From 57be57e49ce2eee18399dfa373e4aea6ad73ce61 Mon Sep 17 00:00:00 2001 From: =?utf8?q?S=C3=A9bastien=20Miqu=C3=A9e?= Date: Mon, 10 Jan 2011 14:36:26 +0100 Subject: [PATCH] Nouvelle passe. MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Corrections de quelques fautes plus divers ajouts. Modification de l'image neurad_gridif, et ajustement du tableau de résultats. Il reste à mettre en place quelques citations et ajouter des explications dans Neurad et XWCH. --- figures/neurad_gridif.eps | 166 +++++++++++++++++++------------------- figures/neurad_gridif.pdf | Bin 12386 -> 12380 bytes gpc2011.tex | 144 +++++++++++++++++---------------- xwch.tex | 18 ++--- 4 files changed, 166 insertions(+), 162 deletions(-) diff --git a/figures/neurad_gridif.eps b/figures/neurad_gridif.eps index 3d3cf0d..2c51089 100644 --- a/figures/neurad_gridif.eps +++ b/figures/neurad_gridif.eps @@ -1,7 +1,7 @@ %!PS-Adobe-2.0 EPSF-2.0 %%Title: neurad_gridif.fig %%Creator: fig2dev Version 3.2 Patchlevel 5c -%%CreationDate: Fri Nov 19 14:39:29 2010 +%%CreationDate: Mon Dec 6 17:10:30 2010 %%BoundingBox: 0 0 271 277 %Magnification: 1.0000 %%EndComments @@ -606,157 +606,157 @@ n 2355 -360 m 2250 -360 2250 255 105 arcto 4 {pop} repeat cp gs col0 s gr % Polyline gs clippath -2616 715 m 2577 861 l 2634 877 l 2674 731 l 2674 731 l 2614 839 l 2616 715 l cp +4073 441 m 3972 328 l 3927 368 l 4028 481 l 4028 481 l 3971 372 l 4073 441 l cp eoclip -n 2745 360 m - 2610 855 l gs col0 s gr gr +n 4680 1170 m + 3960 360 l gs col0 s gr gr % arrowhead -n 2616 715 m 2614 839 l 2674 731 l 2616 715 l cp gs 0.00 setgray ef gr col0 s +n 4073 441 m 3971 372 l 4028 481 l 4073 441 l cp gs 0.00 setgray ef gr col0 s % Polyline + [60] 0 sd gs clippath -4055 1479 m 4180 1563 l 4214 1513 l 4088 1429 l 4088 1429 l 4172 1521 l 4055 1479 l cp +1320 3284 m 1320 3435 l 1380 3435 l 1380 3284 l 1380 3284 l 1350 3404 l 1320 3284 l cp +1380 2476 m 1380 2325 l 1320 2325 l 1320 2476 l 1320 2476 l 1350 2356 l 1380 2476 l cp eoclip -n 2700 540 m - 4185 1530 l gs col0 s gr gr - +n 1350 2340 m + 1350 3420 l gs col0 s gr gr + [] 0 sd +% arrowhead +n 1380 2476 m 1350 2356 l 1320 2476 l 1380 2476 l cp gs col7 1.00 shd ef gr col0 s % arrowhead -n 4055 1479 m 4172 1521 l 4088 1429 l 4055 1479 l cp gs 0.00 setgray ef gr col0 s +n 1320 3284 m 1350 3404 l 1380 3284 l 1320 3284 l cp gs col7 1.00 shd ef gr col0 s % Polyline + [60] 0 sd gs clippath -3967 1517 m 4089 1608 l 4124 1559 l 4003 1469 l 4003 1469 l 4082 1565 l 3967 1517 l cp +2563 2111 m 2431 2036 l 2402 2088 l 2533 2163 l 2533 2163 l 2444 2078 l 2563 2111 l cp +4681 3378 m 4813 3453 l 4842 3401 l 4711 3326 l 4711 3326 l 4801 3412 l 4681 3378 l cp eoclip -n 2700 540 m - 4095 1575 l gs col0 s gr gr - +n 4815 3420 m + 2430 2070 l gs col0 s gr gr + [] 0 sd +% arrowhead +n 4681 3378 m 4801 3412 l 4711 3326 l 4681 3378 l cp gs col7 1.00 shd ef gr col0 s % arrowhead -n 3967 1517 m 4082 1565 l 4003 1469 l 3967 1517 l cp gs 0.00 setgray ef gr col0 s +n 2563 2111 m 2444 2078 l 2533 2163 l 2563 2111 l cp gs col7 1.00 shd ef gr col0 s % Polyline + [60] 0 sd gs clippath -3837 1553 m 3951 1652 l 3990 1606 l 3876 1508 l 3876 1508 l 3948 1609 l 3837 1553 l cp +2432 2200 m 2412 2051 l 2353 2059 l 2373 2208 l 2373 2208 l 2387 2086 l 2432 2200 l cp +2517 3289 m 2537 3438 l 2596 3430 l 2576 3281 l 2576 3281 l 2563 3404 l 2517 3289 l cp eoclip -n 2700 540 m - 3960 1620 l gs col0 s gr gr - +n 2565 3420 m + 2385 2070 l gs col0 s gr gr + [] 0 sd % arrowhead -n 3837 1553 m 3948 1609 l 3876 1508 l 3837 1553 l cp gs 0.00 setgray ef gr col0 s +n 2517 3289 m 2563 3404 l 2576 3281 l 2517 3289 l cp gs col7 1.00 shd ef gr col0 s +% arrowhead +n 2432 2200 m 2387 2086 l 2373 2208 l 2432 2200 l cp gs col7 1.00 shd ef gr col0 s % Polyline gs clippath -3883 1641 m 3996 1742 l 4036 1697 l 3923 1596 l 3923 1596 l 3993 1699 l 3883 1641 l cp +3969 2389 m 4077 2283 l 4035 2241 l 3927 2346 l 3927 2346 l 4034 2284 l 3969 2389 l cp +2956 3303 m 2848 3409 l 2890 3451 l 2998 3346 l 2998 3346 l 2892 3409 l 2956 3303 l cp eoclip -n 2745 585 m - 4005 1710 l gs col0 s gr gr +n 2880 3420 m + 4046 2273 l gs col0 s gr gr % arrowhead -n 3883 1641 m 3993 1699 l 3923 1596 l 3883 1641 l cp gs 0.00 setgray ef gr col0 s +n 2956 3303 m 2892 3409 l 2998 3346 l 2956 3303 l cp gs 0.00 setgray ef gr col0 s +% arrowhead +n 3969 2389 m 4034 2284 l 3927 2346 l 3969 2389 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -4096 1398 m 4228 1473 l 4257 1421 l 4126 1346 l 4126 1346 l 4216 1432 l 4096 1398 l cp +4002 467 m 3935 333 l 3881 360 l 3949 494 l 3949 494 l 3922 374 l 4002 467 l cp eoclip -n 2655 540 m - 4230 1440 l gs col0 s gr gr +n 4392 1305 m + 3915 360 l gs col0 s gr gr % arrowhead -n 4096 1398 m 4216 1432 l 4126 1346 l 4096 1398 l cp gs 0.00 setgray ef gr col0 s +n 4002 467 m 3922 374 l 3949 494 l 4002 467 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -4073 441 m 3972 328 l 3927 368 l 4028 481 l 4028 481 l 3971 372 l 4073 441 l cp +4055 461 m 3978 331 l 3926 362 l 4003 492 l 4003 492 