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
63063801fffe89a234ccb3984b62ae5928dca792
subsection
75
121
Conclusion and open problems
Similar to the previous proof, we have the following decomposition for the error term: R^t &= \sum _{i=1}^n \left[\nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \nabla f_{\sigma _t\left(i\right)}\left(x^t_{0}\right)\right]\\ &= \sum _{i=1}^n \left[\int _{x^t_0}^{x^t_{i-1}} H_{\sigma _t\left(i\right)}\left...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ 0.012566659599542618, 0.017900049686431885, -0.023973554372787476, 0.021699804812669754, 0.003008139319717884, -0.004059175960719585, 0.0009094041888602078, 0.01918189600110054, -0.00764528987929225, 0.008957655169069767, -0.0319240465760231, -0.018815653398633003, -0.010186090134084225, 0...
e3d5b5177cf181792afbcb8154381639b681d1bb
subsection
76
121
Conclusion and open problems
Here we define random variables A^t = -\gamma \sum _{i=1}^n \left[H_{\sigma _t\left(i\right)}\sum _{j=1}^{i-1} \nabla f_{\sigma _t\left(j\right)} \left(x^t_0\right)\right], B^t = - \gamma \sum _{i=1}^n \left\lbrace H_{\sigma _t\left(i\right)}\sum _{j=1}^{i-1}\left[\nabla f_{\sigma _t\left(j\right)} \left(x^t_{j-1}\righ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.04345737397670746, -0.005790756084024906, -0.023742862045764923, -0.008796760812401772, -0.0036220834590494633, -0.0072250934317708015, 0.00025534830638207495, 0.035278599709272385, 0.0003333117929287255, 0.02519245818257332, -0.04681433364748955, 0.03634672239422798, 0.006137896329164505...
4f528573952f582134c2e7e4619fe167d7cf4d66
subsection
77
121
Conclusion and open problems
\left\Vert C^t\right\Vert &\le \sum _{i=1}^n \left[\int _{0}^{\left\Vert x^t_{i-1}-x^t_0\right\Vert } \left\Vert H_{\sigma _t\left(i\right)}\left(x^t_0+t\frac{x^t_{i-1}-x^t_0}{\left\Vert x^t_{i-1}-x^t_0\right\Vert }\right)-H_{\sigma _t\left(i\right)}\right\Vert dt\right]\\ &\le \sum _{i=1}^n \left[L_H\max \left\lbrace ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.0267487782984972, 0.023101910948753357, -0.03213515505194664, -0.011413626372814178, -0.0015592646086588502, -0.011421255767345428, -0.022903546690940857, 0.040619079023599625, 0.01634223759174347, -0.0129395117983222, -0.02095041237771511, 0.0025444147177040577, 0.0017690738895907998, ...
ccd1ed928559b2ff756890cbf776220ed84501aa
subsection
78
121
Conclusion and open problems
Using (REF ) and  (REF ), we can decompose the inner product of x^t_0 - x^* and \operatorname{\mathrm {\mathbb {E}}}\left[R^t\right] into: -2\gamma \left\langle x^t_0 - x^*, \operatorname{\mathrm {\mathbb {E}}}\left[R^t\right] \right\rangle &= -2\gamma \left\langle x^t_0 - x^*, \operatorname{\mathrm {\mathbb {E}}}\lef...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.016512149944901466, 0.01646636798977852, -0.020342601463198662, 0.017885617911815643, 0.001797909033484757, -0.00850024726241827, 0.010026323609054089, 0.03574071452021599, 0.007615122478455305, 0.03268856182694435, -0.039891645312309265, 0.022891148924827576, -0.01587119698524475, -0.0...
c5689eadb55256165d3d097d8464ed802efc76f8
subsection
79
121
Conclusion and open problems
For the first term in (REF ), there is &\ \ \ \ \gamma ^2 n\left(n-1\right) \left\langle x^t_0 - x^*, \operatorname{\mathrm {\mathbb {E}}}\limits _{i\ne j} H_i \nabla f_j\left(x^t_0\right)\right\rangle \\ &= \gamma ^2 n\left(n-1\right)\operatorname{\mathrm {\mathbb {E}}}\limits _{i\ne j} \left\langle H_i \left(x^t_0 -...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.057683516293764114, 0.003836030140519142, -0.025011451914906502, -0.004574242047965527, -0.022478260099887848, -0.0014707008376717567, 0.016938811168074608, 0.059209536761045456, 0.02324126847088337, 0.012604916468262672, -0.0370212197303772, 0.011223869398236275, -0.031252868473529816, ...
aed4d90f2a7166d874e7bbc7ea35ee5530187062
subsection
80
121
Conclusion and open problems
The last inequality is because of \left\Vert H_i\left(x^t_0 - x^*\right) - \left(\nabla f_i\left(x^t_0\right) - \nabla f_i\left(x^*\right)\right)\right\Vert &=\left\Vert H_i\left(x^t_0 - x^*\right) - \int _{x^*}^{x^t_0} H_i\left(x\right) dx\right\Vert \\ &=\left\Vert \int _{x^*}^{x^t_0} \left(H_i - H_i\left(x\right)\r...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.02051360346376896, -0.0007936810725368559, -0.013759682886302471, -0.028343573212623596, -0.03009882941842079, 0.03455565124750137, 0.0002625729830469936, 0.04218720272183418, 0.007787995506078005, -0.009661540389060974, -0.04676613211631775, 0.016270462423563004, -0.01897202990949154, ...
6725a260ce0c651e0d982da1a935789c2f9f84eb
subsection
81
121
Conclusion and open problems
Furthermore, we have \operatorname{\mathrm {\mathbb {E}}}\left[\left\Vert R^t\right\Vert ^2\right] &\le \left[3n^2L\gamma (\left\Vert \nabla F(x^t_0)\right\Vert + \delta )\right]^2\\ &\le 18n^4L^2\gamma ^2 (\left\Vert \nabla F(x^t_0)\right\Vert ^2 + \delta ^2). Inequality (REF ) can be simplified to: \operatorname{\...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.04038400202989578, 0.016590101644396782, -0.009134476073086262, -0.058027200400829315, -0.012309031561017036, 0.008241632953286171, 0.01042413990944624, 0.04853406175971031, -0.00931762345135212, 0.005032737739384174, -0.051250748336315155, 0.03159292787313461, -0.03647685796022415, -0....
8fb7310bae906509343ec2c6ebddd7a5133aeffd
subsection
82
121
Conclusion and open problems
So there is \operatorname{\mathrm {\mathbb {E}}}[\left\Vert x^{t}_n - x^*\right\Vert ^2] &\le \left\Vert x^t_0-x^*\right\Vert ^2 - (2n\gamma -16n^2\gamma ^2L)\left[F(x^t_0) - F(x^*)\right] + \gamma ^2n\left(n-1\right)D\left\Vert \Delta \right\Vert \\ &\ \ \ \ +\gamma ^2n^2L_H (LD^3 + 2D^2G)+ 12\gamma ^4n^4L^4D^2 + 6\g...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.033298999071121216, 0.004849109333008528, -0.04242493957281113, -0.044317275285720825, -0.018343444913625717, 0.000665274157654494, 0.0003381214337423444, 0.039311740547418594, -0.007844035513699055, 0.0013810995733365417, -0.02109038271009922, 0.05213078856468201, -0.04422570765018463, ...
353be87860ffdcd0d646a7ae8bb824fbc510472d
subsection
83
121
Conclusion and open problems
Substituting the step size into (REF ), we have F(\bar{x}) - F(x^*) &\le \frac{D^2}{T}\max \left\lbrace 16nL, \sqrt{\frac{ Tn\left(\left\Vert \Delta \right\Vert +L_H LD^2 + 2L_HDG\right)}{D}}, \left(\frac{Tn^2L^2\delta }{D}\right)^\frac{1}{3}, (Tn^3L^4)^\frac{1}{4}\right\rbrace \\ &\ \ \ \ + \frac{D\sqrt{nD\left(\left...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.0427895151078701, 0.02198990248143673, -0.023988986387848854, -0.04266743361949921, -0.0060659172013401985, 0.03366393223404884, 0.00975887756794691, -0.010758419521152973, -0.00019766687182709575, -0.005566146690398455, -0.01030824426561594, 0.041904423385858536, -0.040347885340452194, ...
c2285bfa4305bd5507ce2748302623eacb838841
subsection
84
121
Conclusion and open problems
We have the following inequality ||x_{t} - x^*||^2 &= ||x_{t-1} - x^*||^2 -2\gamma \langle x_{t-1} - x^*, \nabla f_{s(t)}(x_{t-1}) \rangle + \gamma ^2 ||\nabla f_{s(t)} (x_{t-1})||^2\\ &= ||x_{t-1} - x^*||^2 -2\gamma \langle x_{t-1} - x^*, \nabla f_{s(t)}(x_{t-1}) - \nabla f_{s(t)}(x^*) \rangle + \gamma ^2 ||\nabla f_...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03716812655329704, 0.012358860112726688, -0.038815975189208984, -0.0037305448204278946, -0.0031145091634243727, 0.00660283537581563, 0.016737863421440125, 0.04998472332954407, -0.0035894098691642284, 0.006560876499861479, -0.0433017835021019, -0.01890447922050953, -0.03115653432905674, ...
c49c77e62d28ccb14ce4e86d9e74a3abe705962d
subsection
85
121
Conclusion and open problems
By the AM-GM inequality, we know the term \operatorname{\mathrm {\mathbb {E}}}[\prod _{i=1}^T (1-\gamma \mu _{s(t)})^2] for RandomShuffle is no larger than that of Sgd. Also, this bound is tight when we consider f_i(x) = \frac{\mu _i}{2}||x-x^*||^2, which completes the proof. Sgd under Polyak-Łojasiewicz condition For...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06336595863103867, 0.015963582322001457, -0.04523523524403572, -0.055887799710035324, 0.01404825784265995, 0.010316808708012104, -0.010988316498696804, -0.0041091726161539555, -0.0015280627412721515, 0.03333122655749321, -0.030172085389494896, -0.004605173133313656, -0.011636932380497456,...
ee4da081d3ee6eed0f0b04cfeec8a04bd3d9b08d
subsection
86
121
Proof of Theorem
Assume T = nl where l is positive integer. Notate x^t_i as the ith iteration for tth epoch. There is x_0^1 = x_0, x^t_n = x^{t+1}_0, x^l_n = x_T. Assume the permutation used in tth epoch is \sigma _t\left(\cdot \right).
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.0607658326625824, 0.021303139626979828, -0.00838544126600027, 0.03454893082380295, 0.012742207385599613, -0.013665445148944855, 0.009713072329759598, 0.0089653255417943, 0.019670305773615837, -0.006931913550943136, 0.011300125159323215, -0.0030958985444158316, -0.014428451657295227, 0.0...
b9545810b6bae2031d838f037348627890e136c0
subsection
87
121
Proof of Theorem
Define error termR^t = \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)}\left(x^t_0\right).For one epoch of RandomShuffle, We have the following inequality\left\Vert x^{t}_n - x^*\right\Vert ^2 &= \left\Vert x^t_0 - x^*\right\Vert ^2 - 2\gamma \le...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1518, "openalex_id": "https://openalex.org/W3141595720", "raw": "Y. Nesterov. Introductory lectures on convex optimization: A basic course, volume 87. Springer Science & Business Media, 2013.", "source_ref_id": "ac12aab6cc5b...
