brain2013 posterweb.eecs.umich.edu/~dkoutra/papers/are_all_brains_wired...Title brain2013_poster.ppt...
1
!"# !%% &"'()* +("#, -./'%%01 !"#"$ &'()*" + ,( -'#. + /012$*" 345"# 6 307 8(#. 6 8'92(" :; <'.0=9)0$# > ?2*$9)'9 @"='()9'9 + "#$#%&'()'*+),"+ -%./%0#).1$.2+&.*#%3)'1* (/2*#$&+$*),"+ /4+$.&(#3+")+$*),"+ 4151&(6#6)"+7,),"+ '8/%(61(&'()'*+),"+ " #$%&'()' *'++,& -&).'%/)01 2 -&).'%/)01 ,3 4'5 *'6)7, 8 9:;' -&).'%/)01 234(5'4(3) •91$$,'61*%'( :: '/,#01$ 1; </#%$ '1$$,'05%62 *#-() •=$#32>%$. 68, '1$$,'61*,( 61 +$",/(6#$" 81? 68, </#%$ ;+$'01$() • =/, 68,/, "%@,/,$',( <,6?,,$ "%@,/,$6 -,1-3,A • B,*#3, 5() *#3, • C%.8 *#68 (7%33( 5() $1/*#3 • =/0(6( 5() ('%,$0(6(D • =/, #33 -#/6( 1; 68, </#%$ '1$$,'6," (%*%3#/32A • E,F 8,*%(-8,/, • G%.86 8,*%(-8,/, -67#"(8#)4'% 9#*/%4* 2#4:3,3%3;0 <=><?@AB=>A H A'B0= "11*'"C2I #$#32>, %$5#/%#$6( 1; $+*,/1+(J <%.:('#3, </#%$ ./#-8( K'1$$,'61*,(L %$ 1/",/ 61 "1 '3+(6,/%$. #$" '3#((%M'#01$) HDE090*B"F'#9I !N8, (%>, K$+*<,/ 1; ,".,(LJ #( ?,33 #( 68, *#O%*+* ,%.,$5#3+,( 1; 68, </#%$ ./#-8( "%@,/ (%.$%M'#$632 <,6?,,$ *#3,( #$" ;,*#3,() !N8, ",./,, "%(6/%<+01$J #$" 68, $+*<,/ 1; 6/%#$.3,( #/, ;,#6+/,( 68#6 '#$ '1$6/%<+6, 61?#/"( 68, '3#((%M'#01$ 1; 68, ('#$( <2 .,$",/) C'4'*#4 =/" !77"3'D: 1 2 • P'#3#/ B,#6+/,( =$#32(%( ① $+*<,/ 1; $1",( ② $+*<,/ 1; ,".,( ③ 3#/.,(6 ,%.,$5#3+, ④ $+*<,/ 1; 6/%#$.3,( ⑤ $+*<,/ 1; '1$$,'6," '1*-1$,$6( ⑥ *#O%*+* -#.,/#$7 ⑦ *%$%*+* -#.,/#$7 • Q,'61/ B,#6+/,( ① ",./,, "%(6/%<+01$ ② -#.,/#$7 "%(6/%<+01$ ③ /#"%+( "%(6/%<+01$ ④ #--/1O%*#6, /#"%+( "%(6/%<+01$ ⑤ M/(6 $:68 ,%.,$5#3+,( "%(6/%<+01$ ⑥ 6/%#$.3, "%(6/%<+01$ A4#7 EF &"'() C(GG#"#)D# F G = • 91$$,'61*,( 1; RRS -,1-3, • T<6#%$," <2 U+30*1"#3 U#.$,0' G,(1$#$', V*#.%$. KUGVL • W#'8 '1$$,'61*, %( /,-/,(,$6," #( +$?,%.86," #$" +$"%/,'6," ./#-8) • Q1O,3(X$1",(I SYZ[:YR\[ • 91$$,'01$(X,".,(I Y)RSU:R])SZU • =^/%<+6,( -,/ -,/(1$I #.,J .,$",/J V_J '/,#05%62 %$",OJ D • N1137%6 • `Wa=PbP • c,6?1/7O C#;"## C(*4"(H/4(3) <)%/0 ./= /'7,&> ?%)&7)?$+ 7,@?,&'&0 ,3 0A' @$0%)6 B%$)& /7$& C &,>'D>'(%''D>)/0%= DE90*B"F'#I N8, ,O6/,*, '#(,( 1; ;,*#3, #$" *#3, '1$$,'61*,( #/, ?,33 (,-#/#6,") C1?,5,/J 68, ",./,, "%(6/%<+01$ '#$$16 (,-#/#6, '1$$,'61*,( ?)/)6) 68, *#(6,/2 1; *#68) I"(');%#* J%34* DE90*B"F'#I N/%#$.3, "%(6/%<+01$ '#$$16 (,-#/#6, 68, ./1+-() N16#3 d 1; 6/%#$.3,( 5() d 1; ,".,( '#$ 8,3- (,-#/#6, ;,*#3,( ;/1* *#3,(J <+6 '#$$16 "%(0$.+%(8 68, ./1+-( ?)