l 3968 374 l 4055 461 l cp eoclip -n 4680 1170 m +n 4455 1193 m 3960 360 l gs col0 s gr gr % arrowhead -n 4073 441 m 3971 372 l 4028 481 l 4073 441 l cp gs 0.00 setgray ef gr col0 s +n 4055 461 m 3968 374 l 4003 492 l 4055 461 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -4002 467 m 3935 333 l 3881 360 l 3948 494 l 3948 494 l 3922 374 l 4002 467 l cp +3908 472 m 3846 334 l 3791 358 l 3853 496 l 3853 496 l 3832 375 l 3908 472 l cp eoclip -n 4365 1260 m - 3915 360 l gs col0 s gr gr +n 4275 1368 m + 3825 360 l gs col0 s gr gr % arrowhead -n 4002 467 m 3922 374 l 3948 494 l 4002 467 l cp gs 0.00 setgray ef gr col0 s +n 3908 472 m 3832 375 l 3853 496 l 3908 472 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -3908 471 m 3846 333 l 3791 358 l 3854 496 l 3854 496 l 3832 375 l 3908 471 l cp +4054 1470 m 4180 1554 l 4214 1504 l 4087 1420 l 4087 1420 l 4171 1512 l 4054 1470 l cp eoclip -n 4275 1350 m - 3825 360 l gs col0 s gr gr +n 2700 540 m + 4185 1521 l gs col0 s gr gr % arrowhead -n 3908 471 m 3832 375 l 3854 496 l 3908 471 l cp gs 0.00 setgray ef gr col0 s +n 4054 1470 m 4171 1512 l 4087 1420 l 4054 1470 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -4056 460 m 3977 331 l 3926 362 l 4004 492 l 4004 492 l 3968 374 l 4056 460 l cp +3945 1508 m 4067 1599 l 4102 1550 l 3981 1460 l 3981 1460 l 4060 1556 l 3945 1508 l cp eoclip -n 4455 1170 m - 3960 360 l gs col0 s gr gr +n 2700 540 m + 4073 1566 l gs col0 s gr gr % arrowhead -n 4056 460 m 3968 374 l 4004 492 l 4056 460 l cp gs 0.00 setgray ef gr col0 s +n 3945 1508 m 4060 1556 l 3981 1460 l 3945 1508 l cp gs 0.00 setgray ef gr col0 s % Polyline - [60] 0 sd gs clippath -1320 3284 m 1320 3435 l 1380 3435 l 1380 3284 l 1380 3284 l 1350 3404 l 1320 3284 l cp -1380 2476 m 1380 2325 l 1320 2325 l 1320 2476 l 1320 2476 l 1350 2356 l 1380 2476 l cp +4097 1397 m 4227 1473 l 4258 1421 l 4127 1345 l 4127 1345 l 4216 1432 l 4097 1397 l cp eoclip -n 1350 2340 m - 1350 3420 l gs col0 s gr gr - [] 0 sd -% arrowhead -n 1380 2476 m 1350 2356 l 1320 2476 l 1380 2476 l cp gs col7 1.00 shd ef gr col0 s +n 2691 540 m + 4230 1440 l gs col0 s gr gr + % arrowhead -n 1320 3284 m 1350 3404 l 1380 3284 l 1320 3284 l cp gs col7 1.00 shd ef gr col0 s +n 4097 1397 m 4216 1432 l 4127 1345 l 4097 1397 l cp gs 0.00 setgray ef gr col0 s % Polyline - [60] 0 sd gs clippath -2563 2111 m 2431 2036 l 2402 2088 l 2533 2163 l 2533 2163 l 2444 2078 l 2563 2111 l cp -4681 3378 m 4813 3453 l 4842 3401 l 4711 3326 l 4711 3326 l 4801 3412 l 4681 3378 l cp +3917 1592 m 4033 1688 l 4071 1642 l 3955 1546 l 3955 1546 l 4029 1646 l 3917 1592 l cp eoclip -n 4815 3420 m - 2430 2070 l gs col0 s gr gr - [] 0 sd -% arrowhead -n 4681 3378 m 4801 3412 l 4711 3326 l 4681 3378 l cp gs col7 1.00 shd ef gr col0 s +n 2700 540 m + 4041 1656 l gs col0 s gr gr + % arrowhead -n 2563 2111 m 2444 2078 l 2533 2163 l 2563 2111 l cp gs col7 1.00 shd ef gr col0 s +n 3917 1592 m 4029 1646 l 3955 1546 l 3917 1592 l cp gs 0.00 setgray ef gr col0 s % Polyline - [60] 0 sd gs clippath -2432 2200 m 2412 2051 l 2353 2059 l 2373 2208 l 2373 2208 l 2387 2086 l 2432 2200 l cp -2517 3289 m 2537 3438 l 2596 3430 l 2576 3281 l 2576 3281 l 2563 3404 l 2517 3289 l cp +3839 1639 m 3950 1742 l 3991 1698 l 3880 1595 l 3880 1595 l 3948 1699 l 3839 1639 l cp eoclip -n 2565 3420 m - 2385 2070 l gs col0 s gr gr - [] 0 sd -% arrowhead -n 2517 3289 m 2563 3404 l 2576 3281 l 2517 3289 l cp gs col7 1.00 shd ef gr col0 s +n 2693 536 m + 3960 1710 l gs col0 s gr gr + % arrowhead -n 2432 2200 m 2387 2086 l 2373 2208 l 2432 2200 l cp gs col7 1.00 shd ef gr col0 s +n 3839 1639 m 3948 1699 l 3880 1595 l 3839 1639 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -3855 2196 m 3947 2076 l 3900 2039 l 3808 2159 l 3808 2159 l 3905 2083 l 3855 2196 l cp -2939 3293 m 2847 3413 l 2894 3450 l 2986 3330 l 2986 3330 l 2890 3407 l 2939 3293 l cp +2616 720 m 2577 866 l 2635 882 l 2674 735 l 2674 735 l 2614 844 l 2616 720 l cp eoclip -n 2880 3420 m - 3915 2070 l gs col0 s gr gr +n 2745 360 m + 2610 860 l gs col0 s gr gr % arrowhead -n 2939 3293 m 2890 3407 l 2986 3330 l 2939 3293 l cp gs 0.00 setgray ef gr col0 s -% arrowhead -n 3855 2196 m 3905 2083 l 3808 2159 l 3855 2196 l cp gs 0.00 setgray ef gr col0 s +n 2616 720 m 2614 844 l 2674 735 l 2616 720 l cp gs 0.00 setgray ef gr col0 s % Polyline gs clippath -4620 2142 m 4568 2000 l 4511 2021 l 4564 2162 l 4564 2162 l 4551 2040 l 4620 2142 l cp -4964 3257 m 5016 3399 l 5073 3378 l 5020 3237 l 5020 3237 l 5034 3360 l 4964 3257 l cp +4618 2154 m 4568 2012 l 4511 2032 l 4562 2175 l 4562 2175 l 4550 2052 l 4618 2154 l cp +4966 3306 m 5016 3448 l 5073 3428 l 5022 3285 l 5022 3285 l 5035 3409 l 4966 3306 l cp eoclip -n 5040 3375 m - 4545 2025 l gs col0 s gr gr +n 5040 3424 m + 4545 2037 l gs col0 s gr gr % arrowhead -n 4964 3257 m 5034 3360 l 5020 3237 l 4964 3257 l cp gs 0.00 setgray ef gr col0 s +n 4966 3306 m 5035 3409 l 5022 3285 l 4966 3306 l cp gs 0.00 setgray ef gr col0 s % arrowhead -n 4620 2142 m 4551 2040 l 4564 2162 l 4620 2142 l cp gs 0.00 setgray ef gr col0 s +n 4618 2154 m 4550 2052 l 4562 2175 l 4618 2154 l cp gs 0.00 setgray ef gr col0 s /Times-Roman ff 317.50 scf sf 1327 3825 m gs 1 -1 sc (W) col0 sh gr diff --git a/figures/neurad_gridif.pdf b/figures/neurad_gridif.pdf index 8a47bb3beea974307e3dc7b8ae74038a74143425..ab2ac2530cae9d7600fc8b12c24e86bcd496d205 100644 GIT binary patch delta 6728 zcma)=RZtw<+Gc~hJEWl-ce-(h#@#}2Z5)C-0UBwX5F`)?8X!0c?oQ*u-Q6t^AOr~c zznQ6;srt^voPD#OTD5l7Q}27XcdXW|l2za%NiT;;I)547r11B6U5w4+q2F5>F`_Uq z1Mu}Ic{EHZU3-6QN^B&oRaC^>y$E6SDc~&Ow6WypuY-4e8n+;LcKO@YZXWu_x9j=G zr24uX`_;o`LThMX?X>%JTO|SL4Xc?uYOU!A1#x{=z(D=gwfhgxvA~cf)7VA-k2tk( zPw!