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.0318867452442646, 0.0003115727740805596, -0.024761812761425972, -0.0058433604426681995, -0.0036540073342621326, 0.01433377806097269, 0.007712320424616337, 0.035273756831884384, 0.014928794465959072, -0.01890319399535656, -0.03896590694785118, -0.00575563358142972, -0.028743840754032135, ...
4da5f65589392f9073e266ed3752c160cd119820
subsection
88
121
Proof of Theorem
Firstly, we give a bound on the norm of R^t:\left\Vert R^t\right\Vert &= \left\Vert \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)}\left(x^t_0\right)\right\Vert \\ &\le \sum _{i=1}^n \left\Vert \nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.030921747907996178, 0.027809740975499153, -0.023614633828401566, 0.000611150695476681, 0.020655177533626556, 0.0038213604129850864, -0.0031005649361759424, 0.025414716452360153, -0.007349063642323017, 0.026879189535975456, -0.05067688226699829, 0.01600242406129837, -0.0014349169796332717,...
021c57dca3dec5e113be0a5f0ce3b24f2fc0cb04
subsection
89
121
Proof of Theorem
We begin with the following decomposition:R^t &= \sum _{i=1}^n \left[\nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \nabla f_{\sigma _t\left(i\right)}\left(x^t_{0}\right)\right] \\ &= \sum _{i=1}^n \left[H_{\sigma _t\left(i\right)}\left(x^t_{i-1} - x^t_0\right)\right] \\ &= \sum _{i=1}^n \left\lbrace H_{\s...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.02135998010635376, 0.014768900349736214, -0.02714243158698082, 0.004786923993378878, 0.005000523757189512, 0.004531367216259241, 0.007502692751586437, 0.0267304889857769, 0.012938044965267181, 0.043086130172014236, -0.04201813042163849, 0.011984474025666714, -0.016126785427331924, 0.005...
2287647631a9db45c33f1a502334dafe757e63d9
subsection
90
121
Proof of Theorem
The first inequality is by \left\langle x^t_0 - x^*, H_iH_i\left(x^t_0 - x^*\right)\right\rangle \ge 0 and AM–GM inequality, where \lambda _1 is any positive number.
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.01607002690434456, 0.009736635722219944, -0.004280762281268835, -0.015589299611747265, -0.009629807434976101, -0.024753641337156296, 0.005341414362192154, 0.026646027341485023, 0.021487746387720108, 0.005158280022442341, -0.05539809912443161, -0.05961018428206444, -0.0045745400711894035, ...
29f37c2b9cdd6ee35f6bb0f4349d14657117ee86
subsection
91
121
Proof of Theorem
The second inequality comes from noticing that \operatorname{\mathrm {\mathbb {E}}}\limits _{i, j}H_iH_j = H^2 (with i, j uniformly sampled from all pairs of indices), and let \lambda _1 = \frac{1}{2}\mu \gamma ^{-1} n^{-1}.For the second term in (REF ), we use the bound-2\gamma \left\langle x^t_0 - x^*, \operatorname{...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06371182948350906, 0.024563230574131012, -0.013311440125107765, -0.014730310067534447, -0.0005263549392111599, 0.01754516363143921, -0.014570115134119987, 0.04985877871513367, 0.015622824430465698, 0.005976798478513956, -0.06316259503364563, 0.009993115440011024, -0.03240515664219856, 0...
d49c4f1df0bb6bcbc2fe44524bfc3a3918799bc3
subsection
92
121
Proof of Theorem
Toward this end, we use the following important fact:\left\Vert \Delta \right\Vert &= \left\Vert \operatorname{\mathrm {\mathbb {E}}}\limits _{i\ne j} H_i \nabla f_j\left(x^*\right)\right\Vert \\ &= \left\Vert \frac{1}{n\left(n-1\right)} \sum _{i\ne j} H_i \nabla f_j\left(x^*\right)\right\Vert \\ &= \left\Vert \frac{-1...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.060832999646663666, 0.030019696801900864, -0.015704253688454628, 0.018832895904779434, -0.03035545162856579, -0.01336158812046051, -0.013247125782072544, 0.02402186207473278, 0.005589584819972515, 0.008722041733562946, -0.06654085963964462, 0.014368858188390732, -0.01394916232675314, 0....
8c9f7423513173530e247dbd449c2e4111beaede
subsection
93
121
Proof of Theorem
Substituting (REF ) (REF ) back to (REF ) and using (REF ) , we finally get a recursion bound for one epoch:&\ \ \ \ \operatorname{\mathrm {\mathbb {E}}}\left\Vert x^t_n - x^*\right\Vert ^2 \\ &\le \left(1-2n\gamma \frac{L\mu }{L+\mu } +\frac{1}{2}\gamma \mu \left(n-1\right)\right)\left\Vert x^t_0 - x^*\right\Vert ^2-\...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03484300896525383, 0.023157481104135513, -0.016811294481158257, -0.016902824863791466, -0.013760242611169815, -0.020472556352615356, -0.0030224479269236326, 0.03328697010874748, 0.010960902087390423, 0.005869460292160511, -0.03252420946955681, 0.038534779101610184, -0.025491535663604736, ...
5f48043731ab88eda60c1d876b5ea88768eb441e
subsection
94
121
Proof of Theorem
Expanding (REF ) over all epochs leads to a final bound of RandomShuffle:\operatorname{\mathrm {\mathbb {E}}}\left[\left\Vert x_T - x^*\right\Vert ^2\right] \le \left(1-n\gamma \frac{L\mu }{L+\mu }\right)^{\frac{T}{n}}\left\Vert x_0 - x^*\right\Vert ^2 + \frac{T}{n} \left(\gamma ^3nC_1 + \gamma ^{5}n^{5}C_2 + \gamma ^4...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.056931525468826294, 0.02435997687280178, -0.02055945433676243, -0.032968390733003616, -0.012714197859168053, 0.024008924141526222, 0.0031194048933684826, 0.02724471129477024, 0.01901787705719471, -0.010928410105407238, -0.004594205878674984, 0.003995127510279417, -0.029396813362836838, ...
ef20abc028cc5ffc60410da129ed04dab4752b2b
subsection
95
121
Proof of Theorem
Or in the expanding version with constant dependence, we have\operatorname{\mathrm {\mathbb {E}}}\left[\left\Vert x_T - x^*\right\Vert ^2\right] \le \frac{\left(\log T\right)^2}{T^2}\left(D^2 + 128\frac{L^2G^2}{\mu ^4}\right) + \frac{n^3\left(\log T\right)^4}{T^3}128\frac{L^2G^2}{\mu ^4} + \frac{n^4\left(\log T\right)^...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.05425288900732994, 0.007826696150004864, -0.007727527525275946, 0.037928201258182526, 0.0051110005006194115, 0.018872559070587158, -0.021511971950531006, 0.01995578594505787, 0.013990409672260284, 0.01853691227734089, -0.018964100629091263, -0.00047462695511057973, 0.02016938105225563, ...
a4cb57cae9f81d537477e63b8c3859be69d7938b
subsection
96
121
Proof of Theorem
Again, define error termR^t = \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \sum _{i=1}^n \nabla f_{\sigma _t\left(i\right)}\left(x^t_0\right).We have the following decomposition for the error term:R^t &= \sum _{i=1}^n \left[\nabla f_{\sigma _t\left(i\right)} \left(x^t_{i-1}\right) - \nabla ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.01809144951403141, -0.004511422012001276, -0.03383375704288483, 0.022911768406629562, 0.010807582177221775, -0.03163715824484825, -0.0011879188241437078, 0.025398200377821922, 0.013934687711298466, 0.008984708227217197, -0.03944729268550873, 0.0006368617760017514, -0.008832166902720928, ...
ea0e14e66efdd80bd5331052ef6cdb0ff09f27d8
subsection
97
121
Proof of Theorem
There is\operatorname{\mathrm {\mathbb {E}}}\left[A^t\right] = -\frac{n\left(n-1\right)}{2} \gamma \operatorname{\mathrm {\mathbb {E}}}\limits _{i\ne j}\left[ H_{i} \nabla f_{j}\left(x^t_0\right)\right],\left\Vert B^t\right\Vert &\le \gamma \sum _{i=1}^n H_{\sigma _t\left(i\right)} \sum _{j=1}^{i-1} \left(\nabla f_{\si...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.025946762412786484, 0.0035085363779217005, -0.013209953904151917, -0.0016426589572802186, -0.008654008619487286, -0.008173230104148388, -0.031655050814151764, 0.03769911825656891, -0.005681577604264021, 0.03287607431411743, -0.046307336539030075, 0.026664113625884056, -0.00153391144704073...
ed2e6f4f3f312f4c04db519629c2628119cb36bc
subsection
98
121
Proof of Theorem
The last inequality is because of\left\Vert H_i\left(x^t_0 - x^*\right) - \left(\nabla f_i\left(x^t_0\right) - \nabla f_i\left(x^*\right)\right)\right\Vert &=\left\Vert H_i\left(x^t_0 - x^*\right) - \int _{x^*}^{x^t_0} H_i\left(x\right) dx\right\Vert \\ &=\left\Vert \int _{x^*}^{x^t_0} \left(H_i - H_i\left(x\right)\rig...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.04299446567893028, 0.026837212964892387, -0.013723747804760933, -0.030254799872636795, -0.016859078779816628, 0.03101765364408493, 0.007941313087940216, 0.0434216633439064, 0.029095260426402092, -0.0038772067055106163, -0.04268932342529297, 0.014662058092653751, -0.0075675141997635365, ...
019881a5fc7aa2ae02a6a521c1ff0b173d96e23b
subsection
99
121
Proof of Theorem
Further assume n\gamma \frac{L\mu }{L+\mu }<1, which we call assumption 3, expanding (REF ) over all the epochs we finally get a bound for RandomShuffle:\operatorname{\mathrm {\mathbb {E}}}\left\Vert x_T - x^*\right\Vert ^2 \le \left(1-\frac{1}{2}n\gamma \frac{L\mu }{L+\mu }\right)^{\frac{T}{n}}\left\Vert x_0 - x^*\rig...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06987273693084717, 0.03231232613325119, -0.013280396349728107, -0.017330879345536232, -0.0008533856016583741, -0.003415449522435665, -0.01974133588373661, 0.03167157247662544, 0.01728511042892933, -0.004721749108284712, -0.013303280808031559, 0.007589887827634811, -0.032098740339279175, ...
e0bb60f8008789179841bc0b917f81f46621fa89
subsection
100
121
Proof of Theorem
The first is satisfied whenn\gamma \frac{L\mu }{L+\mu }>\frac{1}{2}\gamma \mu \left(n-1\right),which is naturally satisfied and\frac{1}{2}n\gamma \frac{L\mu }{L+\mu }>\gamma ^2n^2\left(L_H LD + 3L_H G\right),which is equivalent to\frac{T}{\log T} > 16\frac{L+\mu }{L\mu ^2}\left(L_H LD + 3L_H G\right)n,which is obviousl...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03234345465898514, 0.022442085668444633, -0.031092435121536255, -0.006514459848403931, 0.010595533065497875, -0.002118725096806884, -0.0071895532310009, -0.00010560252121649683, -0.024272847920656204, 0.018246591091156006, 0.0027499564457684755, 0.016705699265003204, -0.023098109290003777...