/)6) 68, 3,5,3 1; *#68 (7%33() <3))#D4#, <3873)#)4* DE90*B"F'#9I RL N8, "%(6/%<+01$ 1; '1$$,'6," '1*-1$,$6( G'09 #') "%@,/ (%.$%M'#$632 <,6?,,$ *#3,( #$" ;,*#3,() ZL N8, d1; ,".,( 5() $1",( %$ 68, .%#$6 '1$$,'6," '1*-1$,$6 1; ,#'8 '1$$,'61*, /,5,#3( 6?1 '3+(6,/( '1//,(-1$"%$. 61 *#3,( #$" ;,*#3,() ?'";#*4 -(;#)5'%/# 3G K"'7: 2'4"(6 9#G#"#)D#* Ha/#2 eJ f1.15%' g=JQ1.,3(6,%$ gNJ E#$"*#$ f=J `/%$',gEJ Q1.,3(6,%$ GgJ hU#.$,0' /,(1$#$', '1$$,'61*, #+61*#6," -%-,3%$,I=$ 15,/5%,?hJ `+3(,J VWWWJ 513) iJ $1) ZJ --) SZSjJ ZkRZ) H[#$. bJ N(1+/#7#7%( 9J B#31+6(1( 9)J h`Wa=PbPI = `,6#:P'#3, a/#-8 U%$%$. P2(6,* : V*-3,*,$6#01$ #$" T<(,/5#01$()hJ VWWW V$6,/$#01$#3 91$;,/:,$', 1$ l#6# U%$%$. KV9lULJ U%#*%J B31/%"#J bP=J ZkkY) H*"$# -*"12 /2"B$#. @0")(*0 07)*"CF'# ... ... 1 2 m . . . 3 1 2 . . . n g r a p h s Useful features Connectomics a/#-8( #/, •+$?,%.8," •+$"%/,'6," Manually divide the graphs into different groups according to labeled attributes p-value significance Plot the different invariants and test whether they help distinguish the groups (e.g., female/male). ... ... 1 2 d . . . 3 1 2 . . . n g r a p h s the single vector feature nxk U kxk Left singular vectors u i S s i Singular values kxd v i Right singular vectors V T u 1 u 2 ALC J%34 3/4 / E LA / M u 1 u 2 u 1 u 2 N8, ;,#6+/, '#$$16 "%(0$.+%(8 "%@,/,$6 ./1+-( Possible conclusion 1 N8, ;,#6+/, '1$6/%<+6,( %$ -/,"%'0$. 68, ./1+- *,*<,/(8%-() Possible conclusion 2 J"#%(8()'"(#* - 114 connectomes in total - Groups divided in two ways: R)a,$",/ mk ;,*#3,(I /," "16( #$" \S *#3,(I ./,,$ "16( Z)P+<4,'6 N2-,I n E,5,3 1; *#68 (7%33( c1/*#3I ./,,$ "16( E1?I /," "16( C%.8I <3+, "16( u 1 VS u 2 of the matrix of the connected components distribution N8/,(813" 1$ 68, ,"., ?,%.86( 61 /,*15, $1%(,) C/N# @)(5#"*(40 @)(5#"*(40 3G >#O 2#6(D3 N8, 6?1 .,$",/( "%@,/ (%.$%M'#$632 %$ 68%( ;,#6+/, KZ:(#*-3, 6:6,(6I -:5#3+, k)kkkZL) Graph analysis Wo'%,$6 %$ (,-#/#0$. "%@,/,$6 ./1+-(
Transcript of brain2013 posterweb.eecs.umich.edu/~dkoutra/papers/are_all_brains_wired...Title brain2013_poster.ppt...
!"#$!%%$$&"'()*$+("#,$-./'%%01$!"#"$%&'()*"+%%%%%%%%%%%%%%%%%%%%,(%-'#.+%%%%%%%%%%%%%%%%%%/012$*"%345"#6%%%%%%%%%%%%%%%%%307%8(#.6%%%%%%%%%%%%%%%%8'92("%:;%<'.0=9)0$#>%%%%%%%%%%%%%%%?2*$9)'9%@"='()9'9+%%%%%%%%%%
!!!!!!!!!!!!!!!"#$#%&'()'*+),"+!!!!!!!!!!-%./%0#).1$.2+&.*#%3)'1*!!!!!!!!!(/2*#$&+$*),"+!!!!!!!!!!!!!/4+$.&(#3+")+$*),"+!!!!!!!!!!!!!!!!!4151&(6#6)"+7,),"+!!!!!!!!!!!!!!!!!!!!!!!!!!!'8/%(61(&'()'*+),"+!