*rwAH#ak;oL0r}=}rFV5c<78&8md>dvtJlT_!mG|Rl-|C&z2`gLLMi#Xw$0l~g zu%V~*y%m?dWUxNgzv3U8spE#i_^A-h%s2O`%_bD?S9Gf0lt8HI8KujMD?5R^llW!+ zg$DY+Ny<#HdfAm1UAXHPCl#xVlsBz#w}`Q9#vKd-G4~BpHrtmKXy0AGj)`LFWQ^MH zZ#p)SqULpuw#W>%1!~jy$Aayo>~i=Q$eo~f$(|emTBhY_e7>j7D~YDQcy$>3f`d~4!dYDW3**r%%3R-8fmz10-(@ER3d!rN+R%)D0a2?3u(OdL(OwUamvH7C4e2D=wu{$)q=j#E#1$wWHt*zt?zd<+wi_78zB=MjhxOKL~JlY>D-#lqNd?Nf=$^ zk^(AqUK3k4>XekJp^jdk&HZLmm?OjMP`-15QMHyEJ0W(iPd|K3lG9}!s1oW5)6ERi>Qwt?8rbRfqXcwcO*QrHkZi z;KotF_<;Ei(4c59n4%UHW>sAamT|Dh1Vy7z1-{C;86IjMgJWeTvk#-zq7y6kQz*0| zl{%PbW!bD2u7`8V?X(K;e5e+^nwIF-`c?AKq_0Bv#2lUs1g=c(qk}<7Wi;y$C5ddI z6RdYSH34m?fu-I(Y@{F>gCnFUlXrcwD74i*$Mi>c$?_K zT0GT0EcQnhwSou%-VH2~4Th zM9Cs$psr2qAKe&$1A}F$*rd0rl9pUub6+vpRa{RPSkSeVRo{22}SZLC>k{4x@NN&7m zo`}BfIjCX~q@L)LZVih|`D*aS*gmm=Hv4TN@~*!M4j~;`x~E~HDZWljvEWK(Q3`5N z$$`uy%qcEn@g_m^j>_!br4&m|P~#L)5&@v3rY|oSfX8Z35|PN_=`G2KV!IDU5zK%0 z?2~Q7^X-#wOgy-uq{E1F5@yi1lW)q>cXzQtl`Zo?8rD&;N~Ws$Oj8=H=VGRbiEI{V zZZ!i3U>-9=Tg+FKMzF{I680!-wGwPUa8xR`_3^kjtFTEPW;M}~y`3e7^Y={-d@AEohQv$rti?@QGaRu*_F+kGIZPEAQ?3x|2s$(80>pcQ|oVM zJ)ssA0l2p!eC%idV^t3VykVpf&R~au&M{JDTDln79m&%Zp=4QsP#k8`)z-?Ra0N8Z z4fC`#EHIp~^WpA#+_)7E5Asc4(82? zb}7GaR%;GR2R&-&j?Nq!usBw=6BN>JIW-dX5B{+r z!h8+dpuNUhsy|2vpwYZ_w$OuApfKI&4W?x!O{y>0+mr_iF8D1n@TC@c)t%K!d}3b- zTY_aCP?p6Jdx2S+9G`%Zl^~ZI?!n!HY!|O1911j0$pi<`$dYVG*zw`r;2IBZmn%_)j)7RB^$`s9FmZ^`Ja>XjR; zv!#rF06*cmNVBudsCMiz-00Y?QZe~;^RW8P`2OmFSuAEDf3PWOhT z)1KpTC)d>8QW_Ihq?mk?>0^1f?*E6;f|b_F4Zbd1os7Y_Ls>f2<-9MR<-qs7HcOb( z%X=ZI>JiD?&UTG+Y}TtArxvU|APsC3`Z!&{Bgh;1M8iVpb6_h+r}7|))Z$X&I>f63$sLf7 zOzXlchR57%Sx?hvqrg35@Xio`sB&lCxM3D-gpwT9PfZ28v@8>;CSaJ+xL#QC8gSu( zB7%CmgVna1%z1K(WUWBeQAd&0dF<3_2DO&<>d??*F_Q|rBD%0ttmqj7VH$@?GKM5X z!5J!G=0Mm|KEDW{{Z^IfByAFmtx)KS=%&Mgi{I=6Im&>AbhR!IUV^&W=wW3c>M8|b z@t)>Ezz z5mT;o3TpP&%O@P{8NC?;UhgkIaZm-OvBrk-krkdPlT$cTip!F~?K$NcqO>9iFm}LR z*r#)-uW%(|iD^G%Jf|@CZTS(`8Xm|LcokC}))tup( zy|i|eF72ZeUdA9jD7oKFTLY(>rk@MEV_NjpmnNI}iC1rMS8Gu%tuN7|%qW}hPwHO;-nd#1=*17d>Y{pg3&`NHkFDzHED9(yqG^ zjeJ7c)VXZ1pFp)T4IkIDJ+uVXAPP%xV@n&6U`%ZCu!X>hQ)T(m&lLJo<=cJm+*v)d z>{rCtA~~3*IpQr&%X6*D%e7y9$>s?4R^;$}zZ(}N30A)`FCbmRFJMm$?X z5X2*wfy*9`@%8WGGv~_yza5>iIJwOY-fz!0SMJT1GS5Hq`8J*de*ak=gBwy!*FUjk z0-STLlzT5ou*N9lZuHhdo}RB4EziYaGEMH?&)0wN4et6y{a_=Nmur-FKxeSvf_l0eL;QP)O1$Wy4zDG&^*sRQp+7NE zJgQ^i;bK!m(QA0ENbY1(E0GH`%^DzbAdRzS=Vc%pS`@dtYHdt713>&zw?vxm!}H;) zyM6kSMH>(CSoGIw4(HYSqfW*g=;Lqh)bEPB{%|H6FUTatVEs-H6(Du_fu8GVJU+b8 zvVmHV-72*#pCL=9@L6E#1BJ`-mUR_u>mT)R8|H!yBdKGg4-s01@T58?m&+I6qxiUN7cM##3;VL22!@ z$z~iGT!c3;7w)RApN$zK#+G9Ph^YhkJD-UDB5L9E31msbUMt4{`sEdg{l38SP_M|m zr6JDjR9oDkM$cfYyl^X;jo-Z-(pVUTWns^g-EGh2@vf%H69|D#-emin8=i<@@84?o z_uFD#L;(|X00)G_C}SN++C#Cl?YMz+J+Ng3 z70L=kj~6OJFYM^ZlZm`_EzVOG@duKw+`^SgU9cLI^=0|sRidAVbn@rbuUlxt18vNn z5_4*x?~N)@&8jC2lm{f0VW>?fNIxP*LhjF`{ysa!Yd7w>YBu;}AE-_~!JHF@=L~?gu7YDd%!3n&Mtvc6C$w02{TXBh8IaJ$Mci zFXoqhBuzPE(L(Pn?*!Ib_@19W*Q!%)sgxC?AC4S!qi+o>=`Ugzkrj%cjz!-z@#^*7 zDe(<(!8-u&XHXJWu|H6gv;Y13vOIwyJyM6iC1Unxb5#@$RMD_AZy>k#$ChRiPxDuJCwO1*#b?#iVAuHBp&B5BT1Ve7HwLe&@j^vf z`<2wljzZo0K@dH(x|H~qj-U#8Hd!1eLv)djrrtP z*4$<;Pky!|^OAFL#0tMuG1SkE`N?Uycfo?#j(M5ABSyEeUVRdIs{xnz*1bL@5+Euz zFr^-XnR@FV!Fr6PpHSv0LjW(W?BIX+u5`rvMVZ`zUrVJGC+NE#pMdjZ%kfJG+Hir_ zguP;kV7AVJux)jGmdt`e?evNRK8CeG?l=9{E#l3O_!7(lW%I}@NVeYe+XK~6?tJaR zVAgBIt}GIVh37ZQX~$IJtp|ukD_|En*6`K$Z1-L?xySrN9|QZ&nG_slq`BS5|3$it zg{{qnp)O%;G=j@se)?f>%0v(KW)1CC(Yz*?G-(jlP+TIDp3shR)eS%0LJ8h`iPcMu zfajN}tp2YnGl`HExMU+ah+%1?~GnaZ2xRus_+4kW9lx~1rVw3YEwU< z8Oo|-GNAowlM$39wG03HnLFmQ;)BaAK4ptk*&^4BaMwS%52LyD}%NM=R4tmJISzuAHsbd7swemF_aFXIOEGVSEHA+IPR!-*9B5#A_t8_;05Yk{3Gad|a z-;i=hJjs1}<}NIK{EiZb%m)O`Oe<&^|1$PYit#!E8335Ov95AP4lnUNo}R+umppeD ztcpry`Zc!dwS1tIY5kg zmdx*i9Id8++%4ic0oZ!cEn-(=2#4!Xk)fXaP+$^Q67RVxCz4aNJDrCV1L z@5C{_bqgiff6usyQDRYi`yTSk-Qv&hghj%3#oep!THplh@6F4N=j{kLzx~gE(4#@~ z&93f_huvTB?%=1fw;BUvc)TI6fq^{6BfE>_7Q4TSJvaZX5+x+({#60u=cTIR}kk1L%@*#YorR{=e#A z0Re%uVLnL2mDh!2)h>3wk!4M&N5vYi?YyOZGBKyYic36LNOjk`Mpx8UyX7CeyP7ThgJ2gu}p z-&D=idw|JZ^Q*~C&x6R{~;VN;=BW~S2W|y&Y$eg>~Cd^c5A9>%rL!x1& ziKZVSnB~f}A_)D{_47Bwk8&Y7!PPIwcmt8u*rTXT zg76z9sVWg?V0sqns~Y!*@B}X$R(Bb%*RpXu52bpSr0{qFJ7mJfA=)|&d^23D^Xha) zE4Aaxb1wk~Fm}<)&5D`Rv+8z6OSnw#F@0+%8-wXg{Ph->o9uADXz%J)d$c=}O&PWR zbIcwq{pM0d^l$gdu)AcpFif5AI>T&MY6+y)cob=Y`yR=n7uqe=B?GN`EkI+SoVz+x z@&MvQZVrjNSoxAb6a0=N{3(Y||2JfKc9C(g2tfi&|qfzDLzMc_UXW43HbfP8o* zKxDg0R!