.05565092712640762, -0.003609759034588933, -0.029305215924978256, -0.005491867661476135, -0.018321480602025986, -0.007265283260494471, 0.024454066529870033, 0.03560560941696167, 0.05263039842247963, 0.04094492644071579, -0.06199708580970764, 0.04655883461236954, 0.008657319471240044, 0.0...
1a7971cebe0c8073c1e1c3882400199751010b9f
subsection
64
69
Conclusions
This paper considered the problem of adaptive (feedback control based) polling in hierarchical social networks, and the problem was formulated as a partially observed Markov decision process (POMDP). We presented three main results. First, we presented an adaptive generalization of intent polling to hierarchical social...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ 0.001341806841082871, -0.023772964254021645, -0.012878293171525002, -0.00502009317278862, -0.018585026264190674, -0.004577592480927706, 0.011932257562875748, 0.016113124787807465, 0.019378475844860077, 0.03814660385251045, -0.03979453817009926, -0.0028667172882705927, -0.020751751959323883, ...
ae4aeee74ff93efe720985255c3783b4f439ad8c
subsection
65
69
Proofs
Proof of Theorem  REF:Denote by y^{(u)} as the observations recorded when using action u. Then O(u+1) = O(u) R implies the following\mathbb {P}\left(y^{(u+1)}|x\right) = \sum _{y^{(u)}}\mathbb {P}\left(y^{(u+1)}|y^{(u)}\right)\mathbb {P}\left(y^{(u)}|x\right)For notational convenience, let T(\pi ,y,u) be written as T(\...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.011376844719052315, 0.021487053483724594, -0.03183380141854286, 0.01571851223707199, 0.013963532634079456, -0.002903346437960863, -0.02078506164252758, 0.032718922942876816, 0.03168119490146637, 0.02913265861570835, 0.002447433304041624, 0.0050322129391133785, 0.010835090652108192, 0.00...
3c7f45857cfa119ad9cbf90091c714f58cec8292
subsection
66
69
Proofs
Observe that,\begin{aligned}T\left(\pi ,y^{(u+1)}=y\right) =& \cfrac{O_{u+1}(y)P^\prime \pi }{\sigma \left(\pi ,y^{(u+1)} = y\right)} =\sum _{r}\Lambda (r)T(\pi ,y^{(u)} = r) \end{aligned}where \Lambda (r) is a probability mass function w.r.t r and defined as\Lambda (r) = \mathbb {P}\left(y^{(u+1)}= y|y^{(u)} = r\right...
{ "cite_spans": [] }
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.040272295475006104, 0.01833609864115715, -0.011715576983988285, -0.012302881106734276, 0.043231699615716934, 0.023858284577727318, -0.01435463223606348, 0.01479701790958643, 0.047075871378183365, 0.04759453237056732, -0.016917414963245392, 0.02253112941980362, -0.0017590527422726154, 0....
fbe33d376b8bad615adabdd3e5fe3c8eb0fc908d
subsection
67
69
Proofs
From the definition of Rényi Divergence (REF ) we have :{\hspace{-14.22636pt}}D_\alpha (O_i(u+1)|| O_j(u+1)) \le \min \Big \lbrace (1-\alpha ) D(O_i(u+1)|| O_j(u+1)), \\ \alpha D(O_j(u+1)|| O_i(u+1)) \Big \rbrace .We know that :{\vspace{0.0pt}}O(u) \succeq _B O(u+1) \Rightarrow \\ D(O_i(u)|| O_j(u)) \ge D(O_i(u+1)|| O_...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/ciss.2012.6310920", "end": 215, "openalex_id": "https://openalex.org/W2128428699", "raw": "M. Naghshvar and T. Javidi, “Active hypothesis testing: Sequentiality and adaptivity gains,” in 46th Annual Conference on Information Science...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.02488291822373867, 0.06462540477514267, -0.05693626403808594, -0.05714985355734825, -0.007795935496687889, 0.03743879869580269, 0.02474561147391796, -0.009359699673950672, 0.018170174211263657, 0.029520811513066292, -0.022624995559453964, 0.0020157299004495144, -0.010252188891172409, 0....
a9a95143236445031f56921abb2f732a2d2058fd
subsection
68
69
EM Algorithm with Ultrametric Constraints
The parameters of the POMDP are computed using a sequence of observations obtained from level {N} in Fig.REF . Specifically, a modified version of the EM algorithm  is used to compute the maximum likelihood estimate of the tuple \left(P, B_{N+1}\right), where B_{N+1} is restricted to the space of ultrametric stochastic...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 330, "openalex_id": "", "raw": "A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via the em algorithm,” Journal of the Royal Statistical Society. Series B (methodological), pp. 1–38, 1977.", ...
1810.00571
Adaptive Polling in Hierarchical Social Networks using Blackwell Dominance
[ "Sujay Bhatt", "Vikram Krishnamurthy" ]
[ "cs.SI" ]
2,018
en
Computer Science
[ -0.019098788499832153, -0.02004457451403141, -0.005117925815284252, 0.004770883359014988, -0.011784195899963379, -0.015109703876078129, 0.03166859596967697, 0.020639505237340927, 0.029944824054837227, 0.029273621737957, -0.05061483755707741, 0.05311659723520279, 0.021570036187767982, 0.023...
63f2b26139f4197b2dd13354503d4c04ff02e32f
abstract
0
27
Abstract
Smoothing (and decay) spacetime estimates are discussed for evolution groups of self-adjoint operators in an abstract setting. The basic assumption is the existence (and weak continuity) of the spectral density in a functional setting. Spectral identities for the time evolution of such operators are derived, enabling r...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.026705974712967873, 0.009408134035766125, -0.059485651552677155, 0.016862915828824043, 0.0035557099618017673, -0.04040995612740517, 0.0313146635890007, -0.017244430258870125, 0.06665811687707901, 0.029758086428046227, -0.005375531502068043, -0.0022871759720146656, 0.016817133873701096, ...
3f8d13cb28a7cf6a09630ad7b40cb95724f876f5
subsection
1
27
Introduction
In this paper we present an abstract framework for global spacetime and smoothing estimates for evolution groups generated by self-adjoint operators. In particular, this approach leads to such estimates for various classes of pseudodifferential operators.Global spacetime estimates for solutions of partial differential ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 618, "openalex_id": "", "raw": "R.S.Strichartz , Restriction of Fourier transform to quadratic surfaces and decay of solutions of wave equations, Duke Math.J. 44 (1977), 705–713.", "source_ref_id": "9cd355f12cb61a58a27d49262...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.04963250085711479, 0.02564040571451187, -0.05088399723172188, 0.008142354898154736, 0.016696788370609283, -0.02930332161486149, 0.03433983027935028, -0.025258852168917656, 0.06617666780948639, 0.023534230887889862, -0.025243591517210007, 0.01166791096329689, -0.017765138298273087, 0.018...
67e43a5630d74f34c0f93e261b2b270d5bbcdfca
subsection
2
27
The basic setup and notation
Notation. The following notations are used throughout the paper.{\left\langle {x}\right\rangle }=(1+|x|^2)^\frac{1}{2}. The Fourier transform in \mathbb {R}^n: \mathcal {F}f(\xi )=\widehat{f}(\xi )=(2\pi )^{-\frac{n}{2}}\int _{\mathbb {R}^n}f(x)e^{-i\xi x}\,dx. It is useful in multi-variable formulas to indicate by a...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.048290301114320755, 0.015285314992070198, -0.05934029445052147, 0.06459056586027145, 0.010653169825673103, -0.024374086409807205, -0.012141256593167782, 0.007417536806315184, 0.047466132789850235, 0.017841771245002747, -0.03226475790143013, -0.009271920658648014, 0.04300950467586517, 0....