!"#$%&'()'!*'++,&!-&).'%/)01 ! ! !!2-&).'%/)01!,3!4'5!*'6)7,! !89:;'!-&).'%/)01%
234(5'4(3)!• 91$$,'61*%'(!::!'/,#01$!1;!</#%$!'1$$,'05%62!*#-()!!!
• =$#32>%$.!68,!'1$$,'61*,(!61!!+$",/(6#$"!81?!68,!</#%$!;+$'01$()!!
• =/,!68,/,!"%@,/,$',(!!<,6?,,$!"%@,/,$6!-,1-3,A!• B,*#3,!5()!*#3,!
• C%.8!*#68!(7%33(!5()!$1/*#3!
• =/0(6(!5()!('%,$0(6(D!• =/,!#33!-#/6(!1;!68,!</#%$!'1$$,'6,"!(%*%3#/32A!
• E,F!8,*%(-8,/,!
• G%.86!8,*%(-8,/,!!!
-67#"(8#)4'%$9#*/%4*$
2#4:3,3%3;0$$
<=><?@AB=>A$
H %A'B0=%"11*'"C2I!!#$#32>,!%$5#/%#$6(!1;!$+*,/1+(J!<%.:('#3,!</#%$!./#-8(!
K'1$$,'61*,(L!%$!1/",/!61!"1!'3+(6,/%$.!#$"!'3#((%M'#01$)!
H DE090*B"F'#9I!! N8,!(%>,!K$+*<,/!1;!,".,(LJ!#(!?,33!#(!68,!*#O%*+*!,%.,$5#3+,(!1;!68,!</#%$!
./#-8(!"%@,/!(%.$%M'#$632!<,6?,,$!*#3,(!#$"!;,*#3,()!
! N8,!",./,,!"%(6/%<+01$J!#$"!68,!$+*<,/!1;!6/%#$.3,(!#/,!;,#6+/,(!68#6!'#$!
'1$6/%<+6,!61?#/"(!68,!'3#((%M'#01$!1;!68,!('#$(!<2!.,$",/)!
C'4'*#4$
=/"$!77"3'D:
1
2
• P'#3#/!B,#6+/,(!=$#32(%(!!① $+*<,/!1;!$1",(!
② $+*<,/!1;!,".,(!
③ 3#/.,(6!,%.,$5#3+,!④ $+*<,/!1;!6/%#$.3,(!!
⑤ $+*<,/!1;!'1$$,'6,"!!
!!'1*-1$,$6(!
⑥ *#O%*+*!-#.,/#$7!
⑦ *%$%*+*!-#.,/#$7!
• Q,'61/!B,#6+/,(!!① ",./,,!"%(6/%<+01$!② -#.,/#$7!"%(6/%<+01$!!③ /#"%+(!"%(6/%<+01$!!④ #--/1O%*#6,!/#"%+(!"%(6/%<+01$!
⑤ M/(6!$:68!,%.,$5#3+,(!"%(6/%<+01$!!!⑥ 6/%#$.3,!"%(6/%<+01$!!
A4#7$EF$&"'()$C(GG#"#)D#$$
FG =
…
• 91$$,'61*,(!1;!!RRS!-,1-3,!
• T<6#%$,"!<2!U+30*1"#3!U#.$,0'!G,(1$#$',!V*#.%$.!KUGVL!
• W#'8!'1$$,'61*,!%(!/,-/,(,$6,"!#(!!+$?,%.86,"!#$"!
+$"%/,'6,"!./#-8)!
• Q1O,3(X$1",(I!SYZ[:YR\[!• 91$$,'01$(X,".,(I!Y)RSU:R])SZU!
• =^/%<+6,(!-,/!-,/(1$I!#.,J!.,$",/J!V_J!'/,#05%62!%$",OJ!D!• N1137%6!
• `Wa=PbP!• c,6?1/7O!
C#;"##$C(*4"(H/4(3)$<)%/0!./=!/'7,&>!?%)&7)?$+!7,@?,&'&0!,3!0A'!@$0%)6!B%$)&!/7$&!C!&,>'D>'(%''D>)/0%=!!
DE90*B"F'#I!!N8,!,O6/,*,!'#(,(!1;!;,*#3,!#$"!*#3,!'1$$,'61*,(!#/,!