=5)K{lg+9l-jvL1#Y~U+cA#tYV&<%)GYVXf5wnWsb-3;LFJpW4mn&vNci0 zNSTuwgbq)KeytsNIWljjHA~L@7XKRjVshmY6J{M#(W(vYDjl zc7o6r)>j~9WvwWzjBf5hk|to#0RgH&gE)3tn=vuHB()_mZ<%@GFiHf#K^yiMDCKHj z6%2$}gRD6Mm2D{D)hYN?-$(X-6A@x_`jXeDNQGa7(Z8NodbCJp@7GpfsHI3&NtR7S zsv0%YTTk6BbX}6iS&CF$FlJ&A+3Sjrvjh-a_FprFQ|@)44FOCrIhDccKcgNMtgJ3V zQkerv{pa%Pf|^I8S+vXW@hb&M_x2|yfP=i746cz$IS*d&559Orgy;}pD$=%nYanad zVtAT0zz04tl^&e|oF_!a91mSlccg%X0gct5W14EEYsF&^p7^FW-E zL?a@?M!h)r^)?`+l0FQ$1Zd*Lvj!(_y_ObzcV&SbX&#NC>LtcO1Q4wc=BPGc*Ph0- zlf~+a%KC=Kl!uVU56Eq`uLL0Ut93aUJ0nmR-wY6H#7_suTBdJvqBa)^ zI%v&VHQk`O*%?Nq0a;Qb1oSgLGg(mbOcP3Je?Nj3Cu$RWV*0Ye*z1E^3-6485q(HH z#Y0%wfbHux^f}-g<(N(8y?L~*sGXtEDJNF<0rjo5@72$HU5kFEk7|La95H9Ey+mL* zDwi9AYBACqLB;20vNjoICPTBf2E-M%{~$!{2(CH9;MU6Y_z)Cs(fJCjJ^|u#>O)hc zP6M0wQ@5ZI6=f&t?`@J;mxJ~h<)< z{J;N}zTa67OTQB^YZa#mD*G^5N(B{#Ri*+}qSPGYZ`cA8Xqmc*7YM(leK72hlnp&}QpXdqLw@OqWZTuLB=3x-NzpEsJySJpZ-mFKZZ z;24V3B=E3l(=d=J9L<6H*mF=%CM>CdTm}fgRr3~Zqui-j zsXcR4pHS`2><%qzYNfoYSTj+x3_TeOtALAWtcO*|*seZ#8vHan@em|^a%<20su$I^ z<$z1DqG90P{zJswZTls|UZ~H7ViTuZa7c6Z1mr%%I71E#(~mEVDd1uuV_`U=}q(w-ebo zDftC2T)i{j*|G(wGQ=5DgdUS57%{}UMY_&(+Dre!^SfXX?k+iZR+ z&EjZ&UCC#i%ckm`8Bf_}?3-yuQfI4zhK<$**t$>|qQGaAS1SPMu1yO8D8czchMGUn zy?N-(3tSDb#d>=U1eU+!EdMqlvI~{XHE1WKB#P@6*O@s?3{GnkdSjm@zA+Q8#;Yxm z)Q=en_m-8tVlDs0l`l!={M<3ol6g`4P(}P(S3YP{ECC-yb;j^cJlw(m%Yq~h)m;ZI zPPuHS;0{@>-6V$?2H)*`XOX2Ys^UR$^ecXv+_cKcVeM=C%F{EcH72E9>m{oX0$Pp8 zbwj+9T$)5w0uKdFN%fIyw5uBi@A;5Q!(PP@!qFK{i~>`?Gv+HUun3|dV}_y(Y*xSW z2i}<9YuP)uYI?N;uhE@g?u;1KDeTNZGyZJd2y zD5-mqA4EPnwC>u5AS7gKC`BWR7Wg?nLdmUGIvJ{e{khohH@3ox54vh%o@xRm znm6Pj`?ZbF*2`?=>6V+KZ9~%lwlz7XBT!EPw7YS5@h z7N0TeNAQVI{w)qo-^Dc7l!)9wnkC6{PUxmS8lj zLe!6rZzfzf^;qZpBixw!Yh_JgO2OnO&MauQ#N+*eGbIm8%8z-GI(J(51VVDBo+rAw zQIz3d=G?X+LOKM?P;^4Y#w^M+A>cr}cc-kchJ9LPYE-(}d$A_gwe+)Ekv;B3(117* zT(-p-pDiQArs7MUA%-#OSBY-N5|bUZOdz7`O?W+>mToEJ07>C7qYUR0u_ypc96m5g zU#WxrhRq3|UVxSBiBToE&l$hYU{5OVT@l%S9mqL;aYZyc>u4E@07A-@_ImpCQ=})% zWB8()-rCr;RpOTcvBZ4d8j%#UbO^4)?X$|XSLz)0%lI4(qg#r~vtHn+4}fFE*cdv zMmECEwKeYvKn5t%B8$04(*tiTXS2Z*){cNT9SNO;cLt$3qCm&PK$~AhswQE(YP%6{ z7l!pOshpP_Jy4%k9b2<-ZQ!Gh#bG%X`iT7Fm*$>$1Cq9D@?n=8B(2W-_Jp+pS-^6# z9vKtI#$PohE$>HLs(Nm=Mpk&V{jUG4NS)`L!@YceKMBc(bLuc_bUhtU0!u8YZi+D> z?0KxDt@)7p?_`ErGQR8&QwHxuOJLQQtHrw#(NNgHgeyumZdsp2+=r^vf znpo~^hEj0uOBC;bFWXijmzBaZ8Ls$FZL=Nc9P%Ui1p$Q;U&*o~pB2B<*TA-a((GJ4 z@ZI?t?$2D07qbaVw7J90i73e|@y;wY9R;wqz0i;&d_yi#j1Rt|h-aBx@PsDWIOlk2 zV8-f{*b{-7Eb7;C$YyVa+BcU`erka@6DPIDw>a~8;vKhE4V7kd$i{AkFk+lTh{!sY zKH64Zl2?xekFBt&ihnX{`)b_W9+`?Crezt$L%4LiDbPPFb=eL#S1jMI4eGw>l!lpv zwMlv~M&^uN`#03ne9NVOUHE+!DDSk7bq#Ur>`+s-IsZ^O^OXNHo%UN9Irl=D6Gzs) z<^aMeV`Y05GCPgZ=k;>`J%b(c4A&-paE{sE|C?5%qHL|@8HG}zYyzE|n6o5FeBX-n zkR^0$nD`bX@}?QSa?P-eK@s@#*;V{r@)JLBG2l1b4<_b(HX<2c)1BoF@|z|Sb?Otv zYl^yGelfH`H4STj)0F*9lh@+hWHby52`0@Vv$u~(Q@&ZLvDy|sP_ghm^bQc7ni9KV zzJ-rq!O#Ra#A0u8eu3F3t+mTI zvK}-G|CNc6fTRls#m1u0eK=)60XM9(;o0-d_sg6=<6yyXAuk3YG$qFznCL01xLBm$ zpubib-csXQ_;p+_%pGoi)HRd?KgDo&H1~Va>g@y%j$V#+qh#_e>U&b3dMCI9y=nIh z;iv0>AAWp4(PhRsws}U)nWNiBvOoG6hq%6zd59Ln&*Pv|DFZQ^ck|g5E>bx#+y^@?2p%5h#gu)Vm zc*p=Wl}{%%a8c4bZ#^>nd^_>Ust85yh}@$X?x*E~9l0A*+XI{p%WjY3apPmG+n^Si zgL0{|g+jqu3aS(@Y1O@JZ7QZW=dN2}omv&^VfGx+Gy~@Dbrbr0+=y!*W}DIK-BbEi zNa_2-JE8XX2&^Y>;&P%uY@o{#-m6i2<@bLx&zc8uu6S3&_N;`areoXgqK0tx)LGy+ zjV@FM04K#>H;P6UP-X_`B&aWjf9EjOnRM>TzewBb$CwQc<={hfwtIiGr!)*6Bry)y zQn~n4c?iPTI!AT%4drC~@*sw$L()kg2qfHFX{@Lg$tB^?)(jFD&tWN=mHh71`Lgm9 zvilSzkCxUza0%qSu=aVNdj?oZRKdMDyepfnk|;e!3sYAq?^}fftT0b%gcCIxmzVF^ z^{Qg&q>VWA*?m_;=fv2}|9m??gZDtJW_W33eR9-{2ZQq#?tb)-#2LJ!%k!2V2h@43 z9CO|fszlKinhqb`ojM6tWo^pydc>#_$jP?Fh`aSqTu3@CWCSV2kB7Dn<-<*!SsL;w z3#x^i<@5rtgNvB#RU*GZ$vzj+#UqeLQeye_b!lrF=cAailUy!gS=9zX0m%D#v-Ju{ zIEm9U)Zz5UDgg+R=Dd=Zuf_e7^Vq~4YOW6hkBJJnXAe$+8zq{oM@o&RFcUU0tB_JN zXc<|AsD%UWEm0MQ6+lU^23{X$(GZ<2X1(FTmTwjzUDLZ9CvA;7`yf`EvA6_*A?1~WqFe7_q#;H#8}j0SjD(9vVIim=U_0=W~Rt7HnDY5g+kD; z2{2Q)(5<3z*muRs>EX6vueV@w={3(rA)!%_rm|$ZM4x`Q;*;o5fJ>6w!Shm0aG^%zWFN_`^0rpv%f3t`OmtK4@(U~W287}i4CXR`0f7`ci z^Ax^_oyb9N*BDJz(=h+l`c&lAF7H)=Q^|OKj>31qfVHr7N++BO&>KJ*v1xNQ65qc^ zh`06b^W3`TPc*K&MHi z90(UAC0H7>e{2|d&B@ykXz`|NOQDP*j3vLRs+W>h4Aaza#M0sY_wy~kJ}@vmQV1NT zS7#D;*IFIdX?UB_pB5@1trO0P8TvphM&#yC3nP6&kHxVd4HxYb_anp- z$y|>QCUj9GsbZyqs5*b={4D)N>f3QMdEpHnvvFEucz3yb;7X}(U(2!lBq~ewXbqa_ zXaK?t%GO2ep?|DEPnMD`h?Zz-k&$NjlRBIqPF>~GO=1%Kjq0oO`peHa->R<~+PPc# z4p!F8rqU(+4Zr#M3^%FF@w4ZpNxfqXSW(i4Phx*)bpsLhXb!XvYw~Pi^dP5JtwS%f z*6>)#B2Ee9v5=_F@3j(EUHVU2C2bR;!P*&~zk`&A8|~+4UT|I^pW?wx4@GtQ__u#R z;8>aPyx9u&>Z5j6rBR=%VpHpQ4F1n2eeP-g$I*)LccpQSdk0O9U`4A07CuqOdm1cU zJ}Lhlu#SUH$Nh(X`4ht6z|Ih4&j~iH5kOj;+iyV_yGJa%ppHCHJ=$lTuWyK9L}ltl zJWI;fh@CDS{|G||zlnC{UbW?8&qtGukjQhNTl z@}?g5&F(q*)s!c07ezkNQ5Do>R&?)^OXC4RC87h_v0;|&;6?n2xAC0l)8v;g*xcJ> z0x}6{N8=8K%?TtS z!_D})M_V%fH8F^~25rObVWO7q5srkvm@_Q6bMQ-XGZuP=3A&nGZEX_&?){CPzE6#mx0lwM%AP@SR#R_4jDevYtu@?