6a842434fec39c7b95c3f5f0924eb2cb13b7c524
subsection
3
27
The basic setup and notation
The basic tool in our treatment is the hypothesis that the weak spectral derivativeA(\lambda )=\frac{d}{d\lambda }\Big (E(\lambda )P_{ac}(H)\Big ),\quad \lambda \in J,exists and is bounded from \mathcal {X} into \mathcal {X}^\ast . Thus{\left\langle {A(\lambda )f,g}\right\rangle }=\frac{d}{d\lambda }\Big (E(\lambda )P_...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.043976522982120514, 0.022186627611517906, -0.032898467034101486, -0.010933094657957554, 0.01976044289767742, -0.04797438904643059, 0.03088427521288395, 0.006176092196255922, 0.024933254346251488, 0.01823453977704048, -0.045563459396362305, 0.01672389544546604, 0.0077554015442729, 0.0202...
a13d1bf8e3d28921952f0f862fb943941bd835d8
subsection
4
27
Global smoothing estimates
The following assumption is fundamental in what follows.ASSUMPTION 3.1 Let J\subseteq \mathbb {R} be an open set . The operator-valued function A(\lambda ):\mathcal {X}\hookrightarrow \mathcal {X}^\ast , is weakly continuous on J.In particular, it is also locally bounded on J (in the uniform operator topology).The clas...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.049693942070007324, 0.023519152775406837, -0.03940717130899429, 0.017521081492304802, 0.02625109814107418, -0.02712104842066765, 0.03345488756895065, 0.02063458412885666, 0.02177925407886505, 0.012217444367706776, -0.020466700196266174, -0.006677241530269384, 0.02466382272541523, 0.0016...
5fbc4f2fbe4b05b9ac5d76ab4fdf6738e856a1ad
subsection
5
27
Global smoothing estimates
Otherwise, the set J should be replaced by J\setminus \mathcal {N}.To prove  (REF ) we use the definition of A(\lambda ) and the spectral calculus to get(\sigma (H)e^{ita(H)}P_{ac}(H)E(J)\phi ,\psi )_{\mathcal {H}}=\int \limits _{J}e^{ita(\lambda )}\sigma (\lambda ){\left\langle {A(\lambda )\phi ,\psi }\right\rangle }d...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.031034396961331367, 0.04034166410565376, -0.040494244545698166, -0.01480771042406559, -0.012084189802408218, -0.03613050654530525, -0.01595967635512352, 0.010154076851904392, 0.02929500676691532, 0.013747292570769787, -0.024839723482728004, 0.024916013702750206, -0.01864505186676979, 0....
4de8e84866253ee5102fd862bd7a9a52a74b376b
subsection
6
27
Global smoothing estimates
The method of proof used for Proposition  REF in conjunction with the role of A(\lambda ) in the spectral calculus enable us to obtain a general spacetime estimate for any initial data \phi \in \mathcal {H}.THEOREM 3.4 Assume the conditions of Assumptions  REF - REF and let \phi \in \mathcal {H}. Then\Big \Vert \sigma...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.02635962702333927, 0.024269772693514824, -0.024269772693514824, -0.02288162149488926, 0.0015864589950069785, -0.0514989010989666, 0.02567317895591259, -0.00790178682655096, 0.045824259519577026, 0.020227352157235146, -0.008672134019434452, 0.02094431035220623, -0.009823842905461788, 0.0...
9bbca5f6f9ec2c2bf156bf29c3f2a9226fa73c79
subsection
7
27
Global smoothing estimates
As in  (REF ) we write, assuming first that \phi \in \mathcal {X},\aligned \Big \Vert \sigma (H)e^{ita(H)}P_{ac}(H)E(J)\phi \Big \Vert _{ L^2(\mathbb {R}_t,\mathcal {X}^\ast )}\hspace{100.0pt}\\ =\sup \limits _{\Vert g\Vert _{ L^2(\mathbb {R}_t,\mathcal {X})}=1} \sqrt{2\pi }\,\Big |\int _{J}\sigma (\lambda ) {\left\lan...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.04306621104478836, 0.030949071049690247, -0.006211179308593273, -0.02383749932050705, -0.012338424101471901, -0.021777279675006866, 0.014963296242058277, 0.010644465684890747, 0.013582185842096806, 0.01395607739686966, -0.017046406865119934, 0.019869670271873474, -0.005642711650580168, ...
e7f09c7bfeb4da654b5120aaa05c7056b8c3c385
subsection
8
27
Global smoothing estimates
However, this assumption can be dropped due to the density of \mathcal {X} in \mathcal {H}.Remark 3.5 Note that in the above statements we could “absorb” E(J) into \sigma (H), but we preferred to emphasize the localization aspect of the estimates (in “energy” space) and to leave \sigma (\lambda ) as a continuous functi...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.029395833611488342, 0.030357379466295242, -0.014736074022948742, -0.008020514622330666, -0.00937888864427805, -0.03351674601435661, 0.011057778261601925, 0.013293754309415817, 0.025885427370667458, -0.008150246925652027, 0.01720098964869976, 0.022237658500671387, -0.007623685989528894, ...
a9aa6faa943edec45edb15b6f5e398ce392adabe
subsection
9
27
Global smoothing estimates
Then for all \phi \in L^2(\mathbb {R}^n_x), \aligned \Big \Vert {\left\langle {x}\right\rangle }^{-s}\sigma (H)e^{ita(H)}P_{ac}(H)E(J)\phi \Big \Vert _{L^2(\mathbb {R}_t\times \mathbb {R}^n_x)}\hspace{100.0pt}\\ \le \sqrt{ 2\pi }\,\,\sup \limits _{\lambda \in J}\Big [\frac{|\sigma (\lambda )|}{|a^{\prime }(\lambda )|^...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.04972744360566139, 0.03983690217137337, -0.016011076048016548, -0.003132767742499709, -0.012355544604361057, -0.023688456043601036, -0.014454231597483158, -0.017155814915895462, 0.017735816538333893, 0.021597400307655334, -0.026954777538776398, 0.014072651974856853, 0.015469233505427837, ...
b0e5d535d3713450a0daea73af2f3f8a39b252d0
subsection
10
27
Global smoothing estimates
Then  (REF ) can be rewritten as\Vert \mathfrak {H}_J\Vert \le \sqrt{2\pi }\,\,\sup \limits _{\lambda \in {J\setminus \mathcal {N}}}{\left[{\frac{|\sigma (\lambda )|}{|a^{\prime }(\lambda )|^\frac{1}{2}} \Vert A(\lambda )\Vert _{B(\mathcal {X},\mathcal {X}^\ast )}^\frac{1}{2}}\right]}.In fact, we now show that the oper...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.07860063761472702, 0.019192470237612724, -0.01896362565457821, -0.030848311260342598, -0.003928506281226873, -0.015386014245450497, 0.002938751596957445, 0.0002860562817659229, 0.028788704425096512, 0.03521162271499634, -0.042839791625738144, 0.016171716153621674, 0.031138181686401367, ...
4fbbad4de826c0011fa488209af2a12165513606
subsection
11
27
Global smoothing estimates
Using Equation  (REF ) we therefore write, for \phi \in \Gamma ,\aligned \Big \Vert \sigma (H)e^{ita(H)}P_{ac}(H)E(J)\phi \Big \Vert _{ L^2(\mathbb {R}_t,\mathcal {X}^\ast )}\hspace{100.0pt}\\ =\sup \limits _{\Vert g\Vert _{ L^2(\mathbb {R}_t,\mathcal {X})}=1}\Big |\int _\mathbb {R}\int _{J_\lambda }\sigma (\lambda ) e...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.03700758144259453, 0.031546447426080704, -0.0057776677422225475, -0.036641474813222885, -0.014110475778579712, -0.033834632486104965, 0.002490308368578553, -0.0052094352431595325, 0.007600588724017143, 0.016871552914381027, -0.023919163271784782, 0.018885156139731407, -0.02912097051739692...