?,33! (,-#/#6,")!C1?,5,/J! 68,!",./,,!"%(6/%<+01$!'#$$16!
(,-#/#6,!'1$$,'61*,(!?)/)6)!68,!*#(6,/2!1;!*#68)!I"(');%#*$J%34*$
DE90*B"F'#I!N/%#$.3,! "%(6/%<+01$! '#$$16! (,-#/#6,! 68,! ./1+-()!
N16#3!d!1;!6/%#$.3,(!5()!d!1;!,".,(!'#$!8,3-!(,-#/#6,!
;,*#3,(! ;/1*! *#3,(J! <+6! '#$$16! "%(0$.+%(8! 68,!
./1+-(!?)/)6)!68,!3,5,3!1;!*#68!(7%33()!
<3))#D4#,$<3873)#)4*$$
DE90*B"F'#9I!RL! N8,! "%(6/%<+01$! 1;! '1$$,'6,"! '1*-1$,$6(!
G'09%#')!"%@,/!(%.$%M'#$632!<,6?,,$!*#3,(!#$"!
;,*#3,()!
ZL! N8,! d1;! ,".,(! 5()! $1",(! %$! 68,! .%#$6!
'1$$,'6,"! '1*-1$,$6! 1;! ,#'8! '1$$,'61*,!
/,5,#3(!6?1!'3+(6,/(!'1//,(-1$"%$.!61!*#3,(!#$"!
;,*#3,()!?'";#*4$-(;#)5'%/#$3G$$K"'7:$2'4"(6$$
9#G#"#)D#*$H a/#2!eJ!f1.15%'!g=JQ1.,3(6,%$!gNJ!E#$"*#$!f=J!`/%$',gEJ!Q1.,3(6,%$!GgJ!hU#.$,0'!/,(1$#$',!
'1$$,'61*,!#+61*#6,"!-%-,3%$,I=$!15,/5%,?hJ!`+3(,J!VWWWJ!513)!iJ!$1)!ZJ!--)!SZSjJ!ZkRZ)!!
H [#$.!bJ!N(1+/#7#7%(!9J!B#31+6(1(!9)J!h`Wa=PbPI!=!`,6#:P'#3,!a/#-8!U%$%$.!P2(6,*!:
V*-3,*,$6#01$!#$"!T<(,/5#01$()hJ!VWWW!V$6,/$#01$#3!91$;,/:,$',!1$!l#6#!U%$%$.!KV9lULJ!
U%#*%J!B31/%"#J!bP=J!ZkkY)!
H*"$#%-*"12%/2"B$#.%
@0")(*0%07)*"CF'#%%
...
...
1 2 m . . . 3 1 2 . . . n
graphs
Useful features
Connectomics
a/#-8(!#/,!
• +$?,%.8,"!!• +$"%/,'6,"!
Manually divide the graphs into different groups according to labeled attributes
p-value significance
Plot the different invariants and test whether they help distinguish the groups (e.g., female/male).
...
...
1 2 d . . . 3 1 2 . . . n
graphs
the single vector feature
nxk
U kxk
Left singular vectors
ui S si
Singular values
kxd
vi
Right singular vectors
VT
u1 u2
ALC$
J%34$3/4$/E$LA$/M$$
u1
u2
u1
u2
N8,! ;,#6+/,! '#$$16! "%(0$.+%(8!
"%@,/,$6!./1+-(!!
Possible conclusion 1 N8,!;,#6+/,!'1$6/%<+6,(!%$!
-/,"%'0$.!68,!./1+-!*,*<,/(8%-()!
Possible conclusion 2
J"#%(8()'"(#*$$$
- 114 connectomes in total - Groups divided in two ways: R) a,$",/!
mk!;,*#3,(I!/,"!"16(!#$"!!
\S!*#3,(I!./,,$!"16(!
Z) P+<4,'6!N2-,I!!n!E,5,3!1;!*#68!(7%33(!
c1/*#3I!./,,$!"16(!!
E1?I!/,"!"16(!
C%.8I!<3+,!"16(
u1 VS u2 of the matrix of the connected components distribution
N8/,(813"!1$!
68,!,".,!
?,%.86(!61!
/,*15,!$1%(,)!
C/N#$@)(5#"*(40$
@)(5#"*(40$3G$$>#O$2#6(D3$
N8,!6?1!.,$",/(!"%@,/!(%.$%M'#$632!%$!68%(!
;,#6+/,!KZ:(#*-3,!6:6,(6I!-:5#3+,!k)kkkZL)
Graph analysis Wo'%,$6!%$!
(,-#/#0$.!
"%@,/,$6!
./1+-(