-npSGD^ydYlutC^> zRQGX6>(VE}+cnp9qwC=tnVIOxbIiz};lST=D$jR3u76HL9qO)baC^Ldp1HX8^mIMk z(fK`HSXX~$gcTGe#s5#@EtaGtmb-_=I}fjSHnv!QS18Db#muar Hp@{Wgu`f95 diff --git a/gpc2011.tex b/gpc2011.tex index eaf7101..83a04b4 100644 --- a/gpc2011.tex +++ b/gpc2011.tex @@ -46,21 +46,21 @@ \title{Gridification of a Radiotherapy Dose Computation Application with the XtremWeb-CH Environment} -\author{Nabil Abdennhader\inst{1} \and Mohamed Ben Belgacem{1} \and Raphaël Couturier\inst{2} \and +\author{Nabil Abdennhader\inst{1} \and Mohamed Ben Belgacem\inst{1} \and Raphaël Couturier\inst{2} \and David Laiymani\inst{2} \and Sébastien Miquée\inst{2} \and Marko Niinimaki\inst{1} \and Marc Sauget\inst{2}} \institute{ University of Applied Sciences Western Switzerland, hepia Geneva, Switzerland \\ -\email{nabil.abdennadher@hesge.ch, mohamed.benbelgacem@unige.ch, markopekka.niinimaeki@hesge.ch} +\email{nabil.abdennadher@hesge.ch,mohamed.benbelgacem@unige.ch,markopekka.niinimaeki@hesge.ch} \and Laboratoire d'Informatique de l'universit\'{e} de Franche-Comt\'{e} \\ IUT Belfort-Montbéliard, Rue Engel Gros, 90016 Belfort - France \\ -\email{raphael.couturier, david.laiymani, sebastien.miquee@univ-fcomte.fr} +\email{\{raphael.couturier,david.laiymani,sebastien.miquee\}@univ-fcomte.fr} \and FEMTO-ST, ENISYS/IRMA, F-25210 Montb\'{e}liard , FRANCE\\ -\email{marc.sauget@femtost.fr} +\email{marc.sauget@femto-st.fr} } @@ -70,11 +70,12 @@ Laboratoire d'Informatique de l'universit\'{e} This paper presents the design and the evaluation of the gridification of a radiotherapy dose computation application. Due to the inherent characteristics of the application and its execution, - we choose the architectural context of global (or volunteer) computing. - For this, we used the XtremWeb-CH environement. Experiments were - conducted on a real global computing testbed and show good speed-ups - and very acceptable platform overhead letting XtremWeb-CH a good candidate -for deploying parallel applications over a global computing environment. + we choose the architectural context of global (or volunteer) + computing. For this, we used the XtremWeb-CH + environment. Experiments were conducted on a real global computing + testbed and show good speed-ups and very acceptable platform + overhead letting XtremWeb-CH be a good candidate for deploying + parallel applications over a global computing environment. \end{abstract} @@ -94,14 +95,14 @@ techniques, using analytic methods, models and databases, are rapid but lack precision. Enhanced precision can be achieved by using calculation codes based, for example, on Monte Carlo methods. The main drawback of these methods is their computation times which can be -rapidly be huge. In [] the authors proposed a novel approach, called +rapidly huge. In \cite{} the authors proposed a novel approach, called Neurad, using neural networks. This approach is based on the collaboration of computation codes and multi-layer neural networks used as universal approximators. It provides a fast and accurate evaluation of radiation doses in any given environment for given irradiation parameters. As the learning step is often very time -consuming, in \cite{bcvsv08:ip} the authors proposed a parallel -algorithm that enable to decompose the learning domain into +consuming, in \cite{} the authors proposed a parallel +algorithm that enables to decompose the learning domain into subdomains. The decomposition has the advantage to significantly reduce the complexity of the target functions to approximate. @@ -113,15 +114,15 @@ approach. Here, the computing power is obtained by aggregating unused (or volunteer) public resources connected to the Internet. For our case, we can imagine for example, that a part of the architecture will be composed of some of the different computers of the hospital. This -approach present the advantage to be clearly cheaper than a more +approach presents the advantage to be clearly cheaper than a more dedicated approach like the use of supercomputers or clusters. The aim of this paper is to propose and evaluate a gridification of the Neurad application (more precisely, of the most time consuming part, the learning step) using a Global Computing approach. For this, -we focus on the XtremWeb-CH environment []. We choose this environment +we focus on the XtremWeb-CH environment\cite{}. We choose this environment because it tackles the centralized aspect of other global computing -environments such as XtremWeb [] or Seti []. It tends to a +environments such as XtremWeb\cite{} or Seti\cite{}. It tends to a peer-to-peer approach by distributing some components of the architecture. For instance, the computing nodes are allowed to directly communicate. Experiments were conducted on a real Global @@ -130,7 +131,7 @@ interesting speed-up and show that the overhead induced by the use of XtremWeb-CH is very acceptable. The paper is organized as follows. In Section 2 we present the Neurad -application and particularly it most time consuming part i.e. the +application and particularly its most time consuming part, i.e. the learning step. Section 3 details the XtremWeb-CH environment and Section 4 exposes the gridification of the Neurad application. Experimental results are presented in Section 5 and we @@ -148,26 +149,26 @@ end in Section 6 by some concluding remarks and