9e1945ad2ca2167681fae5cfdaa639c358b815c0
subsection
101
121
Proof of Theorem
In this setting, we have:x_{t} &= x_{t-1} - \gamma A(x_{t-1} + (-1)^{\sigma (t)}b)\\ &= (I - \gamma A) x_{t-1} -(-1)^{\sigma (t)} \gamma A b.Expanding (REF ) over iterations leads to:x_T &= (I-\gamma A)^T x_0 - \sum _{t=1}^T (-1)^{\sigma (t)} \gamma (I-\gamma A)^{T-t} A b.Taking expectation of (REF ) over the randomnes...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.008133991621434689, 0.022860636934638023, -0.030033200979232788, -0.015146316960453987, 0.006577393040060997, 0.0072107152082026005, 0.04855978116393089, 0.034214653074741364, 0.00685208709910512, -0.0032600825652480125, -0.0007339477888308465, 0.024463018402457237, -0.00516958674415946, ...
1f01de5bad6e997b882c3583129ab0fde02e1e0b
subsection
102
121
Proof of Theorem
We can writeb = \sum _{i=1}^d b_i e_i.Since \left\langle e_i, e_j\right\rangle = 0 for i\ne j, we can simplify the last term in (REF ):\left\Vert \sum _{t=1}^T (-1)^{\sigma (t)} \gamma (I-\gamma A)^{T-t} A b\right\Vert ^2 &= \left\Vert \sum _{t=1}^T (-1)^{\sigma (t)} \gamma (I-\gamma A)^{T-t} A (\sum _{i=1}^d b_ie_i)\r...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.0374201275408268, 0.02827874943614006, -0.014726925641298294, -0.022006453946232796, 0.010667483322322369, -0.0162530317902565, 0.02176227606832981, 0.03677916154265404, 0.01405543927103281, 0.02647794410586357, -0.027546219527721405, 0.045416925102472305, 0.009034549817442894, 0.003553...
4996e76281ef03eb7254b43c5dbd29cc9a822638
subsection
103
121
Proof of Theorem
Then for any index pair t\ne u, over randomness of \sigma , there is\operatorname{\mathrm {\mathbb {E}}}\left[s_ts_u\right] &= \frac{2\frac{(\frac{T}{2})(\frac{T}{2}-1)}{2}}{\frac{T(T-1)}{2}} - \frac{(\frac{T}{2})(\frac{T}{2})}{\frac{T(T-1)}{2}}\\ &= -\frac{1}{T-1}.Using this fact, we can simplify the last term in (REF...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.055794112384319305, 0.030949249863624573, -0.013086282648146152, 0.008660601451992989, 0.040716271847486496, 0.015131252817809582, 0.00897345133125782, -0.0080577926710248, 0.008530883118510246, 0.006920850370079279, -0.01811477355659008, 0.032444823533296585, -0.01880151592195034, 0.01...
5fd4564fdb362a2f935b93baa660b3ed0e3896e6
subsection
104
121
Proof of Theorem
Now for the faster convergence rate (REF ) to hold, from (REF ) we know there must be(1-\gamma \lambda _i)^{2T} = o(\frac{1}{T}),\gamma ^2 \operatorname{\mathrm {\mathbb {E}}}\left[\left[ \sum _{t=1}^T (-1)^{\sigma (t)} (1-\gamma \lambda _i)^{T-t}\right]^2\right] = o(\frac{1}{T}),hold for any i.However with (REF ), we ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.043806418776512146, 0.03577778488397598, -0.038799967616796494, -0.026405958458781242, 0.004113529343158007, -0.019934209063649178, -0.010699755512177944, 0.03535040467977524, 0.012653491459786892, 0.024238532409071922, -0.030893445014953613, 0.007975210435688496, -0.01830100826919079, ...
ae828129ca343080b33771db417505d8661caee7
subsection
105
121
Proof of Theorem
Since (REF ), there is (1-\gamma \lambda _i)^{T} = o(1) , so\gamma ^2\left[- \frac{1}{T-1} \frac{-2(1-\gamma \lambda _i)^T+(1-\gamma \lambda _i)^{2T}}{\gamma ^2\lambda _i^2}\right] = o(\frac{1}{T}).Again, since |1-\gamma \lambda _1|<1, for i=2,3 there is|\frac{\gamma ^2}{2\gamma \lambda _i - \gamma ^2\lambda _i^2}|\le ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06059154123067856, 0.0376790314912796, -0.02365999110043049, -0.029639821499586105, 0.007962936535477638, -0.04796067625284195, 0.00798581913113594, 0.043628349900245667, -0.009618068113923073, 0.004259866196662188, -0.04350631311535835, 0.005914997775107622, -0.02950252778828144, 0.006...
fc36d78bec3eeaea70da7c3aadee47a3c56eb142
subsection
106
121
Proof of Theorem
As a result, no step size can leads to convergence of o(\frac{1}{T}).The idea is similar to the proof of theorem REF , with a slightly different analysis on the R^t term adopting the sparsity parameter. For any i, we use H_i to denote H_i(x^*).
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.013721507973968983, 0.03595981374382973, -0.046063970774412155, -0.01064599771052599, -0.0211088377982378, 0.015995705500245094, -0.039073482155799866, 0.01602623239159584, 0.021871991455554962, 0.004323266912251711, -0.011050469242036343, -0.013599403202533722, 0.01871253363788128, 0.0...
e6cec35197dfdae32f37560916e0fa7d901cb2b4
subsection
107
121
Proof of Theorem
Again, we have the following decomposition for the error term:R^t &= \sum _{i=1}^n \left[\nabla f_{\sigma _t(i)} (x^t_{i-1}) - \nabla f_{\sigma _t(i)}(x^t_{0})\right]\\ &= \sum _{i=1}^n \left[\int _{x^t_0}^{x^t_{i-1}} H_{\sigma _t(i)}(x) dx\right]\\ &= \sum _{i=1}^n \left[\int _{x^t_0}^{x^t_{i-1}} H_{\sigma _t(i)} dx\r...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.009118512272834778, -0.0009786186274141073, -0.022342262789607048, 0.008500437252223492, 0.013612909242510796, -0.008752245455980301, 0.014078373089432716, 0.019717352464795113, -0.012628567405045033, 0.03406279534101486, -0.03845799341797829, 0.023151101544499397, -0.014009697362780571, ...
fb56739cf26dbe7fea44c21fe3c9d23a3b2923ca
subsection
108
121
Proof of Theorem
The introduction of \rho in (REF ) is similar: if f_{\sigma _t(i)} and f_{\sigma _t(j)} depend on disjoint dimensions of variables and j<i, then there must be \int _{x^t_{j-1}}^{x^t_j} (H_{\sigma _t(i)}(x)-H_{\sigma _t(i)}) dx=0.With (REF ) (REF ), we can decompose the innerproduct of x^t_0 - x^* and \operatorname{\mat...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03926714509725571, 0.016780830919742584, -0.02849690243601799, -0.011319433338940144, -0.0002996236435137689, 0.020594656467437744, 0.010518530383706093, 0.044453948736190796, -0.01746731996536255, 0.025689927861094475, -0.041555438190698624, 0.012585623189806938, -0.02022852934896946, ...
bc6163ca4e6c3e393fdcd8b37a5f6ced250295f4
subsection
109
121
Proof of Theorem
The idea is similar to the proof of theorem REF . For any vector v not being zero, define vector value directional functiondir\left(v\right) = \frac{v}{\left\Vert v\right\Vert },with norm being \ell _2 norm. For the convenience of notation, we define dir\left(\vec{0}\right) = \vec{0}, where \vec{0} is the zero vector. ...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06438145041465759, -0.03383840247988701, 0.004409061744809151, 0.03148893639445305, 0.006365677807480097, 0.0498574860394001, -0.0060681807808578014, 0.03725580871105194, 0.003619549563154578, 0.01803290843963623, -0.010816694237291813, 0.0019194292835891247, 0.0300701055675745, -0.0049...
184a6fddb24ea8668d9ee0184f41b47f4f667a90
subsection
110
121
Proof of Theorem
For one epoch of RandomShuffle, we haveF(x^{t+1}_0) - F^* &\le F(x^t_0) - F^* - \gamma \left\langle \nabla F(x^t_0), n\nabla F(x^t_0) + R^t\right\rangle + \frac{L}{2}\gamma ^2 \left\Vert n\nabla F(x^t_0) + R^t\right\Vert ^2 \\ &\le (1-2n\mu \gamma )\left[F(x^t_0) - F^* \right] - \gamma \left\langle \nabla F(x^t_0), R^t...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06584832817316055, 0.027843652293086052, -0.02000165916979313, -0.03249697387218475, 0.017163895070552826, 0.008604832924902439, 0.018765859305858612, 0.03213081136345863, 0.013441236689686775, 0.0049851578660309315, -0.05364289507269859, 0.009993201121687889, -0.00945158489048481, 0.00...
eb38420fd2533351cdd0ab29debc360374400ab4
subsection
111
121
Proof of Theorem
Further assume n\gamma \mu <1, which we call assumption 2, expanding (REF ) over all the epochs we finally get a bound for RandomShuffle:\operatorname{\mathrm {\mathbb {E}}}\left\Vert x_T - x^*\right\Vert ^2 \le \left(1- n\gamma \mu \right)^{\frac{T}{n}}[F(x^t_0) - F^*] + \frac{T}{n} \left(\gamma ^3nC_1 + \gamma ^{4}n^...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.06058661267161369, 0.03467324376106262, -0.007558065466582775, -0.013941025361418724, -0.003510055597871542, -0.010072333738207817, -0.015222958289086819, 0.03265877813100815, 0.0151313915848732, -0.0015547256916761398, -0.010438600555062294, 0.0019066850654780865, -0.033818621188402176, ...
55cda189f3576fa4adc38bf5e4d2541536c28709
subsection
112
121
Proof of Theorem
The first is satisfied when\frac{T}{\log T} > 16 \frac{L^2}{\mu ^2} n.The second assumption is satisfied when\frac{T}{\log T} > 2 n.Since 2<\frac{L}{\mu }, the theorem is proved.For one epoch of RandomShuffle, We have the following inequality\left\Vert x^{t}_n - x^*\right\Vert ^2 &= \left\Vert x^t_0 - x^*\right\Vert ^2...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.055557552725076675, 0.029442450031638145, -0.0028809034265577793, -0.015690429136157036, -0.016300952062010765, 0.019582511857151985, -0.010768091306090355, 0.015446220524609089, -0.0058381217531859875, -0.00338458432815969, 0.012309661135077477, -0.008982312865555286, -0.0284350886940956...
11bb55c2adb7c420291f6de4116ae394d282acf7
subsection
113
121
Proof of Theorem
Obviously R^t_n = R^t.