4cd89603407cb8f67e501921019d0be35e0b6181
subsection
12
27
Global smoothing estimates
In view of  (REF ) there exists \psi \in \mathcal {X} such that \Vert \psi \Vert _{\mathcal {X}}=1 and{\left\langle {A(\lambda _0)\psi ,\psi }\right\rangle }\,\,\,\ge \Vert A(\lambda _0)\Vert _{B(\mathcal {X},\mathcal {X}^\ast )}-\delta .Let h>0 so that D_h=(\lambda _0-\frac{h}{2},\lambda _0+\frac{h}{2})\subseteq J, an...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.01784089393913746, 0.028264615684747696, -0.023090910166502, -0.010858679190278053, -0.017673015594482422, -0.008699152618646622, 0.011629394255578518, 0.005192783661186695, 0.05976470559835434, 0.04001610353589058, -0.022923031821846962, 0.022129423916339874, 0.03497975319623947, 0.043...
490f22f7a193c9fb6d2a0ce112f36fb2fb5deacd
subsection
13
27
Global smoothing estimates
In the case of the Laplacian H=-\Delta such a result was obtained in  .
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1016/0022-1236(92)90100-w", "end": 72, "openalex_id": "https://openalex.org/W2036342747", "raw": "B. Simon, Best constants in some operator smoothness estimates, J. Funct. Anal.107 (1992), 66–71.", "source_ref_id": "6509b98ba4c47d0...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.023112259805202484, 0.038383133709430695, -0.03585069999098778, 0.00294052017852664, -0.05147245153784752, 0.00862705148756504, -0.0009611038840375841, 0.06437870860099792, -0.012563001364469528, -0.005812390241771936, -0.007723156362771988, 0.0032990272156894207, 0.015331895090639591, ...
58c024d119e76397cd86b64339a360d611eeb375
subsection
14
27
Comparison principles
As an application of the spectral identities established in Section  , we present here applications of spectral comparison principles for self-adjoint operators. Given two such operators, these principles enable us to carry smoothing estimates for one of them to the other, provided their spectral densities can be “comp...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1112/plms/pds006", "end": 564, "openalex_id": "https://openalex.org/W1972497712", "raw": "M. Ruzhansky and M. Sugimoto, Smoothing properties of evolution equations via canonical transforms and comparison principle, Proc. Lond. Math.Soc. ...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.05104311183094978, 0.04127993434667587, -0.043629199266433716, 0.014850707724690437, 0.0236757043749094, -0.04124942421913147, 0.04957863688468933, -0.005156178027391434, 0.03514743968844414, 0.03542202711105347, -0.023797744885087013, -0.0030719684436917305, 0.013851507566869259, 0.002...
2c08f8256b19b6729209a715b810aea76da8dea6
subsection
15
27
Comparison principles
Suppose that we have\frac{|\sigma (\lambda )|}{{\left|{{a}^{\prime }(\lambda )}\right|}^{1/2}} {\left|{ {\left\langle { A(\lambda )\phi ,\,\psi }\right\rangle }}\right|} \ge \frac{|\widetilde{\sigma }(\lambda )|}{{\left|{\widetilde{a}^{\prime }(\lambda )}\right|}^{1/2}} {\left|{ {\left\langle { \widetilde{A}(\lambda )\...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.006507397163659334, 0.06261748820543289, -0.005946677643805742, -0.004069983959197998, 0.0020121054258197546, -0.032865025103092194, -0.0018204309744760394, 0.011702634394168854, 0.07793618738651276, 0.046841055154800415, -0.003284214064478874, 0.04140932112932205, -0.011946757324039936, ...
800ebf943e2b67c1480f10b88dcb2c68646026b1
subsection
16
27
Comparison principles
The absence of such a scalar product in the second estimate means that it is “space-global comparison” .The following theorem is a direct consequence of Theorem  REF (and  (REF )).THEOREM 4.4 (uniform comparison) Assume the conditions of Assumption  REF .Suppose that for almost all \lambda \in J and for every unit vec...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.04860600084066391, 0.02381480485200882, -0.03405166044831276, 0.005270989611744881, 0.017514027655124664, -0.022472266107797623, -0.011159852147102356, 0.030802108347415924, 0.039391305297613144, 0.012746489606797695, -0.0018145248759537935, 0.001817385433241725, -0.0021949743386358023, ...
f40353a2c665f59086b53588d34fcce10715475c
subsection
17
27
Comparison principles
Suppose also that there is a constant C_0>0 such that{\left\Vert {\sigma (H)e^{itH}P_{ac}(H)E(J)\phi }\right\Vert }_{ L^2(\mathbb {R}_t,\mathcal {X}^\ast )} \le C_0 \Vert \phi \Vert _{\mathcal {H}}for all \phi \in \mathcal {H}.Then we have the estimate{\left\Vert {|a^{\prime }(H)|^{1/2}\sigma (H)e^{ita(H)}P_{ac}(H)E(J)...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.01914733648300171, 0.03740977495908737, 0.006175206508487463, -0.019864406436681747, -0.0038485382683575153, -0.058860894292593, -0.000002860657787095988, 0.030849331989884377, 0.04244453087449074, 0.028026817366480827, -0.026470620185136795, -0.0014017223147675395, -0.012594521977007389,...
d3b3e90b0b01405e793f1e94c791434cc28aa721
subsection
18
27
Comparing unperturbed and perturbed operators
Comparing unperturbed and perturbed operatorsThe comparison principles presented above provide a very effective way of dealing with global spacetime estimates for perturbations, as we shall now see.Turning back to the setup in Section   we impose on the self-adjoint operator H an assumption which is stronger than Assum...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/978-3-0348-0024-2_3", "end": 731, "openalex_id": "https://openalex.org/W204789615", "raw": "M. Ben-Artzi, Smooth spectral calculus , In: ``Partial Differential Equations and Spectral Theory”, M. Demuth, B.-W. Schulze and I. Witt (Ed...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.0329008623957634, 0.022768128663301468, -0.03958480432629585, 0.023073330521583557, 0.013627304695546627, -0.01754916086792946, 0.017304997891187668, -0.00637110834941268, 0.02846015989780426, 0.01776280254125595, -0.06149836257100105, -0.048038918524980545, 0.033206064254045486, -0.008...
ad0fe7fae7a9989b5832e56d5efc97faaa9e3a0d
subsection
19
27
Comparing unperturbed and perturbed operators
\end{array}\right.}Remark 4.8 The compactness hypothesis on V can be replaced by the assumption that the operators VR^\pm (\lambda ):\mathcal {X}\hookrightarrow \mathcal {X} are compact for any \lambda \in J.Under these assumptions, the operator \widetilde{H}=H+V is essentially self-adjoint on \mathcal {D}   and we us...