perspectives. The \emph{Neurad}~\cite{Neurad} project presented in this paper takes place in a multi-disciplinary project, involving medical physicists and computer scientists whose goal is to enhance the treatment -planning of cancerous tumors by external radiotherapy. In our -previous works~\cite{RADIO09,ICANN10,NIMB2008}, we have proposed an -original approach to solve scientific problems whose accurate modeling -and/or analytical description are difficult. That method is based on -the collaboration of computational codes and neural networks used as +planning of cancerous tumors by external radiotherapy. In our previous +works~\cite{RADIO09,ICANN10,NIMB2008}, we have proposed an original +approach to solve scientific problems whose accurate modeling and/or +analytical description are difficult. That method is based on the +collaboration of computational codes and neural networks used as universal interpolator. Thanks to that method, the \emph{Neurad} software provides a fast and accurate evaluation of radiation doses in any given environment (possibly inhomogeneous) for given irradiation parameters. We have shown in a previous work (\cite{AES2009}) the interest to use a distributed algorithm for the neural network -learning. We use a classical RPROP algorithm with a HPU topology to do -the training of our neural network. +learning. We use a classical RPROP (DEFINITION)algorithm with a HPU +topology to do the training of our neural network. Figure~\ref{f_neurad} presents the {\it{Neurad}} scheme. Three parts are clearly independent: the initial data production, the learning process and the dose deposit evaluation. The first step, the data -production, is outside the {\it{Neurad}} project. They are many +production, is outside of the {\it{Neurad}} project. They are many solutions to obtain data about the radiotherapy treatments like the measure or the simulation. The only essential criterion is that the -result must be obtain in a homogeneous environment. +result must be obtained in an homogeneous environment. % We have chosen to % use only a Monte Carlo simulation because this kind of tool is the @@ -187,30 +188,30 @@ result must be obtain in a homogeneous environment. % \end{figure} The secondary stage of the {\it{Neurad}} project is the learning step -and this is the most time consuming step. This step is off-line but it -is important to reduce the time used for the learning process to keep -a workable tool. Indeed, if the learning time is too huge (for the -moment, this time could reach one week for a limited domain), this -process should not be launched at any time, but only when a major -modification occurs in the environment, like a change of context for -instance. However, it is interesting to update the knowledge of the -neural network, by using the learning process, when the domain evolves -(evolution in material used for the prosthesis or evolution on the -beam (size, shape or energy)). The learning time is related to the -volume of data who could be very important in a real medical context. -A work has been done to reduce this learning time with the -parallelization of the learning process by using a partitioning method -of the global dataset. The goal of this method is to train many neural -networks on sub-domains of the global dataset. After this training, -the use of these neural networks all together allows to obtain a -response for the global domain of study. +and this is the most time consuming step. This step is performed +off-line but it is important to reduce the time used for the learning +process to keep a workable tool. Indeed, if the learning time is too +huge (for the moment, this time could reach one week for a limited +domain), this process should not be launched at any time, but only +when a major modification occurs in the environment, like a change of +context for instance. However, it is interesting to update the +knowledge of the neural network, by using the learning process, when +the domain evolves (evolution in material used for the prosthesis or +evolution on the beam (size, shape or energy)). The learning time is +related to the volume of data who could be very important in a real +medical context. A work has been done to reduce this learning time +with the parallelization of the learning process by using a +partitioning method of the global dataset. The goal of this method is +to train many neural networks on sub-domains of the global +dataset. After this training, the use of these neural networks all +together allows to obtain a response for the global domain of study. \begin{figure}[h] \centering \includegraphics[width=0.5\columnwidth]{figures/overlap.pdf} \caption{Overlapping for a sub-network in a two-dimensional domain with ratio - $\alpha$.