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.019027741625905037, 0.05377206206321716, -0.0005302437930367887, 0.012466451153159142, -0.03326421603560448, 0.001524927793070674, -0.00031781248981133103, 0.01556399092078209, 0.003238683333620429, -0.002365116262808442, 0.006248484365642071, -0.024749796837568283, -0.01713564805686474, ...
942f5c2f929d3d8f62055bbeb21fa7f308dab4db
subsection
114
121
Proof of Theorem
We firstly show that \left\Vert R^t_k\right\Vert \le 3n^2L\gamma (\left\Vert \nabla F(x^t_0)\right\Vert + \delta ), which is an important fact to be used in further analysis.For any index 1\le id \le n, there is\left\Vert \nabla f_{id}(x^t_1) - \nabla f_{id}(x^t_0)\right\Vert \le L\gamma (\left\Vert \nabla F(x^t_0)\rig...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.055538687855005264, 0.03295702487230301, -0.024290546774864197, -0.029508743435144424, -0.0244888998568058, 0.031187109649181366, 0.01673789508640766, 0.02400064840912819, -0.01876719482243061, -0.010337214916944504, -0.03774799779057503, 0.012801362201571465, -0.0326518677175045, 0.003...
e43111f392b4148f4d5e834df9da418cba9891bf
subsection
115
121
Proof of Theorem
Since \gamma \le \frac{1}{16nL} \le \frac{1}{nL}, there is 1+\gamma L \le \frac{1}{n}.
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03705033287405968, 0.019150810316205025, -0.006264070048928261, -0.02156182937324047, -0.00919391494244337, -0.019135551527142525, 0.0031759063713252544, 0.0448632538318634, -0.018799839541316032, -0.034822430461645126, 0.011116625741124153, 0.014298254624009132, -0.04367300495505333, 0...
539fb31d343b197efe52efe91a6a718219a2325c
subsection
116
121
Proof of Theorem
Therefore, we have\left\Vert \nabla f_{id}(x^t_i) - \nabla f_{id}(x^t_0)\right\Vert &\le \left[\sum _{j=0}^{i-1} (1+L\gamma )^j\right] L\gamma (\left\Vert \nabla F(x^t_0)\right\Vert + \delta ) \\ &\le \left[n (1+\frac{1}{n})^n\right] L\gamma (\left\Vert \nabla F(x^t_0)\right\Vert + \delta )\\ &\le 3nL\gamma (\left\Vert...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.04532194137573242, 0.002693374641239643, -0.03689846768975258, -0.006618986837565899, 0.014786855317652225, 0.0328088104724884, 0.0027200793847441673, 0.017854098230600357, -0.0058865114115178585, 0.02073822170495987, -0.053165536373853683, 0.024232741445302963, -0.014039119705557823, -...
00becf8d2e08cda1029e0b8f1ff9f86380b21616
subsection
117
121
Proof of Theorem
The last inequality is because of\left\Vert H_i\left(x^t_0 - x^*\right) - \left(\nabla f_i\left(x^t_0\right) - \nabla f_i\left(x^*\right)\right)\right\Vert &=\left\Vert H_i\left(x^t_0 - x^*\right) - \int _{x^*}^{x^t_0} H_i\left(x\right) dx\right\Vert \\ &=\left\Vert \int _{x^*}^{x^t_0} \left(H_i - H_i\left(x\right)\rig...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.025826046243309975, 0.012058764696121216, -0.017542796209454536, -0.036214429885149, -0.02410227619111538, 0.03294994682073593, 0.005327670834958553, 0.04686214402318001, 0.013446934521198273, 0.007520520128309727, -0.04249932989478111, 0.010929924435913563, -0.01908351108431816, 0.0132...
ff597fcc741cc7072bbdf9b27422a8074cf42331
subsection
118
121
Proof of Theorem
\\ &\le \left\Vert x^t_0-x^*\right\Vert ^2 - (2n\gamma -3n^2\gamma ^2L)\left[F(x^t_0) - F(x^*)\right] + \gamma ^2n\left(n-1\right)D\left\Vert \Delta \right\Vert +\gamma ^2n^2L_H (LD^3 + 2D^2G) \\ &\ \ \ \ +12\gamma ^2n^2 \left\Vert \nabla F(x^t_0)\right\Vert ^2 + 12\gamma ^4n^4L^4D^2 + 6\gamma ^3n^3L^2D\delta + 36n^4L^...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.07370632141828537, 0.020729903131723404, -0.02780766226351261, -0.04176488146185875, -0.011318313889205456, 0.018579117953777313, 0.01825878955423832, 0.018243534490466118, -0.007665030658245087, 0.016169019043445587, -0.0177401602268219, 0.039232753217220306, -0.02677040547132492, 0.00...
13da47691a3d7530d6af50aa38a3b69dbfdeb819
subsection
119
121
Proof of Theorem
We have the following inequality||x_{t} - x^*||^2 &= ||x_{t-1} - x^*||^2 -2\gamma \langle x_{t-1} - x^*, \nabla f_{s(t)}(x_{t-1}) \rangle + \gamma ^2 ||\nabla f_{s(t)} (x_{t-1})||^2\\ &= ||x_{t-1} - x^*||^2 -2\gamma \langle x_{t-1} - x^*, \nabla f_{s(t)}(x_{t-1}) - \nabla f_{s(t)}(x^*) \rangle + \gamma ^2 ||\nabla f_{s...
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.056720368564128876, 0.005826964974403381, -0.026681164279580116, -0.02408631332218647, 0.0002659244346432388, 0.0018869902705773711, 0.008021139539778233, 0.053270746022462845, -0.00888354517519474, 0.002535702660679817, -0.043959811329841614, -0.011180751025676727, -0.03718267381191254, ...
327c8bd9642f09bdb139211beaa6ef072e0db947
subsection
120
121
Proof of Theorem
Also, this bound is tight when we consider f_i(x) = \frac{\mu _i}{2}||x-x^*||^2, which completes the proof.
{ "cite_spans": [] }
1806.10077
Random Shuffling Beats SGD after Finite Epochs
[ "Jeff Z. HaoChen", "Suvrit Sra" ]
[ "math.OC", "stat.ML" ]
2,018
en
Mathematics
[ -0.03258850798010826, 0.04555677995085716, 0.0005745135713368654, -0.0175758246332407, -0.04235285520553589, -0.00541234714910388, -0.028926877304911613, 0.037836845964193344, 0.030437299981713295, 0.017118120566010475, -0.05065254867076874, -0.024197272956371307, -0.013998107053339481, 0....
4c8df67f0883fcdbedca16d651ea6f90dadffbd7
abstract
0
74
Abstract
In this paper, we derive a theoretical analysis of an interior penalty discontinuous Galerkin methods for solving the Cahn-Hilliard-Navier-Stokes model problem. We prove unconditional unique solvability of the discrete system, obtain unconditional discrete energy dissipation law, and derive stability bounds with a gene...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.0033861109986901283, -0.012979773804545403, -0.018130477517843246, 0.0018122846959158778, -0.022373132407665253, -0.03876381739974022, 0.028584498912096024, -0.0037752753123641014, -0.003903089091181755, 0.03226248547434807, -0.03604729846119881, -0.00903471652418375, -0.02669209241867065...
1f1442f353b89b737ff8c71e845337d754fc35a5
subsection
1
74
Introduction
The Cahn–Hilliard–Navier–Stokes system strikes an optimal balance in terms of thermodynamical rigor and computational efficiency for modeling two-component binary flow. The model that belongs to the class of diffusive interphase or phase-field models, attracts much attention in physics, chemistry, biology, and engineer...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1137/050648110", "end": 738, "openalex_id": "https://openalex.org/W2036822565", "raw": "D. Kay and R. Welford, Efficient numerical solution of Cahn–Hilliard–Navier–Stokes fluids in 2D, SIAM Journal on Scientific Computing, 29 (2007), pp....
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.014132270589470863, -0.029577042907476425, -0.032812509685754776, 0.029424425214529037, 0.023457128554582596, 0.0012400075793266296, 0.02666207030415535, -0.0013745006872341037, -0.016345208510756493, 0.051797978579998016, -0.03467442840337753, -0.007199674844741821, -0.000107069885416422...
88a845d78ca1a37fc8ee510e01bf58265ffd5e5d
subsection
2
74
Introduction
The coupling term in the momentum equation of the Navier–Stokes system may take several forms, that yield different numerical methods and impact their analysis. We note that in , , , the coupling term is the product of the chemical potential and the gradient of the order parameter. In the other works , as well as in ou...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 282, "openalex_id": "", "raw": "X. Feng, Fully discrete finite element approximations of the Navier–Stokes–Cahn–Hilliard diffuse interface model for two-phase fluid flows, SIAM Journal on Numerical Analysis, 44 (2006), pp. 1049–10...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.0008360571227967739, -0.02455613762140274, -0.04248867928981781, 0.021763239055871964, 0.019122958183288574, -0.013285343535244465, 0.013697410002350807, 0.024510351940989494, -0.03174441680312157, 0.03436943516135216, -0.03211069852113724, 0.0131556186825037, 0.0027356664650142193, -0....
1ddcf1672c6bf3064e675749895a9bdce0a7f433
subsection
3
74
Mathematical Model
Let \Omega \subset {R}^d, where d=2 or 3, be an open bounded polyhedral domain and {{n}} denote the outward normal of \Omega . The unknown variables in Cahn–Hilliard–Navier–Stokes equations are the order parameter c, the chemical potential \mu , the velocity {{v}} and the pressure p, satisfying:\partial _t c - \Delta \...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.02598368003964424, 0.001277823350392282, -0.05227250978350639, 0.002919921651482582, 0.009680942632257938, -0.030042199417948723, 0.0011529017938300967, 0.031278062611818314, 0.03860170766711235, 0.01634088158607483, -0.040615711361169815, -0.011382163502275944, -0.0062975711189210415, ...
c7dfa8205ffb1eb16b1a5577d0914bf6584af725
subsection
4
74
Mathematical Model
Under the assumption of the incompressibility constraint (), it is possible to consider employing advection operator {{v}}\cdot \nabla {c} in (REF ) in nonconservative form instead of \nabla \cdot {(c{{v}})} in conservative form. However, for the convenience of proving discrete global mass conservation property, we pro...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1217, "openalex_id": "", "raw": "J. G. Heywood, R. Rannacher, and S. Turek, Artificial boundaries and flux and pressure conditions for the incompressible Navier–Stokes equations, International Journal for Numerical Methods in Flui...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.006509771570563316, -0.022911954671144485, -0.021188220009207726, 0.034901831299066544, -0.003943236079066992, -0.01983058638870716, 0.012241574004292488, 0.03249165415763855, 0.0320645347237587, 0.02707637846469879, -0.03526793420314789, -0.01934244856238365, -0.004404678475111723, -0....