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.04760725051164627, 0.03869615122675896, -0.017196593806147575, 0.049926578998565674, 0.022567668929696083, -0.036071646958589554, 0.025207430124282837, 0.030639538541436195, 0.04113755002617836, 0.005870798137038946, -0.011177022941410542, -0.02679433859884739, 0.002737035509198904, -0....
c391e95e59adadb60083d9fdb6e96b0be6fdd60d
subsection
20
27
Comparing unperturbed and perturbed operators
Under the conditions of the Proposition, we can invoke the resolvent equation to obtain the Limiting Absorption Principle for \widetilde{H} :{\widetilde{R}}^\pm (\lambda )=\lim \limits _{\epsilon \downarrow 0}{\widetilde{R}}(\lambda \pm i\epsilon ))=R^\pm (\lambda )[I+VR^\pm (\lambda )]^{-1},\quad \lambda \in J.The exi...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/978-3-0348-0024-2_3", "end": 901, "openalex_id": "https://openalex.org/W204789615", "raw": "M. Ben-Artzi, Smooth spectral calculus , In: ``Partial Differential Equations and Spectral Theory”, M. Demuth, B.-W. Schulze and I. Witt (Ed...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.015272055752575397, 0.026851966977119446, -0.006705363281071186, 0.02380060777068138, -0.00008027698640944436, -0.049340490251779556, 0.000881556945387274, -0.00865060556679964, 0.023724323138594627, 0.0008334026788361371, -0.04216979444026947, -0.016065409407019615, 0.0120681282132864, ...
4fd3a55921c7f00084756b9ef63f61bb3a8bca68
subsection
21
27
Applications to operators of mathematical physics
We shall now give a few examples that illustrate the scope of the abstract results when applied to a variety of operators that are frequently studied in mathematical physics.
{ "cite_spans": [] }
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.03155525401234627, 0.06045547500252724, -0.052978649735450745, 0.012466462329030037, -0.012512238696217537, -0.023468077182769775, 0.04217539727687836, 0.01937871053814888, 0.033691491931676865, 0.0032405934762209654, -0.042724717408418655, -0.007747669704258442, -0.0000728370578144677, ...
a69db099354fd6c689eb4535ef8d60f6e9022ad0
subsection
22
27
The fractional Laplacian
Consider the operator H=-\Delta in \mathcal {H}=L^2(\mathbb {R}^n), n\ge 3. It is absolutely continuous in J=(0,\infty ) and the condition of Assumption  REF is satisfied with \mathcal {X}=L^2_s(\mathbb {R}^n),\,s>\frac{1}{2}  .Furthermore, it was shown in   that for all \phi \in L^2(\mathbb {R}^n),\int _{\mathbb {R}}\...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 230, "openalex_id": "", "raw": "M. Ben-Artzi and A.Devinatz, Local smoothing and convergence properties for Schrödinger-type equations, J. Funct. Anal. 101 (1991), 231–254.", "source_ref_id": "1bd11b8bc32b88f8c2bb4f120dffb7e...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.060029707849025726, 0.01551098097115755, 0.0010748201748356223, 0.03030477836728096, -0.010528850369155407, 0.0031243220437318087, -0.0005083222058601677, 0.03277676925063133, 0.0072137885726988316, -0.005085129756480455, -0.05429224669933319, -0.006847567856311798, 0.03723245859146118, ...
ae92bc2c6393697d61554e425b838295161d7788
subsection
23
27
The Stark Hamiltonian
Consider the self-adjoint operator in \mathcal {H}=L^2(\mathbb {R}^n), n\ge 1,H=-\Delta -x_1,where x=(x_1,x_2,\ldots ,x_n)\in \mathbb {R}^n.This operator is the well-known “Stark Hamiltonian”, describing the motion of a quantum-mechanical charged particle in a uniform electric field (all physical constants scaled to un...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1006/jfan.1997.3211", "end": 464, "openalex_id": "https://openalex.org/W2076069315", "raw": "M. Ben-Artzi and A. Devinatz, Regularity and decay of solutions to the Stark evolution equation, J. Funct. Anal. 154 (1998), 501–512.", "s...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.050852421671152115, 0.015147367492318153, -0.010927471332252026, -0.005444657988846302, 0.018894147127866745, 0.0025868804659694433, -0.002922640647739172, 0.010271212086081505, 0.02940954826772213, -0.001553845126181841, -0.027272891253232956, -0.0016406468348577619, 0.018741527572274208...
24e088cda548f9a0f86e6bf611dcfdb030cbe47f
subsection
24
27
The Schrödinger operator with potential
Consider the operator \widetilde{H}=H+V in L^2(\mathbb {R}^n),\,n\ge 3, where H=-\Delta . This operator can be studied in terms of Theorem  REF .Employing the notation of Subsection  REF we let \mathcal {X}=L^2_{s}(\mathbb {R}^n), so that \mathcal {X}^\ast =L^2_{-s}(\mathbb {R}^n) and \mathcal {X}^\ast _H=H^2_{-s}(\mat...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 500, "openalex_id": "", "raw": "M. Ben-Artzi and A.Devinatz, Local smoothing and convergence properties for Schrödinger-type equations, J. Funct. Anal. 101 (1991), 231–254.", "source_ref_id": "1bd11b8bc32b88f8c2bb4f120dffb7e...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.05989040061831474, 0.012577594257891178, -0.01236402615904808, 0.018382050096988678, 0.01739048771560192, 0.02297374978661537, 0.014453936368227005, 0.021021133288741112, 0.0558021105825901, -0.005262910854071379, -0.04314061254262924, -0.006673980969935656, 0.029518064111471176, 0.0135...
053463948bbdee7acc0e7701b435fc176c95a2a0
subsection
25
27
The Schrödinger operator with potential
Then there exists a constant C_{\delta ,s} so that\int _{\mathbb {R}}\int _{\mathbb {R}^n}{\left\langle {x}\right\rangle }^{-2s}|(I+\widetilde{H})^\frac{1}{4} {\widetilde{E}}(J_\delta )e^{it{\widetilde{H}}}\phi (x)|^2\,dx\,dt\le C_{\delta ,s}\Vert \phi \Vert _{L^2(\mathbb {R}^n)}^2,for all \phi \in L^2(\mathbb {R}^n).T...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 475, "openalex_id": "", "raw": "M. Ben-Artzi and A.Devinatz, Local smoothing and convergence properties for Schrödinger-type equations, J. Funct. Anal. 101 (1991), 231–254.", "source_ref_id": "1bd11b8bc32b88f8c2bb4f120dffb7e...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.014397467486560345, 0.022981012240052223, -0.021882319822907448, 0.03744714707136154, 0.023438801988959312, 0.0023766884114593267, 0.028489742428064346, 0.02128719352185726, 0.03842376545071602, -0.00244154199026525, -0.020249538123607635, -0.008263093419373035, 0.028855973854660988, -0...
b8504f25b3fd204726e73dd5a63a16c2ce1875b8
subsection
26
27
The Schrödinger operator with potential
For example, we can take V to be the pseudodifferential operator V={\left\langle {x}\right\rangle }^{-s}(I-\Delta )^\beta {\left\langle {x}\right\rangle }^{-s}, for any \beta <\frac{1}{2}.We refer the reader to  , , for related spacetime estimates for the Schrödinger operator, using very different methods.
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1215/s0012-7094-79-04631-3", "end": 309, "openalex_id": "https://openalex.org/W1545201582", "raw": "A. Jensen and T. Kato, Spectral properties of Schrödinger operators and time-decay of the wave functions, Duke Math. J. 46 (1979), 583–61...
1807.11611
Spectral identities and smoothing estimates for evolution operators
[ "Matania Ben-Artzi", "Michael Ruzhansky", "Mitsuru Sugimoto" ]
[ "math.SP", "math.AP", "math.FA" ]
2,018
en
Mathematics
[ -0.06854211539030075, 0.0009860940044745803, -0.04968234896659851, 0.05789153650403023, -0.014694134704768658, 0.0021152382250875235, 0.030334310606122017, 0.029891807585954666, 0.07342489808797836, -0.005947080906480551, -0.02626023441553116, 0.0007052383734844625, -0.004848454147577286, ...
def140cad4d780ad9cc072267636353ae272fa3c
abstract
0
31
Abstract
Blind image deblurring plays a very important role in many vision and multimedia applications. Most existing works tend to introduce complex priors to estimate the sharp image structures for blur kernel estimation. However, it has been verified that directly optimizing these models is challenging and easy to fall into ...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.0351165309548378, -0.01372931245714426, -0.00045764376409351826, 0.00760451378300786, 0.0035619938280433416, -0.03067738749086857, -0.004229390993714333, -0.004790004808455706, 0.038136981427669525, 0.016505684703588486, -0.041737109422683716, -0.025002604350447655, -0.02210419438779354, ...