} + $\alpha$} \label{fig:overlap} \end{figure} @@ -252,27 +253,27 @@ The second step of the application, and the most time consuming, is the learning itself. This is the one which has been parallelized, using the XWCH environment. As exposed in the section 2, the parallelization relies on a partitionning of the global -dataset. Following this partitionning all learning tasks execute in -parallel independently with their own local data part, with no -communication, following the fork-join model. Clearly, this +dataset. Following this partitionning all learning tasks are executed +in parallel independently with their own local data part, with no +communication, following the fork/join model. Clearly, this computation fits well with the model of the chosen middleware. The execution scheme is then the following (see Figure \ref{fig:neurad_grid}): \begin{enumerate} -\item we first send the learning application and its data to the +\item We first send the learning application and its data to the middleware (more precisely on warehouses (DW)) and create the - computation module, -\item when a worker (W) is ready to compute, it requests a task to - execute to the coordinator (Coord.), + computation module; +\item When a worker (W) is ready to compute, it requests a task to + execute to the coordinator (Coord.); \item The coordinator assigns the worker a task. This last one retrieves the application and its assigned data and so can start the computation. -\item At the end of the learning process, the worker sends the result,, to a warehouse. +\item At the end of the learning process, the worker sends the result to a warehouse. \end{enumerate} The last step of the application is to retrieve these results (some weighted neural networks) and exploit them through a dose distribution -process. +process. This latter step is out of the scope of this paper. \begin{figure}[ht] @@ -285,7 +286,7 @@ process. \section{Experimental results} \label{sec:neurad_xp} -The aim of this section is to describe and analyse the experimental +The aim of this section is to describe and analyze the experimental results we have obtained with the parallel Neurad version previously described. Our goal was to carry out this application with real input data and on a real global computing testbed. @@ -294,17 +295,17 @@ data and on a real global computing testbed. \label{sec:neurad_cond} The size of the input data is about 2.4Gb. In order to avoid that data -noise appears and disturb the learning process, these data can be -divided into 25 part, at most. This generates input data parts of +noise appears and disturbs the learning process, these data can be +divided into, at most, 25 parts. This generates input data parts of about 15Mb (in a compressed format). The output data, which are retrieved after the process, are about 30Kb for each part. Unfortunately, the data decomposition limitation does not allow us to use more than 25 computers (XWCH workers). Nevertheless, we used two -distincts deployments of XWCH: +distinct deployments of XWCH: \begin{enumerate} -\item In the first one, called ``ditributed XWCH'' in the following, - the XWCH coordinator and the warehouses were situated in Geneva, +\item In the first one, called ``distributed XWCH'' in the following, + the XWCH coordinator and the warehouses were located in Geneva, Switzerland while the workers were running in the same local cluster in Belfort, France. @@ -313,7 +314,7 @@ distincts deployments of XWCH: the same local cluster. \end{enumerate} -For the both deployments, during the day these machines were used by +For both deployments, during the day these machines were used by students of the Computer Science Department of the IUT of Belfort. In order to evaluate the overhead induced by the use of the platform @@ -324,7 +325,7 @@ by the use of shell scripts. No specific middleware was used and the workers were in the same local cluster. Finally, five computation precisions were used: $1e^{-1}$, $0.75e^{-1}$, -$0.50e^{-1}$, $0.25e^{-1}$ and $1e^{-2}$. +$0.50e^{-1}$, $0.25e^{-1}$, and $1e^{-2}$. \subsubsection{Results} @@ -332,17 +333,19 @@ $0.50e^{-1}$, $0.25e^{-1}$ and $1e^{-2}$. Table \ref{tab:neurad_res} presents the execution times of the Neurad application on 25 machines with XWCH (local and distributed -deployment) and without XWCH. These results correspond to the measure -of the same step for both kind of execution i.e. sending of local data and the -executable, the learning process, and retrieving the results. The -results represent the average time of $x$ executions. +deployment) and without XWCH. These results correspond to the measures +of the same steps for both kinds of execution, i.e. sending of local +data and the executable, the learning process, and retrieving the +results. Results represent the average time of $?? x ??$ executions. \begin{table}[h!] + \renewcommand{\arraystretch}{1.7} \centering \begin{tabular}[h!]