5e9761261b1d8ccedbbd30278ab785d1fe3e0ec4
subsection
5
74
Well-posedness
A weak formulation of the Cahn–Hilliard–Navier–Stokes system (REF ) is proposed as finding the quaternion (c,\, \mu ,\, {{v}},\, p), wherec \in L^{\infty }\big (0,\, T;\, H^1(\Omega )\big ) \cap L^4\big (0,\, T;\, L^\infty (\Omega )\big ), && \mu \in L^2\big (0,\, T;\, H^1(\Omega )\big ), \\ {{v}} \in L^2\big (0,\, T;\...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.060530077666044235, 0.011418035253882408, -0.02024279534816742, -0.008168814703822136, -0.015330828726291656, 0.020654669031500816, 0.034505803138017654, 0.05314687266945839, 0.01121972594410181, 0.03261423856019974, -0.04698403179645538, -0.02730565518140793, 0.017237648367881775, -0.0...
13f8473b959863a045848210514f5f50523d7b51
subsection
6
74
Well-posedness
Standard notation is used for the Sobolev and Bochner spaces and we recall that L_0^2(\Omega ) denotes the space of L^2 functions with zero average. The existence of the weak solution to (REF ) follows the argument as in . A generalized version of Cahn–Hilliard–Navier–Stokes model, in which the deformation tensor {{\va...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1007/s00211-017-0887-5", "end": 222, "openalex_id": "https://openalex.org/W2963167146", "raw": "A. E. Diegel, C. Wang, X. Wang, and S. M. Wise, Convergence analysis and error estimates for a second order accurate finite element method fo...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.05481253191828728, -0.007103350479155779, -0.02957313135266304, -0.01142945047467947, 0.007694660220295191, -0.03354063257575035, 0.05182164907455444, 0.057070955634117126, 0.022401113063097, -0.012299248948693275, -0.020753074437379837, 0.014672119170427322, 0.012428955174982548, 0.008...
0993efff3ab84c9ef3efe0b0e1af396a8cc840ff
subsection
7
74
Mass conservation
Let \bar{c}_0 denote the mass average at time t_0. The solution of the model problem (REF ) enjoys the global mass conservation property . The total amount of the order parameter c is preserved, i. e., for any t\in (0,\, T), we have\frac{1}{|\Omega |} \int _\Omega c = \frac{1}{|\Omega |} \int _\Omega c^0 = \bar{c}_0.
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 138, "openalex_id": "", "raw": "F. Frank, C. Liu, F. O. Alpak, and B. Rivière, A finite volume/discontinuous Galerkin method for the advective Cahn–Hilliard equation with degenerate mobility on porous domains stemming from micro-C...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.007673853542655706, 0.03679788112640381, -0.02794930338859558, 0.0068347640335559845, 0.009413057006895542, -0.06303849816322327, -0.003434545360505581, 0.01403567660599947, 0.02257913164794445, 0.0055303615517914295, -0.031076818704605103, -0.01720895990729332, -0.006171120796352625, 0...
84ae8c7640321da50d6cca34ca77b5266d32a027
subsection
8
74
Energy dissipation
Benefitting from the boundary conditions (REF -) , the Cahn–Hilliard–Navier–Stokes model (REF ) is an energy dissipative system. Analysis of a similar model can be found in . Define the total energy as followsF(c,{{v}})=\underbrace{\int _\Omega \frac{1}{2}|{{v}}|^2}_{\text{kinetic energy}} + \underbrace{\int _\Omega \B...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1137/140971154", "end": 174, "openalex_id": "https://openalex.org/W2078273688", "raw": "J. Shen and X. Yang, Decoupled, energy stable schemes for phase-field models of two-phase incompressible flows, SIAM Journal on Numerical Analysis, 5...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.026431679725646973, 0.00004751111555378884, -0.018465548753738403, -0.0044714100658893585, -0.01649690791964531, -0.06070740148425102, 0.018450288102030754, 0.02981957420706749, 0.03210869058966637, -0.0030521568842232227, -0.01780933514237404, -0.031956084072589874, -0.03038422204554081,...
834d6a851e4fb28f4e766e058412db469938a44d
subsection
9
74
Chemical energy density
The chemical energy density \Phi may take several forms. Two popular expressions of \Phi are the Ginzburg–Landau double well potential ,\Phi (c) = \frac{1}{4}(1+c)^2(1-c)^2,and the logarithmic potential ,\Phi (c) = \frac{\vartheta }{2}\Big ((1+c)\log {(\frac{1+c}{2})}+(1-c)\log {(\frac{1-c}{2})}\Big )+\frac{\vartheta _...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 345, "openalex_id": "", "raw": "A. Novick-Cohen, The Cahn–Hilliard equation, Handbook of Differential Equations: Evolutionary Equations, 4 (2008), pp. 201–228, https://doi.org/10.1016/S1874-5717(08)00004-2.", "source_ref_id"...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.021160487085580826, 0.0311533622443676, -0.022914960980415344, -0.007971416227519512, -0.009298714809119701, -0.10777049511671066, 0.01246439665555954, 0.04140559583902359, 0.0259662214666605, 0.006491555366665125, -0.03432667255401611, 0.03203822672367096, 0.0009697285131551325, -0.013...
f259e5cc2bb29d3314f170edfef4a7fe095d87f0
subsection
10
74
Convex-concave decomposition
Throughout our analysis, we assume the chemical energy density \Phi \in \mathcal {C}^2, i. e., \Phi is a two times continuously differentiable function with respect to c. Any \mathcal {C}^2 function can be decomposed into the sum of a convex part and a concave part . We write\Phi (c) = \Phi _{+}(c)+\Phi _{-}(c),where \...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1162/08997660360581958", "end": 268, "openalex_id": "https://openalex.org/W2166765763", "raw": "A. L. Yuille and A. Rangarajan, The concave-convex procedure, Neural Computation, 15 (2003), pp. 915–936, https://doi.org/10.1162/08997660360...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.025302395224571228, 0.02843085676431656, -0.009095432236790657, 0.009164106100797653, 0.005966970231384039, -0.05316860228776932, -0.011918678879737854, 0.02011372521519661, 0.018755514174699783, 0.026874257251620293, -0.013887321576476097, 0.02507348358631134, -0.010346817784011364, -0...
094676fae1eb9d4b0708b142f3f8dfa6bfb63cb8
subsection
11
74
Numerical Analysis
In this section, we introduce an interior penalty discontinuous Galerkin method for the Cahn–Hilliard–Navier–Stokes system and analyze their numerical properties. These include uniquely solvability of the scheme, discrete mass conservation, energy dissipation, stability and error bounds. Our results are valid for any g...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ 0.008543595671653748, -0.014104560948908329, -0.016202319413423538, 0.012601803988218307, -0.015088600106537342, -0.0499495230615139, 0.04149746522307396, 0.008596993051469326, -0.009870904497802258, 0.035333871841430664, -0.045952338725328445, 0.01163302082568407, -0.025478223338723183, -...
b0e9291c2cdd9464e48a28e3a46afb911f24ec5c
subsection
12
74
Domain and triangulation
Let \mathcal {T}_h = \lbrace E_k\rbrace be a family of conforming nondegenerate (also called regular) meshes of the domain \Omega . The parameter h denotes the maximum element diameter. Let \Gamma _h denote the set of interior faces. For each interior face e \in \Gamma _h shared by elements E_{k^-} and E_{k^+}, we defi...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1377, "openalex_id": "", "raw": "B. Rivière, Discontinuous Galerkin Methods for Solving Elliptic and Parabolic Equations: Theory and Implementation, Frontiers in Applied Mathematics, Society for Industrial and Applied Mathematics,...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.03987669199705124, -0.022089308127760887, -0.03081519342958927, 0.010159861296415329, 0.008474178612232208, -0.0056519946083426476, -0.011593835428357124, -0.0020632450468838215, 0.005632925778627396, 0.032127127051353455, -0.0414327047765255, 0.03902240842580795, -0.016216114163398743, ...
2fffd8bf6134b6ee6524b563964a518116f89638
subsection
13
74
DG forms
We introduce the formsa_{\mathcal {A}}: H^2(\mathcal {T}_h) \times H^2(\mathcal {T}_h)^d \times H^2(\mathcal {T}_h) &\rightarrow {R}, \\ a_{\mathcal {C}}: H^2(\mathcal {T}_h)^d \times H^2(\mathcal {T}_h)^d \times H^2(\mathcal {T}_h)^d \times H^2(\mathcal {T}_h)^d &\rightarrow {R}, \\ a_{\mathcal {D}}: H^2(\mathcal {T}_...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.003953020088374615, 0.031242595985531807, -0.0012400883715599775, 0.012545878998935223, 0.002741549164056778, -0.045513153076171875, 0.006089787930250168, 0.04230799898505211, 0.017994636669754982, 0.041392240673303604, -0.04123961552977562, 0.01665152609348297, 0.031837839633226395, -0...
6883dc1ca9f2d56421e633796cc530d01eda79b7
subsection
14
74
DG forms
\mathrm {ext}) refers to the trace of the function on a face of E coming from the interior of E (resp. coming from the exterior of E on that face), in addition, if the face lies on the boundary of the domain, we take the exterior trace to be zero. For more details related to (), we refer the reader to . The derivation ...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 304, "openalex_id": "", "raw": "V. Girault, B. Rivière, and M. Wheeler, A discontinuous Galerkin method with nonoverlapping domain decomposition for the Stokes and Navier–Stokes problems, Mathematics of Computation, 74 (2005), pp....
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.062390509992837906, -0.010599366389214993, -0.0444730781018734, -0.02869230881333351, -0.019886212423443794, -0.021076638251543045, -0.005131802521646023, 0.023869559168815613, 0.06648068875074387, 0.03137839585542679, -0.008737418800592422, 0.03400344029068947, -0.001856224494986236, 0...
c72c999cac49cbc61521e7925c72e9158d826c71
subsection
15
74
DG scheme
Uniformly partition [0,\,T] into N subintervals and let \tau be the time step length. For any fixed positive integer q\in {N}_+ , the set {P}_q(E) denotes all polynomials of degree at most q on an element E. Define the following broken polynomial spacesS_h &= \big \lbrace \omega \in L^2(\Omega ):~\forall E \in \mathcal...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.04434514790773392, -0.004501658491790295, -0.017701435834169388, 0.0001018118709907867, -0.00986549910157919, 0.009392443113029003, 0.0155726857483387, 0.00579874636605382, 0.001323792734183371, 0.05383678153157234, 0.023240765556693077, 0.00998757779598236, -0.009628971107304096, 0.000...
5f76da7b8550d55e7602f96656796e3766aaa99b
subsection
16
74
DG scheme
The fully discrete mixed convex-concave splitting DG scheme reads:for any 1\le n \le N , given c_h^{n-1} \in S_h and {{v}}_h^{n-1} \in \mathbf {X}_h find (c_h^n, \mu _h^n, {{v}}_h^{n}, p_h^n) \in S_h \times S_h \times \mathbf {X}_h \times Q_h such that(\delta _\tau c_h^n,\chi ) + a_{\mathcal {D}}(\mu _h^n,\chi ) + a_{\...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.015004382468760014, -0.0027842398267239332, -0.03252602368593216, -0.013959338888525963, 0.016705438494682312, -0.019405769184231758, 0.02772034890949726, 0.024333491921424866, 0.03557724505662918, 0.06554023921489716, -0.015080663375556469, -0.0011966510210186243, -0.014005107805132866, ...