8b0ffdaadeb95dc7b542afcfe75639f5ac97756c
subsection
1
31
Introduction
Blind image deblurring is a fundamental component in many multimedia and computer vision applications. This problem involves the estimation of latent sharp image and blur kernel from a blurry observation. The most commonly used formulation for the blurry phenomenon can be given as follows:\mathbf {y}=\mathbf {u}\otimes...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2011.5995308", "end": 1108, "openalex_id": "https://openalex.org/W2036682493", "raw": "Anat Levin, Yair Weiss, Fredo Durand, and William T. Freeman. 2011. Efficient marginal likelihood optimization in blind deconvolution. In CV...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.02836034819483757, -0.012707998976111412, -0.01394371036440134, -0.024912560358643532, 0.016933828592300415, -0.0017544056754559278, 0.004752150736749172, -0.024836281314492226, 0.014347987249493599, 0.028924809768795967, -0.03517964482307434, -0.028528161346912384, -0.008924584835767746,...
9889328695609f7bfad084646f0f604ec9f1563d
subsection
2
31
Introduction
Fig. REF shows the performance of GCM on various applications.Our contributions can be summarized as follows:We establish two fundamental propagative modules (i.e., Generator and Corrector) to respectively learn latent image structures from training data and investigate principled mathematical rules for image propagati...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.06743306666612625, -0.03591344878077507, 0.0076052905060350895, -0.021252095699310303, 0.007754039950668812, -0.03573037311434746, -0.024959389120340347, -0.014699493534862995, 0.004382386803627014, 0.04033778980374336, -0.0454639233648777, -0.030375393107533455, -0.021999657154083252, ...
89905c92e870968bdfb39c1910a6c5d6fb795f1d
subsection
3
31
Related Works
In this section, we briefly review some related works on the prior models and the existing inference strategies for blind image deblurring. Specifically, the most commonly used deblurring formulation is the following regularized variational minimization model\min \limits _{\mathbf {u}}\Psi (\mathbf {u})=f(\mathbf {u}) ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1145/1141911.1141956", "end": 1226, "openalex_id": "https://openalex.org/W2098535678", "raw": "Rob Fergus, Barun Singh, Aaron Hertzmann, Sam T. Roweis, and William T. Freeman. 2006. Removing camera shake from a single photograph. ACM TOG...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.040762558579444885, 0.0012128081871196628, -0.017528509721159935, 0.008298964239656925, 0.016872525215148926, 0.0015932408859953284, 0.017864130437374115, -0.00134057248942554, 0.04601043090224266, 0.024500250816345215, -0.0575435496866703, 0.0006926927017048001, 0.007009878288954496, 0...
c4919d078941e5a405660d45711e82a790fb6e47
subsection
4
31
Related Works
(REF ) can also be utilized to formulate other computer vision and multimedia tasks. For example, by defining f with physical rules of different problems and enforcing other task-related priors for \phi , a variety of applications, such as image interpolation and edge-preserved smoothing, can all be formulated by Eq. (...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.04717462137341499, 0.015043394640088081, 0.006255366839468479, 0.00006257750646909699, -0.009596647694706917, 0.012075909413397312, 0.007376755587756634, 0.0277524683624506, 0.03399257734417915, 0.002654716605320573, -0.010496811009943485, -0.0042567006312310696, -0.010969777591526508, ...
c1f49a001ec5c2bbbf675a587a312dc5ba8063ac
subsection
5
31
The Collaborative Modules
In this section, we propose a collaborative framework to learn Generation and Correction Modules (GCM) for latent image propagation. The strict theoretical analysis on GCM is also established at the end of this section.
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.04636380448937416, 0.003975566476583481, 0.009194928221404552, -0.005883228033781052, -0.0013067481340840459, -0.022479882463812828, -0.01874086633324623, 0.012376907281577587, -0.005402497015893459, 0.02733297273516655, -0.015238399617373943, -0.041144441813230515, -0.012720286846160889,...
86f6bdf40595bf9b317eda022482c7bd5005f649
subsection
6
31
Generator with Fidelity Warm Start
Inspired by the success of deep networks in visual processing areas, we would like to first establish our Generator as a parameterized network architecture (denoted as \mathcal {N}), i.e., at t-th stage, we consider\tilde{\mathbf {u}}^{t+1}=\mathcal {N}^t(\mathbf {u}_0^{t+1};{\omega }^t),where {\omega }^t is the learna...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2014.371", "end": 1484, "openalex_id": "https://openalex.org/W2047123483", "raw": "Jinshan Pan, Zhe Hu, Zhixun Su, and Ming-Hsuan Yang. 2014. Deblurring text images via L0-regularized intensity and gradient prior. In CVPR. 2901...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.008025805465877056, -0.01132157351821661, -0.03372059017419815, 0.03799287974834442, 0.023467088118195534, -0.010787537321448326, 0.012382017448544502, 0.02873116359114647, 0.017501138150691986, 0.040708839893341064, 0.018019916489720345, 0.005759964697062969, -0.025526944547891617, 0.0...
5764e77da4700913bdae5b922bdd3c3970d4df49
subsection
7
31
Corrector by Proximal Prior Descent
Since generating the latent image structure is a highly ill-posed problem, only performing Generator may not guarantee the exact recovery of our desired optimal solution. Moreover, no prior knowledge is enforced into the current scheme, thus it is natural to introduce another module to incorporate our prior assumptions...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 915, "openalex_id": "https://openalex.org/W2147298660", "raw": "Dilip Krishnan and Rob Fergus. 2009. Fast image deconvolution using hyper-Laplacian priors. In NIPS. 1033–1041.", "source_ref_id": "693c5a60265a97fa3c6b22c5f875...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.0676441565155983, -0.006360717583447695, -0.005017602816224098, -0.00017122807912528515, 0.010821232572197914, -0.02735070511698723, -0.008173159323632717, 0.021886669099330902, -0.006585841998457909, 0.033333670347929, -0.02846487984061241, -0.015163462609052658, -0.03724091500043869, ...
67400d2ae4b393436152ca1a809aca5282aaea38
subsection
8
31
GCM with Theoretical Guarantee
We first summarize the complete GCM framework in Alg. REF and express the pipeline of GCM in Fig. REF . Please notice that due to the CNN-based Generator, GCM is indeed not a standard optimization scheme. Thus existing convergence analysis is not available for the propagations generated by our GCM. But fortunately, we ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1291, "openalex_id": "https://openalex.org/W2184634347", "raw": "Huan Li and Zhouchen Lin. 2015. Accelerated proximal gradient methods for nonconvex programming. In NIPS. 379–387.", "source_ref_id": "b3547030027dc2eff604ae84...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.058905862271785736, 0.009797296486794949, -0.022280456498265266, -0.024218549951910973, 0.04044055566191673, 0.004322560969740152, -0.031131597235798836, -0.012200838886201382, -0.011071555316448212, 0.06769596040248871, -0.015916790813207626, -0.017976971343159676, -0.019701417535543442,...
125b24cfd47aac04bb3476b0294556be03443c2b
subsection
9
31
Applications
Now we demonstrate how to apply our GCM to address blind image deblurring. Thanks to the flexibility of our Generation and Correction modules, GCM indeed can also be applied to address other related multimedia applications.