{|c|c|c|c|c|} \hline - Precision & 1 machine & Without XWCH & With XWCH & With local XWCH\\ + ~Precision~ & ~1 machine~ & ~Without XWCH~ & ~With XWCH~ & ~With + local XWCH~ \\ \hline $1e^{-1}$ & 5190 & 558 & 759 & 629\\ $0.75e^{-1}$ & 6307 & 792 & 1298 & 801 \\ @@ -351,6 +354,7 @@ results represent the average time of $x$ executions. $1e^{-2}$ & 11030 & 1035 & 1447 & 1108 \\ \hline \end{tabular} + \vspace{0.3cm} \caption{Execution time in seconds of the Neurad application, with and without using the XWCH platform} \label{tab:neurad_res} \end{table} @@ -381,12 +385,12 @@ though the same steps were done, all transfer processes are inside a local cluster with a high bandwidth and a low latency. Whereas when using XWCH, all transfer processes (between datawarehouses, workers, and the coordinator) used a wide network area with a smaller -bandwidth. In addition, in executions without XWCH, all the machines +bandwidth. In addition, in executions without XWCH, all the machines started immediately the computation, whereas when using the XWCH platform, a latency is introduced by the fact that a computation starts on a machine, only when this one requests a task. -This underline that, unsurprisingly, deploying a local +This underlines that, unsurprisingly, deploying a local coordinator and one or more warehouses near a cluster of workers can enhance computations and platform performances. @@ -396,15 +400,15 @@ enhance computations and platform performances. In this paper, we have presented a gridification of a real medical application, the Neurad application. This radiotherapy application tries to optimize the irradiated dose distribution within a -patient. Based on a multi-layer neural network, this applications -present a very time consuming step i.e. the learning step. Due to the +patient. Based on a multi-layer neural network, this application +presents a very time consuming step, i.e. the learning step. Due to the computing characteristics of this step, we choose to parallelize it using the XtremWeb-CH global computing environment. Obtained experimental results show good speed-ups and underline that overheads induced by XWCH are very acceptable, letting it be a good candidate for deploying parallel applications over a global computing environment. -Our future works, include the testing of the application on a more +Our future works include the testing of the application on a more large scale testbed. This implies, the choice of a data input set allowing a finer decomposition. Unfortunately, this choice of input data is not trivial and relies on a large number of parameters diff --git a/xwch.tex b/xwch.tex index 70a2b12..f90168b 100644 --- a/xwch.tex +++ b/xwch.tex @@ -10,7 +10,7 @@ The coordinator is the main component of the XWCH platform. It controls user access and schedules jobs to workers. It provides a web interface for managing jobs and users, and a set of web services. These are user service and worker/warehouse services -implemented using WSDL \cite{WebServ2002}. +implemented using WSDL \cite{WebServ2002} (DEFINITION). A worker is a Java daemon that runs on the user machine. Assumed to be volatile, the workers report periodically themselves to the @@ -25,32 +25,32 @@ shown in Figure \ref{xwch}. \label{xwch} \begin{centering} \includegraphics [scale=0.2]{figures/xwcharchitecture.pdf} - \caption{The XWCH Architecture} + \caption{The XtremWeb-CH architecture} \end{centering} \end{figure} A warehouse is a file server that acts as a data storage system for workers and client programs. Workers may not necessarily be able to -communicate directly with each other, due to firewalls and NAT +communicate directly with each others, due to firewalls and NAT sub-networks. For these reasons, warehouses are used as intermediaries to exchange, store and retrieve data. Job submission is done by a client program which is written using a flexible API, available for Java and C/C++ programs. The client program runs on a “client node” and calls the user services to submit -jobs (Figure \ref{xwch}, (1)). The main flexibility provided by the use of this +jobs (Figure \ref{xwch}, (1)). The main flexibility provided by the use of this architecture is to control and generate dynamically jobs especially when their number cannot be known in advance. Communications between the coordinator and the workers are always initiated by the workers -following a pull model (Figure \ref{xwch}, (2)): +following a pull model (Figure \ref{xwch}, (2)): \begin{itemize} -\item Workers receive jobs (Figure \ref{xwch}, (3)) only if they send a “work +\item Workers receive jobs (Figure \ref{xwch}, (3)) only if they send a “work request” signal; \item When a worker finishes its job, it stores its output file on - warehouse and sends a “work result” signal to the coordinator; + a warehouse and sends a “work result” signal to the coordinator; \item During its execution, a worker (respectively warehouse) - periodically sends “work alive” to the worker services (respectively - warehouse services) to report itself to the coordinator. + periodically sends “work alive” to the worker service (respectively + warehouse service) to report itself to the coordinator. \end{itemize} As a whole, XWCH is easy to install, maintain and use. Its components -- 2.20.1