93c84effd6b0a287b3caf400a18cf8b346c13ec9
subsection
17
74
Operator properties
Throughout this paper, the semi-norms \Vert \cdot \Vert _{\mathrm {DG}} for any scalar quantity c \in H^1(\mathcal {T}_h) and for any vector quantity {{v}} \in H^1(\mathcal {T}_h)^d are defined as follows, respectively\forall c &\in H^1(\mathcal {T}_h), & \Vert c\Vert _{\mathrm {DG}}^2 &= \sum _{E\in \mathcal {T}_h} \V...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.06970454752445221, -0.007919583469629288, -0.009491292759776115, -0.0042916820384562016, 0.023499343544244766, -0.020554296672344208, -0.009300551377236843, 0.02043222077190876, 0.02262956276535988, 0.00986514613032341, -0.031220555305480957, 0.004528201185166836, 0.032685451209545135, ...
f095f966634adf594cb66b3d6a4b7049bb16bbe0
subsection
18
74
Operator properties
Then, we have the following result [Poincaré's inequality ] For each p \le p_0 (exclude infinity when d=2), there exists a constant C_P > 0 independent of mesh size h such that\Vert \chi - \frac{1}{|\Omega |}\int _{\Omega }\chi \Vert _{L^p(\Omega )} \le C_P\Vert \chi \Vert _{\mathrm {DG}}, \quad \forall \chi \in S_h.We...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1002/num.22092", "end": 318, "openalex_id": "https://openalex.org/W2551004786", "raw": "V. Girault, J. Li, and B. Rivière, Strong convergence of the discontinuous Galerkin scheme for the low regularity miscible displacement equations, Nu...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.039004143327474594, 0.0004768699000123888, -0.041415195912122726, -0.03567750006914139, -0.03424307331442833, -0.04471132159233093, 0.016785820946097374, 0.025468667969107628, 0.04080480337142944, -0.008148753084242344, -0.028093360364437103, -0.02047870121896267, 0.029192067682743073, ...
10ba11fc1ef3ce7eb23b6d6e8e985b3b3971f84d
subsection
19
74
Operator properties
Using Hölder's inequality and Cauchy–Schwarz's inequality we have|\sum _{E\in \mathcal {T}_h} \int _E c\,{{v}}\cdot \nabla {\chi }| \le \Big (\sum _{E\in \mathcal {T}_h} \Vert c\Vert _{L^4(E)}^4\Big )^{\frac{1}{4}}\Big (\sum _{E\in \mathcal {T}_h} \Vert {{v}}\Vert _{L^4(E)}^4\Big )^{\frac{1}{4}}\Big (\sum _{E\in \mathc...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.046026721596717834, 0.05091019347310066, -0.017839932814240456, 0.00021925548207946122, -0.005871611647307873, 0.01095728948712349, 0.008973379619419575, 0.015451610088348389, 0.006031850818544626, 0.0024302902165800333, -0.0520089752972126, 0.009187031537294388, 0.03281082585453987, -0...
74727ce20323ffd6827f8cb2357ef97d177544e3
subsection
20
74
Operator properties
For the inequality eq:CHNS:boundednessaA2, using similar arguments as above, we have|a_{\mathcal {A}}(c,{{v}},\chi )| \le C \Vert c\Vert _{L^6(\Omega )}\Vert {{v}}\Vert _{L^3(\Omega )}\Vert \chi \Vert _{\mathrm {DG}}.Finally, we conclude our proof by applying Poincaré's inequality and interpolation inequality \Vert {{v...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.04886240512132645, 0.030718307942152023, -0.010781164281070232, -0.060582056641578674, -0.025652000680565834, -0.030016349628567696, 0.022142207249999046, 0.01464956533163786, 0.030642008408904076, -0.01663336157798767, -0.027116956189274788, -0.012658140622079372, 0.024965301156044006, ...
390e9f3f6267b68fd47a6165484f5c625d232336
subsection
21
74
Operator properties
Then, there exists a constant K_\alpha >0 independent of mesh size h such thata_{\mathcal {D}}(c,c) \ge K_\alpha \Vert c\Vert _{\mathrm {DG}}^2, \quad \forall c \in S_h.[Continuity of a_{{\varepsilon }}] The bilinear form a_{{\varepsilon }} is continuous on \mathbf {X}_h equipped with the energy norm, i. e., there exis...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.010832175612449646, 0.008665740489959717, -0.055747561156749725, -0.03441885486245155, -0.0385991595685482, -0.04497641324996948, 0.017300967127084732, 0.018826626241207123, 0.033442433923482895, -0.005602980963885784, -0.007902910932898521, 0.010244796983897686, -0.007483354769647121, ...
bea2f3cc1190793d21ddcedea6d35a410013ded6
subsection
22
74
Discrete mass conservation
The DG scheme (REF ) satisfies the discrete global mass conservation property, i. e., for any 1\le n \le N , we have(c_h^n,1) = (c_h^{0},1) = (c^0,1) = \big (c(t^n),1\big ).The proof for the first equality is straightforward and obtained by choosing \chi = 1 in (REF ) and by using a_{\mathcal {D}}(\mu _h^n,1)=0 and a_{...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ 0.0026546521112322807, 0.010969510301947594, -0.013158835470676422, -0.023647762835025787, -0.00016961072105914354, -0.031794797629117966, 0.04009440168738365, -0.0028606164269149303, 0.007143912836909294, 0.05263534188270569, -0.054130490869283676, -0.003337385831400752, -0.0363717861473560...
9d01d066de2977fb106ef7fd3fbcd822bd2b5d79
subsection
23
74
Existence and uniqueness
Investigating the unique solvability of the fully discrete DG method (REF ) is a complicated task. We will design an equivalent scheme, which is based on an auxiliary flow problem, to overcome this challenge. The existence and uniqueness of the solution for our equivalent scheme can be proved by using nonlinear operato...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.03744636848568916, -0.025498321279883385, -0.02012704126536846, -0.019364075735211372, -0.004978352691978216, -0.012009083293378353, 0.01680050976574421, 0.014099610038101673, 0.02998455986380577, 0.05746659263968468, -0.029679374769330025, -0.024704836308956146, -0.023270459845662117, ...
47e19f22942483f1377e1b0d0bb92d59c28d32d3
subsection
24
74
Existence and uniqueness
To recover the discrete pressure \tilde{p}_h^n\in Q_h, we then use the inf-sup condition of lem:infsup.Owing to the last result, we can construct the following scheme by employing the unique discrete solution from the auxiliary flow problem: for any 1\le n \le N , given (y_h^{n-1},{{v}}_h^{n-1}) \in M_h\times \mathbf {...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.02664467878639698, 0.001940933638252318, 0.002659126650542021, -0.03168061375617981, 0.008500049822032452, 0.0014249410014599562, 0.03332873806357384, 0.03940238431096077, 0.016969578340649605, 0.0637885183095932, -0.03751009330153465, 0.014856010675430298, -0.0364418625831604, 0.012643...
3f1e84a4494998224e19889452bec14542acae70
subsection
25
74
Existence and uniqueness
We also denote \hat{y}_h^{n-1} \in M_h the solution of(y_h^{n-1},\mathring{\chi }) - \tau a_{\mathcal {A}}(y_h^{n-1}+\bar{c}_0,\tilde{{{v}}}_h^n,\mathring{\chi }) = (\hat{y}_h^{n-1},\mathring{\chi }), && \forall \mathring{\chi } \in M_h,whose existence and uniqueness are asserted by the Riesz representation theorem. Ou...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ 0.0033349976874887943, 0.0076392050832509995, 0.002064340515062213, -0.020330511033535004, -0.018193671479821205, 0.0004090046277269721, 0.031152082607150078, 0.035593658685684204, 0.030434714630246162, 0.04676628112792969, -0.04227891564369202, 0.006895126774907112, -0.01340104267001152, ...
eb8d0fec5c21c0d848c143334bede01ddd16c447
subsection
26
74
Existence and uniqueness
Due to the translational invariance of the trilinear form a_{\mathcal {A}} with respect to the third argument and using the same techniques as in , we have The unique solvability of the DG scheme (REF ) is equivalent to the unique solvability of the problem: for any 1\le n \le N , given (y_h^{n-1},{{v}}_h^{n-1}) \in M...
{ "cite_spans": [ { "arxiv_id": "", "doi": "", "end": 1336, "openalex_id": "", "raw": "C. Liu, F. Frank, and B. Rivière, Numerical error analysis for non-symmetric interior penalty discontinuous Galerkin method of Cahn–Hilliard equation, Numerical Methods for Partial Differential Equ...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.049579668790102005, -0.019511505961418152, -0.02393553964793682, -0.07047940790653229, -0.007124217227101326, 0.02295920066535473, 0.016170600429177284, 0.04591840133070946, 0.04286734387278557, 0.03606348857283592, -0.04497257247567177, 0.018108021467924118, -0.0035811776760965586, 0.0...
562388e30f21a774e25fe4826141414d50d2c612
subsection
27
74
Existence and uniqueness
(Necessity) If (REF ) has a solution (y_h^n, \mu _h^n, {{v}}_h^{n}, p_h^n). Define w_h^n=\mu _h^n-\frac{1}{|\Omega |}(\mu _h^n,1), then (y_h^n, w_h^n, {{v}}_h^{n}, p_h^n) is a solution of eq:CHNS:solvability:P. If the solution of (REF ) is unique. Assume (y_h^{n,1}, w_h^{n,1},{{v}}_h^{n,1},p_h^{n,1}) and (y_h^{n,2}, w_...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.023470258340239525, 0.007103644777089357, -0.010613501071929932, -0.03598365932703018, -0.02739214152097702, -0.004318649880588055, 0.05096922069787979, 0.012940688990056515, 0.03699083626270294, 0.03253484517335892, -0.045811258256435394, -0.014817698858678341, 0.005520394071936607, 0....
a20737e8d13cc02cf15ff70dfb90caca1a9828ff
subsection
28
74
Existence and uniqueness
Assume (y_h^{n,1}, \mu _h^{n,1}, {{v}}_h^{n,1}, p_h^{n,1}) and (y_h^{n,2}, \mu _h^{n,2}, {{v}}_h^{n,2}, p_h^{n,2}) are two different solutions of eq:CHNS:solvability:Paux, then \big (y_h^{n,1}, \mu _h^{n,1}-\frac{1}{|\Omega |}(\mu _h^{n,1},1),{{v}}_h^{n,1},p_h^{n,1}\big ) and \big (y_h^{n,2}, \mu _h^{n,2}-\frac{1}{|\Om...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.02326611988246441, 0.006506885401904583, -0.030070506036281586, -0.045677877962589264, -0.007776986807584763, 0.00746041489765048, 0.0058928122743964195, 0.022121883928775787, 0.01850610040128231, 0.05754741281270981, -0.022594835609197617, -0.03078756108880043, -0.00427181227132678, 0....