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.03893464058637619, -0.007941018790006638, 0.015897294506430626, -0.005107120610773563, -0.017361920326948166, -0.061849936842918396, -0.016431273892521858, -0.003182891756296158, 0.03524256497621536, 0.028865337371826172, -0.03832438215613365, -0.0276905857026577, -0.013524905778467655, ...
316d2a561ebf3c50d87283a01a66038b5e15f470
subsection
10
31
Blind Image Deblurring
As discussed in Sec. , generating latent image with rich salient edges and sharp structures often plays important role in blind image deblurring. Therefore, we would like to train our Generator in image gradient domain as follows. We first calculate the warm start with the fidelity f(\mathbf {u}) = \Vert \mathbf {u}\ot...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/978-3-642-15549-9_12", "end": 1302, "openalex_id": "https://openalex.org/W1598281290", "raw": "Li Xu and Jiaya Jia. 2010. Two-phase kernel estimation for robust motion deblurring. In ECCV. 157–170.", "source_ref_id": "58808c9b...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.03066183254122734, -0.00541539816185832, 0.002743928926065564, -0.0077302902936935425, 0.019358141347765923, -0.02690918929874897, 0.008054451085627079, -0.008123097009956837, -0.008245134726166725, 0.006422204431146383, -0.00503784604370594, -0.015445325523614883, -0.04448254033923149, ...
ee860baadc1c5907a651a601718034ecbcab3e25
subsection
11
31
Byproduct Applications
As nontrivial byproducts, we would also like to demonstrate how to apply GCM to other image related vision and multimedia applications.Image Interpolation: The purpose of this task is to remove corrections (e.g., text, blocking or mask) from the partially invisible observation. When using GCM to deal with this problem,...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1145/2366145.2366158", "end": 1070, "openalex_id": "https://openalex.org/W2078718577", "raw": "Li Xu, Qiong Yan, Yang Xia, and Jiaya Jia. 2012b. Structure extraction from texture via relative total variation. ACM TOG 31, 6 (2012), 139.",...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.0705128088593483, -0.022588519379496574, 0.006257630418986082, -0.017750302329659462, 0.01404151227325201, -0.0595848523080349, -0.011027165688574314, 0.0026404147502034903, 0.006883393507450819, 0.026159947738051414, -0.03702685609459877, -0.029639801010489464, -0.006059217732399702, 0...
5fc4f0e8c5e27daed03266dd76b00a93997663b2
subsection
12
31
Discussions
Here we would like to discuss and highlight some important aspects of GCM.
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.023544929921627045, 0.02122553251683712, -0.022476784884929657, -0.01652570255100727, -0.0027447466272860765, -0.03332607075572014, 0.01573222503066063, -0.015137115493416786, -0.020615164190530777, 0.05200331658124924, 0.0019083523657172918, -0.0092470683157444, -0.03903300687670708, 0...
1676df9bc18ea7af7d57fcb7b7305532082f454b
subsection
13
31
Theoretically Convergent Ensemble Framework for Image Modeling
Indeed, our GCM can be viewed as a general framework to integrate deep network architectures (i.e., Generator) and physical principles (i.e., Corrector) to address not only blind image deblurring, but also other related vision and multimedia tasks. The main advantage against existing heuristic ensemble strategies (e.g....
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2017.300", "end": 532, "openalex_id": "https://openalex.org/W2613155248", "raw": "Kai Zhang, Wangmeng Zuo, Shuhang Gu, and Lei Zhang. 2017. Learning Deep CNN Denoiser Prior for Image Restoration. In CVPR.", "source_ref_id...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.010799885727465153, -0.0038630946073681116, 0.04036228358745575, 0.003474115626886487, 0.010685480199754238, -0.03725045174360275, -0.04530460387468338, 0.009236343204975128, 0.012607493437826633, 0.042436838150024414, -0.05872819200158119, -0.025901420041918755, -0.030355609953403473, ...
8aa89dd8ed8324ab8e37b474ee0eb7d26c5b2cd0
subsection
14
31
Analogy to Adversarial Learning Methodology
Since there exist two cascaded modules in our GCM (i.e., Generator and Corrector), which is similar to that in the popular adversarial learning methods (i.e., Generator and Discriminator in Generative Adversarial Network, GAN for short), we would like to provide some brief comparisons between these two learning methodo...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.04744218289852142, -0.04603874683380127, -0.006304013542830944, -0.03703846409916878, 0.015559815801680088, -0.07999575883150101, -0.02350752428174019, 0.02288208156824112, 0.004915833938866854, 0.04823542758822441, -0.030021291226148605, -0.04670995473861694, -0.04695403203368187, 0.01...
8945bf454a075258db9c57237cfbc7ebbfd043e1
subsection
15
31
Experimental Results
We first conduct experiments to verify the mechanism of GCM. Then a range of results are demonstrated to evaluate GCM on blind image deblurring. Finally, we show the performance of GCM on other related applications, such as image interpolation and edge-preserved smoothing.
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.03540978953242302, 0.011370816268026829, 0.01851382479071617, -0.0355624184012413, -0.0028808610513806343, -0.02063535898923874, 0.00397215411067009, -0.005937244743108749, 0.012355268932878971, 0.025611044839024544, -0.028343094512820244, -0.022833209484815598, -0.006143293343484402, 0...
bbea91b939d0e427c2f1dbd8dfa7430d0ad98641
subsection
16
31
Experimental Setup
To provide fair comparisons on the blind image deblurring problem, we adopt   as the final non-blind deconvolution process for all the compared methods. We also execute these approaches with their default parameter settings. For the training data of GCM, we randomly select 800 natural images from the training set of th...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/iccv.2011.6126278", "end": 152, "openalex_id": "https://openalex.org/W2172275395", "raw": "Daniel Zoran and Yair Weiss. 2011. From learning models of natural image patches to whole image restoration. In ICCV. 479–486.", "sourc...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.012198961339890957, -0.03280521184206009, -0.0026740061584860086, -0.029082201421260834, -0.01945425383746624, -0.029753563925623894, -0.009803418070077896, -0.018386175855994225, 0.03439207002520561, 0.017180776223540306, -0.021315759047865868, -0.023970691487193108, -0.01942373625934124...
b57e56a0861b3a6e2806484ed0e49e47c1f95249
subsection
17
31
Model Verification
To verify the efficiency of our proposed collaborative learning strategies, we first compare the deblurring performance of GCM with different settings on an example image. In Fig. REF , we denote the naive cascade of the designed Generator and Corrector as “Generator" and “Corrector”, respectively. While “GCM” denotes ...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.05473468452692032, -0.019602583721280098, 0.002059415215626359, -0.045520707964897156, 0.0215399581938982, -0.022455254569649696, -0.025963887572288513, -0.007932562381029129, 0.0003224987303838134, 0.05961626023054123, -0.03003695420920849, -0.06254520267248154, 0.009930958040058613, 0...
a2b4b6690a4177d5dae0ceffeb3ca12a3125befe
subsection
18
31
Synthetic Dataset
We first consider synthetic test data and compare GCM with several state-of-the-art blind deblurring methods (e.g., , , , , ) on widely used Levin et al.'s dataset , including 32 blurry images with size 255\times 255, which are produced by 4 clear images and 8 blur kernels. Quantitative scores (e.g., PSNR, SSIM, Error ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2011.5995521", "end": 274, "openalex_id": "https://openalex.org/W1987075379", "raw": "Dilip Krishnan, Terence Tay, and Rob Fergus. 2011. Blind deconvolution using a normalized sparsity measure. In CVPR. 233–240.", "source...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.046726930886507034, -0.019685741513967514, 0.0043911412358284, -0.06970889121294022, 0.00669544143602252, -0.006264338735491037, -0.03201603516936302, -0.0081184608861804, 0.0011540575651451945, 0.03686879947781563, -0.003673908766359091, 0.00032761882175691426, -0.03686879947781563, 0....
