Introduction Graph estimation Adaptive tests Minimax bounds

Nicolas Verzelen

Description

Introduction Graph estimation Adaptive tests Minimax bounds Estimation and tests for Gaussian graphical models Nicolas Verzelen Workshop on random graphs 1/28

• ?? z


• introduction graph


• markov local


• estimation adaptive


• ?? z?


• tests minimax


• undirected graphical


• interaction between

Informations

• Publié par: profil-urra-2012
• Nombre de lectures: 34
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Intro
duction
Graph
estimation
on
mo
erzelen
A
Gaussian
daptive
Nicolas
tests
o
Minimax
graphs
b
graphical
ounds
dels
Estimation
V
and
W
tests
rkshop
fo
random
r
1/28−= ( ,..., )∼N ( ,Ω )+ +
Ω
Γ :={ ,..., + }
= (Γ, )
( ) :
2
1
3
5
4
0
1
ne
rs
1
tests
p
A
1
G
singula
Minimax
r.
undirected
p
daptive
Z
neighb
1
a
Z
Graph
1
b
Z
E
consider
nite
e
graph.
W
G
del
a
p
estimation
mo
o
1
of
graphical
in
G
.
Undirected
duction
ounds
Intro
non
2/28−= ( ,..., )∼N ( ,Ω )+ +
Ω
Γ :={ ,..., + }
= (Γ, )
( ) :
2
1
3
( ⊥⊥ )|−{ , ( )} ( )
5
4
r.
singula
non
p
e
tests
at
1
a
1
Graph
0
ect
G
del
1
ounds
p
G
1
ert
p
with
Z
G
1
a
nite
mo
undirected
ne
graph.
b
ne
Z
1
a
p
G
Z
rop
a
y
Z
a
consider
resp
neighb
to
o
if
rs
Z
of
W
a
Z
in
graphical
G
Undirected
.
G
Z
a
satises
Minimax
the
daptive
Ma
ne
rk
A
ov
estimation
lo
duction
cal
Intro
E
2/28−= ( ,..., )∼N ( ,Ω )+ +
Ω
Γ :={ ,..., + }
= (Γ, )
( ) :
2
1
3
( ⊥⊥ )|−{ , ( )} ( )
5
4
⇐⇒
∈Γ
p
1
1
Z
Undirected
graphical
G
e
r.
tests
singula
Graph
non
Z
1
a
E
ounds
0
Z
nite
resp
undirected
the
graph.
1
ne
consider
G
G
1
mo
a
ne
p
b
1
A
rk
a
p
del
ert
to
fo
Z
any
cal
p
a
.
Z
.
Z
Z
a
satises
ne
the
W
Ma
del
rk
graphical
ov
Z
lo
G
cal
a
p
Minimax
rop
daptive
ert
estimation
y
is
at
Gaussian
a
mo
with
with
resp
ect
ect
G
to
duction
G
satises
if
lo
Z
Ma
neighb
ov
o
rop
rs
y
of
r
a
a
Intro
in
G
2/28Intro
duction
Graph
genes
een
ro
estimation
interactions
A
genes
daptive
des
tests
ertices
Minimax
et
b
w
ounds
their
Ex.
that
1
:
:
V
Genetic
:
net
b
w
w
o
t
rk
o
of
and
E.
p
Coli
ducts
No
3/28Intro
duction
Graph
estimation
(RNAm)
=
gene
A
ranscriptomic
daptive
of
tests
ression
Minimax
T
b
data
ounds
measures
T
the
ranscriptomic
exp
data
levels
analysis
4/28Intro
duction
Graph
using
in
rt
estimation
the
A
:
daptive
net
tests
Analysis
Minimax
endence
b
data.
ounds
a
T
this
ranscriptomic
o
data
data.
analysis
of
T
dep
ranscriptomic
structure
data
the
=
Goal
measures
Inferring
of
pa
the
of
gene
gene
exp
w
ression
rk
levels
transcriptomic
(RNAm)
4/28Ω = ⇐⇒ ( ⊥⊥ )|, −{ , }
1 2 3 4 521
1 * * 0 0 *
3 2 0* * * *
3 0 0* * *
5
4 4 0 * * * *
5 0 0* * *
A
a
ert
Z
Prop
duction
a
ounds
a
0
y
estimation
.
Z
b
b
b
Z
b
tests
Graph
daptive
Minimax
Intro
5/28Ω = ⇐⇒ ( ⊥⊥ )|, −{ , }
1 2 3 4 521
1 * * 0 0 *
3 2 0* * * *
3 0 0* * *
5
4 4 0 * * * *
5 0 0* * *
b
Z
a
Z
w
graph.
estimation
0
interp
b
.
a
duction
y
b
ert
of
Prop
daptive
ounds
A
b
Graph
Minimax
T
Z
o
:
enets
a
scientic
b
retation
reduction
the
a
dimension
tests
in
analysis.
statistical
Intro
5/28Intro
duction
Graph
:
and
Ex.
estimation
analysis
A
(2002)
daptive
2
tests
Image
Minimax
Rue
b
Tjelmeland
ounds
6/28