6ad255d0c5c4b1d4bd1b8e54ab558384f078ef85
subsection
29
74
Existence and uniqueness
Then (y_h^{n,1},w_h^{n,1},{{v}}_h^{n,1}-\tilde{{{v}}}_h^n,p_h^{n,1}-\tilde{p}_h^n) and (y_h^{n,2},w_h^{n,2},{{v}}_h^{n,2}-\tilde{{{v}}}_h^n,p_h^{n,2}-\tilde{p}_h^n) are two different solutions of (REF ). By contradiction argument, we know the solution of (REF ) is unique.(Sufficiency) If (REF ) has a solution (y_h^n,w_...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.009569760411977768, -0.0029438077472150326, -0.01122577115893364, -0.008661624975502491, 0.0005885707796551287, -0.02669459395110607, 0.02866349183022976, 0.03458544984459877, 0.048901934176683426, 0.041484225541353226, -0.03922533243894577, -0.006482863798737526, -0.010081063024699688, ...
a4c723af34b2a323dfa4d9ca899b17b611dd8d71
subsection
30
74
Existence and uniqueness
We first prove the existence of a solution. For each fixed w_h \in M_h , define the mapping \mathcal {F}:~M_h \rightarrow M_h by\big (\mathcal {F}(y_h),\mathring{\varphi }\big ) = \big (\Phi _{+}\,\!^{\prime }(y_h+\bar{c}_0)+\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),\mathring{\varphi }\big )\\ + \kappa a_{\mathcal {...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.009154790081083775, 0.002353543881326914, -0.01748564839363098, 0.005149569362401962, -0.010245735757052898, 0.008033327758312225, 0.043393705040216446, 0.03948765993118286, 0.051663532853126526, 0.049222253262996674, -0.053006235510110855, 0.007167437579482794, -0.020933952182531357, 0...
537dbdb9f45266dc2bf2dd3930fe0d518576e618
subsection
31
74
Existence and uniqueness
Applying Cauchy–Schwarz's inequality, Young's inequality, and Poincaré's inequality, we have&- \big (\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),y_h\big ) + (w_h,y_h)\\ \le & \Vert \Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0)\Vert _{L^2(\Omega )}\Vert y_h\Vert _{L^2(\Omega )} + \Vert w_h\Vert _{L^2(\Omega )}\Vert y_h\...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.04276032745838165, 0.019457321614027023, -0.022463668137788773, -0.034550100564956665, -0.00454004155471921, -0.01570320315659046, 0.012994439341127872, 0.006466698367148638, 0.052099842578172684, 0.006619304418563843, -0.03799900412559509, 0.009766812436282635, -0.013368324376642704, 0...
b2ea934901d2ad6584c80807debb0e4c68dc9631
subsection
32
74
Existence and uniqueness
By Brouwer's fixed point theorem, there exists a function y_h \in M_h such that \mathcal {F}(y_h) = 0. In particular \big (\mathcal {F}(y_h),\mathring{\varphi }\big ) = 0 for all \mathring{\varphi } \in M_h, i. e., the function y_h is a solution of (REF ). Next, let us prove the solution of (REF ) is unique. Assume y_h...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.0007954190950840712, -0.02075337991118431, -0.011856894940137863, -0.03308333083987236, 0.0014182112645357847, -0.00006777514499844983, 0.03790543973445892, 0.052982158958911896, 0.07007317990064621, 0.03439567610621452, -0.046939264982938766, 0.005909372586756945, 0.005550766363739967, ...
b4fd84fcaf445642b47fc52dc3538e3dc25e2336
subsection
33
74
Existence and uniqueness
Due to the fact that \Vert \cdot \Vert _{\mathrm {DG}} is a norm in M_h, we obtain y_h=\tilde{y}_h, i. e., the solution of (REF ) is unique. For each fixed w_h \in M_h , given (y_h^{n-1},{{v}}_h^{n-1}) \in M_h \times \mathbf {X}_h and \bar{c}_0 \in S_h , there exists a unique solution ({{v}}_h, p_h) \in \mathbf {X}_h...
{ "cite_spans": [ { "arxiv_id": "", "doi": "10.1137/1.9781611972597", "end": 1168, "openalex_id": "https://openalex.org/W333643410", "raw": "P. G. Ciarlet, Linear and nonlinear functional analysis with applications, vol. 130, SIAM, 2013.", "source_ref_id": "c3b732ba12711d5f8f8b...
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.02559538371860981, -0.019612444564700127, -0.033180899918079376, -0.019536131992936134, -0.022359712049365044, -0.0033043534494936466, 0.0012839664705097675, -0.0028025954961776733, 0.03333352506160736, 0.040934301912784576, -0.0319598913192749, -0.011805622838437557, 0.002016571350395679...
d6e59a14869c7af4c7a37f2f224e5560399025ba
subsection
34
74
Existence and uniqueness
By triangle inequality, Cauchy–Schwarz's inequality, Poincaré's inequality, and the continuity of a_{\mathcal {D}}, we have|\langle \mathcal {G}(w_h),\mathring{\chi }\rangle | \le & \Vert y_h\Vert _{L^2(\Omega )}\Vert \mathring{\chi }\Vert _{L^2(\Omega )} + \Vert \hat{y}_h^{n-1}\Vert _{L^2(\Omega )}\Vert \mathring{\chi...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.05485693737864494, 0.02460629679262638, -0.008458890952169895, -0.048663411289453506, -0.016307583078742027, 0.0025952549185603857, 0.012844700366258621, 0.020762039348483086, 0.012791307643055916, -0.001350066508166492, -0.04140203818678856, 0.007734279148280621, 0.02411813661456108, -...
7d7bc77d29b13e421d0dde54d06733d8d8e623a6
subsection
35
74
Existence and uniqueness
Since y_h=y_h(w_h) \in M_h is the unique solution of (REF ) which is defined in CHNS:solvability:uniqc, take \mathring{\varphi }=y_h then\big (\Phi _{+}\,\!^{\prime }(y_h+\bar{c}_0)+\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),y_h\big ) + \kappa a_{\mathcal {D}}(y_h,y_h) - (w_h,y_h) = 0.Recall the nonnegativity of \big...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.008529779501259327, -0.01980617828667164, -0.006420223042368889, -0.029663490131497383, -0.01169602107256651, -0.020309725776314735, 0.03222700208425522, 0.03927665948867798, 0.0352788008749485, 0.029037872329354286, -0.056031037122011185, 0.006012044847011566, 0.0174105167388916, 0.029...
08d4ed76cd3d4067e8efd2218c69ab327d714411
subsection
36
74
Existence and uniqueness
By the coercivity of a_{\mathcal {D}}, Cauchy–Schwarz's inequality and Poincaré's inequality, we haveK_\alpha \kappa \Vert y_h\Vert _{\mathrm {DG}}^2 \le & \big (\Phi _{+}\,\!^{\prime }(y_h+\bar{c}_0),y_h\big ) + \kappa a_{\mathcal {D}}(y_h,y_h)\\ =& (w_h,y_h) - (\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),y_h\big )\\...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.03342302888631821, 0.008920438587665558, -0.016925176605582237, -0.04218321666121483, -0.02541065774857998, -0.002434234134852886, 0.023563997820019722, 0.029989155009388924, 0.03134744241833687, 0.020923729985952377, -0.04050443693995476, 0.02423550933599472, 0.007069962099194527, 0.00...
5ba0d8aff52d3d6a7c27553c9d5e64c8cdd3aa6a
subsection
37
74
Existence and uniqueness
By the positivity of a_{\mathcal {C}}, the coercivity of a_{{\varepsilon }}, and considering {{v}}_h is discrete divergence-free, we obtain\begin{split} &a_{\mathcal {A}}\big (y_h^{n-1}+\bar{c}_0,{{v}}_h,w_h\big ) = b_\mathcal {I}\big (y_h^{n-1}+\bar{c}_0,w_h,{{v}}_h\big ) \\ =& \frac{1}{\tau }({{v}}_h,{{v}}_h) + a_{\m...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.0026914537884294987, 0.0316794216632843, -0.009300870820879936, -0.047458093613386154, 0.002792550250887871, -0.02577386423945427, 0.008827815763652325, 0.002561745001003146, 0.014542624354362488, 0.021592669188976288, -0.03961453586816788, 0.003921777941286564, 0.003406758652999997, -0...
da8f30fdc61a8b8d4e00eee51e94b0b78182d1d9
subsection
38
74
Existence and uniqueness
\end{split}Taking \mathring{\chi } = w_h in (REF ) and combining the result with (REF ), we obtain the following bound\Vert {{v}}_h\Vert _{\mathrm {DG}} \le \frac{C_\gamma }{K_{{\varepsilon }}\mu _\mathrm {s}}(\Vert y_h^{n-1}+\bar{c}_0\Vert _{\mathrm {DG}} + |\Omega ||\bar{c}_0|) \Vert w_h\Vert _{\mathrm {DG}}.Substitu...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.061921071261167526, 0.017578382045030594, -0.020706478506326675, -0.052155300974845886, -0.007160291541367769, -0.03494313731789589, -0.0010290296049788594, 0.00997176393866539, 0.021087953820824623, 0.009857322089374065, -0.05960170179605484, -0.0168917253613472, 0.03842219337821007, -...
0826f07ff472ce81299fa1fef073bde1df7dc29f
subsection
39
74
Existence and uniqueness
By the coercivity of a_{\mathcal {D}}, Cauchy–Schwarz's inequality, Young's inequality, and Poincaré's inequality, we have\begin{split} -(w_h,y_h) \le & - \big (\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),y_h\big ) - \kappa a_{\mathcal {D}}(y_h,y_h)\\ \le & \Vert \Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0)\Vert _{L^2...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.027068879455327988, 0.02764870785176754, -0.016509879380464554, -0.02499369904398918, -0.004417384508997202, -0.004062620457261801, 0.020812824368476868, 0.02714517153799534, 0.025939736515283585, 0.0033511852379888296, -0.047973256558179855, 0.012946981005370617, 0.00010383047629147768, ...