0e9bd7f88ebbd8e007f0aa4bb76918f339b368f8
subsection
19
31
Real Blurry Images
We also evaluate the compared methods on real-world blurry images (collected by ) in Fig. REF . We can see on the top row that GCM can recover more details (e.g., the tail of bird with legible), compared with other methods. On the bottom row, it is also easy to observe that the numbers in license plate have been succes...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2016.188", "end": 95, "openalex_id": "https://openalex.org/W2465552163", "raw": "Wensheng Lai, Jiabin Huang, Zhe Hu, Narendra Ahuja, and Ming-Hsuan Yang. 2016. A comparative study for single image blind deblurring. In CVPR. 170...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.056610800325870514, -0.036652058362960815, 0.01994348131120205, -0.03500409051775932, -0.006332474760711193, -0.0296634491533041, 0.02412443980574608, -0.015320011414587498, -0.00026631681248545647, 0.05535956472158432, -0.060578133910894394, -0.012771762907505035, -0.003400844056159258, ...
0ff73fd24b846f4a4af7846493b0f3742721f275
subsection
20
31
Detection by Deblurring
It is known that motion blurs caused by the shaking of capture device often reduce the performance of detection algorithm. See Fig. REF (b) for an example. Thus a natural strategy to evaluate the effectiveness of deblurring algorithms is just to perform object detection on the restored images. In this experiment, we ad...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/cvpr.2016.91", "end": 418, "openalex_id": "https://openalex.org/W2963037989", "raw": "Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. 2016. You only look once: Unified, real-time object detection. In CVPR. 779–788.",...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.05285513401031494, -0.06066744029521942, -0.011619280092418194, 0.00460804020985961, -0.01573142223060131, 0.005206932779401541, -0.021010830998420715, 0.00613388093188405, -0.026595409959554672, 0.022857099771499634, -0.04070943966507912, -0.01554069109261036, 0.008201396092772484, 0.0...
9ff9b723446d10691f3d52b6596b87daa11ab838
subsection
21
31
Other Applications
To verify the flexibility of our collaborative modules, we also express the performance of GCM on other applications, including image interpolation and edge-preserved smoothing.
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.05809216573834419, 0.007440770510584116, 0.0006912555545568466, -0.02982410229742527, 0.021845582872629166, -0.044362399727106094, -0.007250079419463873, 0.007013622205704451, 0.0006192696746438742, 0.02274564653635025, -0.010297324508428574, -0.020670926198363304, -0.004366828128695488, ...
613b2c2737724e5859118351f12dad73920b2b9b
subsection
22
31
Image Interpolation
The purpose of image interpolation is to recover an image in which some pixels are lost or deteriorated. To evaluate the performance of our method in this task, we compare GCM with other state-of-the-art image interpolation methods , , , on both text and random missing pixels masks. The test images are randomly chosen ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1109/tip.2015.2446943", "end": 284, "openalex_id": "https://openalex.org/W1562968274", "raw": "Kyong Hwan Jin and Jong Chul Ye. 2015. Annihilating filter-based low-rank Hankel matrix approach for image inpainting. IEEE TIP 24, 11 (2015),...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.06110222265124321, -0.028384149074554443, -0.00994971301406622, -0.02842993102967739, 0.003925710916519165, -0.017839284613728523, 0.018236052244901657, 0.026354530826210976, -0.011567303910851479, 0.03778449073433876, -0.02035723440349102, 0.0017835469916462898, -0.01939583569765091, 0...
fe7fac7a0b0c5da5bc3687eae42a73742e6d7375
subsection
23
31
Edge-Preserved Smoothing
Edge-preserved image smoothing is a fundamental tool for image editing and processing, such as pencil sketch rendering  and cartoon artifact removal . Here we compare our method with state-of-the-art image smoothing approaches, including the classic BLF , WLS  and recently proposed L_0 , RTV . Fig. REF illustrates the ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.5555/2330147.2330161", "end": 150, "openalex_id": "https://openalex.org/W2039755782", "raw": "Cewu Lu, Li Xu, and Jiaya Jia. 2012. Combining sketch and tone for pencil drawing production. In NPAR. 65–73.", "source_ref_id": "5a06c38...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.05861515551805496, 0.011755069717764854, -0.007624394726008177, -0.05480105057358742, 0.029170271009206772, -0.03316745162010193, 0.0052825347520411015, 0.03182488679885864, 0.02386103756725788, 0.022350652143359184, -0.003796940902248025, 0.018307700753211975, -0.025325652211904526, 0....
7005f447dc3200b7f44ca55d866eb60d832857d7
subsection
24
31
Conclusions
This paper proposed GCM, a collaborative learning framework to estimate the latent image structures. By integrating the learnable-architecture-based Generator and the model-driven Corrector in a principled manner, we obtained a convergent image propagation, which can promote kernel estimation for blind image deblurring...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.06593262404203415, -0.004960209131240845, 0.02657145820558071, -0.0277771707624197, 0.01350703090429306, -0.02821977250277996, -0.012698134407401085, 0.0036552923265844584, 0.011858714744448662, 0.02583887241780758, -0.031104326248168945, -0.03879646584391594, -0.023610593751072884, 0.0...
87bb5708b5b112b3382c980797f096601c1c203d
subsection
25
31
Proof of Theorem
To present our proof in a clear manner, here we reorganize the results in Theorem REF as the following two successive theorems, referring to “non-increasing properties of the objectives” (i.e., Theorem REF ) and the “critical point convergence” (i.e., Theorem REF ), respectively. Moreover, our theoretical analysis is b...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 539, "openalex_id": "", "raw": "R Tyrrell Rockafellar and Roger J-B Wets. 2009. Variational analysis. Vol. 317. Springer Science & Business Media.", "source_ref_id": "41d3dbbcb556c239a6fc54df9251d1692beb6e04", "start":...
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.04236547648906708, 0.03287292271852493, -0.006310563534498215, -0.009477291256189346, 0.005986260250210762, -0.013178166002035141, 0.012361684814095497, 0.006611974909901619, 0.01985500194132328, 0.053262073546648026, -0.022815700620412827, -0.015322384424507618, -0.011888583190739155, ...
32dff8f896cff9cf9138dd4af78065015fcf09bf
subsection
26
31
Non-increasing Properties of the Objectives
Theorem 2 If \mu ^{t} < 1/L, both \lbrace \mathbf {u}^{t}\rbrace and \lbrace \mathbf {v}^{t}\rbrace are the sequence generated by GCM, we have the objectives \lbrace \Psi (\mathbf {u}^{t})\rbrace is a non-increasing sequence and satisfied the following relationship:\Psi (\mathbf {u}^{t+1}) \le \Psi (\mathbf {v}^{t+1})...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.04143429547548294, 0.02245629020035267, -0.02631596475839615, -0.01311221532523632, 0.02025947906076908, 0.014355549588799477, 0.007856650277972221, -0.02938234806060791, 0.019756043329834938, 0.04585842788219452, -0.017299886792898178, -0.008001578971743584, -0.008741477504372597, 0.02...
f6579290986b4482b881928f2bea8af3d2048915
subsection
27
31
Non-increasing Properties of the Objectives
(REF ) and (REF ), we have\Psi (\mathbf {u}^{t+1}) \le \Psi (\mathbf {v}^{t+1}) - \left( \frac{1}{2\mu ^t} - \frac{L}{2} \right) \Vert \mathbf {u}^{t+1} - \mathbf {v}^{t+1}\Vert ^2.Setting \mu ^t < \frac{1}{L}, we have \Psi (\mathbf {u}^{t+1}) \le \Psi (\mathbf {v}^{t+1}). So far, we get the following relationship of o...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.08240219205617905, -0.008064733818173409, -0.029161222279071808, 0.007683241739869118, -0.012299290858209133, 0.03970564901828766, 0.00493650184944272, -0.001453483127988875, 0.029161222279071808, 0.028566094115376472, -0.02407975308597088, 0.006363280583173037, -0.0009937857976183295, ...