3cc67717985b5f837cad37006d6aa0b752c622ca
subsection
40
74
Existence and uniqueness
\end{split}Using the definition of \mathcal {G}, the coercivity of a_{\mathcal {D}}, the bounds (REF ), (REF ), and (REF ), we obtain\langle \mathcal {G}(w_h),w_h\rangle =& (y_h-\hat{y}_h^{n-1},w_h) + \tau a_{\mathcal {D}}(w_h,w_h) + \tau a_{\mathcal {A}}(y_h^{n-1}+\bar{c}_0,{{v}}_h,w_h)\\ \ge & K_\alpha \tau \Vert w_h...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.028309252113103867, 0.012857436202466488, 0.018710050731897354, -0.08595024794340134, -0.0020487962756305933, -0.04035022109746933, 0.007195586338639259, 0.0027279129717499018, 0.0065240999683737755, 0.027790376916527748, -0.046149421483278275, -0.02237270213663578, 0.005318476818501949, ...
d219b488d211112fd590a32fa5b78a9b09e2080e
subsection
41
74
Existence and uniqueness
From CHNS:solvability:uniqc, for any \mathring{\varphi } \in M_h, we obtain(w_h,\mathring{\varphi }) &= \big (\Phi _{+}\,\!^{\prime }(y_h(w_h)+\bar{c}_0)+\Phi _{-}\,\!^{\prime }(y_h^{n-1}+\bar{c}_0),\mathring{\varphi }\big ) + \kappa a_{\mathcal {D}}(y_h(w_h),\mathring{\varphi }),\\ \,\,(s_h,\mathring{\varphi }) &= \bi...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.007096332963556051, 0.010949717834591866, -0.03198691084980965, -0.04038042575120926, -0.022906657308340073, -0.007516008801758289, 0.03366561606526375, 0.029133116826415062, 0.032536305487155914, 0.03601579740643501, -0.05130724981427193, -0.0021994817070662975, -0.008775035850703716, ...
c2a5862075b85a5fd4301ff0195e0d7af7dbd0ff
subsection
42
74
Existence and uniqueness
\end{split}From CHNS:solvability:uniqvandp, for any {{\theta }} \in \mathbf {X}_h, we obtainb_\mathcal {I}(y_h^{n-1}+\bar{c}_0,w_h,{{\theta }}) =& \frac{1}{\tau }\big ({{v}}_h(w_h),{{\theta }}\big ) + a_{\mathcal {C}}\big ({{v}}_h^{n-1},{{v}}_h^{n-1},{{v}}_h(w_h),{{\theta }}\big ) \\ &+ \mu _\mathrm {s} a_{{\varepsilon...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.013412553817033768, 0.022903576493263245, -0.04434230178594589, -0.005077391862869263, -0.008163499645888805, -0.01686106063425541, 0.052215881645679474, 0.025711210444569588, 0.01417549792677164, 0.031433288007974625, -0.03186053782701492, -0.0018606294179335237, 0.00007957266643643379, ...
14624bde25330e0f73aa931829ade813e30ab1f6
subsection
43
74
Existence and uniqueness
Using the positivity of a_{\mathcal {C}}, the coercivity of a_{{\varepsilon }}, considering {{v}}_h(w_h) and {{v}}_h(s_h) are discretely divergence-free, and by rem:CHNS:relationaAbI, we obtain\begin{split} & a_{\mathcal {A}}\big (y_h^{n-1}+\bar{c}_0,{{v}}_h(w_h)-{{v}}_h(s_h),w_h-s_h\big ) \\ =\,& b_\mathcal {I}\big (y...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.04712660610675812, 0.01726042479276657, 0.01712307333946228, -0.05182705819606781, 0.0016062420327216387, -0.01459734421223402, 0.016619455069303513, -0.0011941894190385938, 0.03302524983882904, 0.032658983021974564, -0.04941578954458237, -0.018923897296190262, 0.005368129350244999, 0.0...
76f1c20c81557aad400a4671066ad95926396189
subsection
44
74
Existence and uniqueness
For any \mathring{\chi } \in M_h with \Vert \mathring{\chi }\Vert _{\mathrm {DG}}=1, by triangle inequality, Cauchy–Schwarz's inequality, the continuity of a_{\mathcal {D}}, the boundedness of a_{\mathcal {A}}, and Poincaré's inequality, we have\begin{split} &|\langle \mathcal {G}(w_h)-\mathcal {G}(s_h),\mathring{\chi ...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.043903566896915436, 0.010396206751465797, -0.013096321374177933, -0.048937682062387466, -0.015514221042394638, -0.013790419325232506, -0.009366501122713089, 0.04558160528540611, 0.038228750228881836, 0.00795542448759079, -0.02309590019285679, -0.01620069146156311, 0.024011194705963135, ...
273e09d0adb59f4de8f7ce3e3c0e783bcaabfab6
subsection
45
74
Existence and uniqueness
By (REF ), Cauchy–Schwarz's inequality, and Poincaré's inequality, we obtainK_\alpha \kappa \Vert y_h(w_h)-y_h(s_h)\Vert _{\mathrm {DG}}^2 \le & \big (y_h(w_h)-y_h(s_h),w_h-s_h\big )\\ \le & \Vert y_h(w_h)-y_h(s_h)\Vert _{L^2(\Omega )}\Vert w_h-s_h\Vert _{L^2(\Omega )}\\ \le & C_P^2\Vert y_h(w_h)-y_h(s_h)\Vert _{\mathr...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.06341157108545303, 0.04830631613731384, -0.006656231824308634, -0.05514182522892952, -0.00834603700786829, -0.005042715463787317, 0.004085286054760218, 0.022657889872789383, 0.034818384796381, 0.03683242201805115, -0.029279790818691254, 0.0045277634635567665, 0.014578863978385925, 0.016...
915210a70d94549a9c1d953204602cda17d60b1c
subsection
46
74
Existence and uniqueness
All conditions of the Minty–Browder theorem are satisfied. We conclude that there exists a unique solution w_h^n such that \langle \mathcal {G}(w_h^n),\mathring{\chi }\rangle = 0 for all \mathring{\chi } \in M_h. Recall CHNS:solvability:uniqc and CHNS:solvability:uniqvandp, this implies that \big (y_h(w_h^n),w_h^n,{{v}...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ 0.022076115012168884, -0.010076900012791157, 0.007212497293949127, -0.0382632315158844, 0.00031466473592445254, -0.025554589927196503, 0.009947219863533974, -0.019711362197995186, 0.015004740096628666, 0.008680932223796844, -0.033899880945682526, 0.010267606005072594, 0.008802983909845352, ...
852c0352881ebf7f405a2a20bfd2bb304d7eb161
subsection
47
74
Stability analysis
In this section, we show the discrete solution of (REF ) satisfies the energy dissipation property and we derive stability bounds valid for any chemical energy density \Phi . Analoguously to the energy (REF ) at the continuous level, we define the discrete energy:F_h(c_h,{{v}}_h) = \frac{1}{2}({{v}}_h,{{v}}_h) + \big (...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.016376787796616554, -0.011866684071719646, 0.004784830380231142, -0.05131291598081589, -0.02715219371020794, -0.06715551018714905, 0.03247884660959244, -0.007722883950918913, 0.033180925995111465, 0.02568698301911354, -0.014667374081909657, 0.00018410530174151063, -0.03208202123641968, ...
74fc09853e46253e7135326fd6c0e85b5ff887ff
subsection
48
74
Stability analysis
Then for any mesh size h, time step size \tau , parameter \kappa  , and parameter \mu _\mathrm {s} , the discrete energy (REF ) is non-increasing in time.\forall 1\le n \le N,\quad F_h(c_h^n, {{v}}_h^n) \le F_h(c_h^{n-1}, {{v}}_h^{n-1}).Take \chi =\mu _h^n in (REF ), \varphi =\delta _\tau c_h^n in (), {{\theta }}={{v}}...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ 0.006183673162013292, 0.0003080392489209771, -0.022934142500162125, -0.04998818784952164, -0.03649931401014328, -0.019531404599547386, 0.02403278276324272, 0.020065465942025185, 0.0300295352935791, 0.017593523487448692, -0.02444477379322052, -0.0150071382522583, -0.04190096631646156, -0.00...
e4df9f103ea3757b48a7be89cb6fed10d99df105
subsection
49
74
Stability analysis
\end{split}For the term (\Phi _{+}\,\!^{\prime }(c_h^n)+\Phi _{-}\,\!^{\prime }(c_h^{n-1}),\,\delta _\tau c_h^n) , we utilize Taylor expansions up to the second order. There exist \xi _h and \eta _h between c_h^{n-1} and c_h^n such that\Phi _+\,\!^{\prime }(c_h^n)(c_h^n-c_h^{n-1}) &= \Phi _+(c_h^n) - \Phi _+(c_h^{n-1})...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.03708905726671219, -0.02165817841887474, -0.03678379952907562, 0.0007412087870761752, -0.004552185535430908, 0.010561986826360226, 0.02967124618589878, 0.05082574486732483, 0.03144175186753273, 0.016621392220258713, -0.01953662373125553, 0.017644012346863747, -0.044964756816625595, -0.0...
a54a5544455d36e64241bf5b9efad8a8bae5e159
subsection
50
74
Stability analysis
\end{split}For the terms (\delta _\tau {{v}}_h^n ,{{v}}_h^n) and \kappa a_{\mathcal {D}}(c_h^n,\delta _\tau c_h^n), since the inner product and a_{\mathcal {D}} are both symmetric bilinear forms, we immediately have(\delta _\tau {{v}}_h^n ,{{v}}_h^n) \ge & \frac{1}{2\tau }({{v}}_h^n,{{v}}_h^n) - \frac{1}{2\tau }({{v}}_...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.007072987966239452, 0.014977643266320229, -0.022157451137900352, -0.05743845924735069, -0.020524635910987854, -0.01255131047219038, 0.009896079078316689, 0.03720376268029213, 0.03564724698662758, 0.029695866629481316, -0.014046786352992058, -0.007156917825341225, -0.019929498434066772, ...
75d8dd6f96e4c5c46e0daf209da0855b59624aaf
subsection
51
74
Stability analysis
Then for any mesh size h, time step size \tau , parameter \kappa , and parameter \mu _\mathrm {s}, and for any 1 \le \ell \le N we have\frac{1}{2}\Vert {{v}}_h^\ell \Vert _{L^2(\Omega )}^2 + \big (\Phi (c_h^\ell ),1\big ) + \frac{K_\alpha \kappa }{2}\Vert c_h^\ell \Vert _{\mathrm {DG}}^2 \\ + \tau K_\alpha \sum _{n=1}^...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.05271025747060776, 0.0008350432617589831, -0.020418738946318626, -0.043004490435123444, -0.030414460226893425, -0.015512442216277122, -0.0017673730617389083, 0.010575625114142895, -0.020784994587302208, 0.028186405077576637, -0.019808312878012657, -0.008240746334195137, -0.022921483963727...
2ede56f6036b3bf249a13a34cab0903344f61fa9
subsection
52
74
Stability analysis
In case \Phi is bounded from below by a constant, since the parameters \kappa , \mu _\mathrm {s} and constants K_\alpha , K_{{\varepsilon }} are all positive, it is straightforward to show (REF ) holds.
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.030280187726020813, 0.012424792163074017, -0.025902146473526955, 0.0005267570377327502, -0.021066468209028244, -0.05808914825320244, 0.03368194401264191, 0.022393610328435898, 0.03310227021574974, 0.015079076401889324, -0.05275006964802742, 0.012356147170066833, 0.014751104637980461, 0....
78c7986a130c3e99a0c86f7427054d3d2cb59ebe
subsection
53
74
Error analysis
In this section, we derive an optimal error estimate for the fully discrete scheme (REF ) in terms of time and space discretization parameters. We show that the method (REF ) converges for any general chemical energy density that satisfies Lipschitz continuity constraints on the first order derivative of the convex and...
{ "cite_spans": [] }
1807.02725
Numerical analysis of a discontinuous Galerkin method for Cahn-Hilliard-Navier-Stokes equations
[ "Chen Liu", "Beatrice Riviere" ]
[ "math.NA" ]
2,018
en
Mathematics
[ -0.04382842779159546, 0.013147002086043358, -0.021609002724289894, -0.04416416212916374, -0.008629865944385529, -0.02601931430399418, 0.03870086744427681, -0.02212786301970482, 0.04343165084719658, 0.022646723315119743, -0.021761607378721237, 0.011491227895021439, -0.015138509683310986, 0....