8523329afc6144502c5086fb980b43d894a23870
subsection
28
31
Critical Point Convergence
Theorem 3 If \lbrace \mathbf {u}^t\rbrace be the image sequence generated by GCM, we have any accumulation point of \lbrace \mathbf {u}^t\rbrace is the critical point of \Psi (i.e., it satisfies the first-order necessary optimal condition of Eq. (REF )).In this proof, we first verify the existence of accumulation poin...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.02872474305331707, -0.00547646451741457, -0.020273596048355103, -0.021295666694641113, 0.005346798803657293, 0.018793882802128792, -0.02124990150332451, -0.020014265552163124, 0.010396129451692104, 0.02282114513218403, -0.02218044362962246, 0.0017285568173974752, -0.02274486981332302, 0...
2009235e5f9f295c8c1c61598de56ce0d73e1322
subsection
29
31
Critical Point Convergence
\end{array}Then sum over t to obtain\min \limits _{t}\left\lbrace \frac{1}{2\mu ^t} - \frac{L}{2} \right\rbrace \sum _{t=0}^{\infty } \Vert \mathbf {u}^{t+1} - \mathbf {v}^{t+1}\Vert ^2 \le \Psi (\mathbf {u}^{0}) - \Psi ^{*} < \infty ,which implies \Vert \mathbf {u}^{t+1} - \mathbf {v}^{t+1}\Vert \rightarrow 0 when t \...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.05305447801947594, 0.025596190243959427, -0.0361735075712204, -0.004075243137776852, -0.009623373858630657, 0.02730565518140793, -0.018575172871351242, 0.017766227945685387, 0.02058989927172661, 0.016056763008236885, -0.039897698909044266, 0.008333642967045307, -0.026481449604034424, 0....
b5312bdc40594992bccac98b784ee613e0dcd355
subsection
30
31
Critical Point Convergence
Note that Lipschitz smooth of f implies f is continuity, which yields \lim \limits _{j\rightarrow \infty } f(\mathbf {u}^{t_j}) = f(\mathbf {u}^{*}). Thus we conclude\lim \limits _{j\rightarrow \infty } \Psi (\mathbf {u}^{t_j}) = \Psi (\mathbf {u}^{*}).Recall that \lim \limits _{j\rightarrow \infty } \Psi (\mathbf {u}^...
{ "cite_spans": [] }
1807.11706
Learning Collaborative Generation Correction Modules for Blind Image Deblurring and Beyond
[ "Risheng Liu", "Yi He", "Shichao Cheng", "Xin Fan", "Zhongxuan Luo" ]
[ "cs.CV" ]
2,018
en
Computer Science
[ -0.036651335656642914, 0.0314939059317112, -0.056212957948446274, -0.03918427601456642, 0.013976945541799068, 0.0009131362894549966, 0.01451862882822752, -0.01962265558540821, 0.040618591010570526, 0.03750582039356232, -0.010078354738652706, 0.028350617736577988, -0.024764828383922577, 0.0...
952afb81bded3d5c88271906e5b213b70b8cc54d
abstract
0
16
Abstract
Given two graphs $G$ and $H$, it is said that $G$ percolates in $H$-bootstrap process if one could join all the nonadjacent pairs of vertices of $G$ in some order such that a new copy of $H$ is created at each step. Balogh, Bollob\'as and Morris in 2012 investigated the threshold of $H$-bootstrap percolation in the Erd...
{ "cite_spans": [] }
1806.10425
On $K_{2,t}$-bootstrap percolation
[ "M. R. Bidgoli", "A. Mohammadian", "B. Tayfeh-Rezaie" ]
[ "math.CO" ]
2,018
en
Mathematics
[ -0.04255402833223343, -0.016011182218790054, -0.006139655131846666, 0.012309836223721504, 0.008287961594760418, -0.011332986876368523, -0.009936396032571793, -0.008036118000745773, -0.009371655061841011, -0.004086743574589491, -0.026130737736821175, -0.03571607917547226, 0.0691121369600296, ...
c8d69586d5df6dc169067e7760983fd15d44fcd4
subsection
1
16
Introduction
Bootstrap percolation on graphs has been extensively investigated in several diverse fields such as combinatorics, probability theory, statistical physics and social sciences. Many different models of bootstrap percolation have been defined and studied in the literature including the r-neighbor bootstrap percolation an...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1002/rsa.20458", "end": 480, "openalex_id": "https://openalex.org/W2091222603", "raw": "J. Balogh, B. Bollobás and R. Morris, Graph bootstrap percolation, Random Structures Algorithms 41 (2012), 413–440.", "source_ref_id": "04af684...
1806.10425
On $K_{2,t}$-bootstrap percolation
[ "M. R. Bidgoli", "A. Mohammadian", "B. Tayfeh-Rezaie" ]
[ "math.CO" ]
2,018
en
Mathematics
[ -0.04053254425525665, -0.03020101971924305, 0.0052649579010903835, -0.015611744485795498, 0.03339051455259323, -0.06342366337776184, 0.004589669872075319, 0.016313739120960236, 0.01585591584444046, 0.01217044610530138, -0.04681997373700142, -0.014574013650417328, 0.045782241970300674, -0.0...
af675609aed1c27845c3129f8f0a231d7758f8f6
subsection
2
16
Introduction
A function { is a {\sl threshold} for a sequence {E}_n of events in {G}(n, p) if {\lim }_{n\rightarrow \infty }{P}({E}_n)= \left\lbrace \begin{array}{ll} 0, & \mbox{ if } p\ll {\mbox{;}\\ \vspace{-2.84526pt}\\ 1, & \mbox{ if } p\gg {\mbox{.} } We say that {E}_n holds {\sl with high probability} if \lim _{n\rightarrow \...
{ "cite_spans": [] }
1806.10425
On $K_{2,t}$-bootstrap percolation
[ "M. R. Bidgoli", "A. Mohammadian", "B. Tayfeh-Rezaie" ]
[ "math.CO" ]
2,018
en
Mathematics
[ -0.05010884255170822, -0.014854128472507, -0.006191127467900515, -0.004894156940281391, 0.016097694635391235, -0.01888999529182911, -0.015006713569164276, 0.00653825793415308, -0.028869040310382843, 0.030104978010058403, -0.016265537589788437, -0.0019092170987278223, 0.01916464790701866, -...
93f42164a204dde15bf07285fa0e1060422c4ff0
subsection
3
16
Introduction
The degree of a vertex v\in V(G), denoted by \deg _G(v), is defined as |N_G(v)|. A graph G is a complete split graph if one can partition V(G) into an independent set I and a clique C such that each vertex in I is adjacent to each vertex in C.
{ "cite_spans": [] }
1806.10425
On $K_{2,t}$-bootstrap percolation
[ "M. R. Bidgoli", "A. Mohammadian", "B. Tayfeh-Rezaie" ]
[ "math.CO" ]
2,018
en
Mathematics
[ -0.03087041899561882, -0.03671489283442497, 0.001518342411145568, 0.003055759472772479, 0.012741868384182453, -0.035066843032836914, 0.018113290891051292, 0.012574011459946632, 0.017319785431027412, 0.0030061653815209866, -0.0119102131575346, -0.007435299456119537, 0.028199966996908188, 0....