Matching in Planar Graphs
Matching in Planar Graphs
in
Planar
Graphs
Mentor
–
Prof.
Surender
Baswana
Shubham
Tulsiani
(Y9574)
Matching
• M
–
a
subset
of
edges
that
do
not
share
a
common
vertex
Maximum
Matching
• Separator
theorems
are
one
of
the
most
common
tools
used
in
many
planar
graph
algorithms
G1
G2
G2
G1
Linear
Reductions
removing
verGces
with
degree
>
3
G1,M1
S1
S2
G2,M2
Combining
Solutions
to
Partitions
–
Basic
Results
CombineBasic(G1,
G2,
S,
M1,
M2)
• M
:=
M1
U
M2
• Let
S
=
{s1,s2,...,sk}
• H
:=
G1
U
G2
• For
i
=
1..k
• H
:=
H
U
{si}
• Pi
:=
AugmenGng
path
from
si
[possibly
empty]
• M
:=
M
(xor)
Pi
• return
M
Time
Complexity
:
O(kn)
Q)
What
if
{Pi
|
i=1..k}
are
vertex
disjoint
?
Combining
Solutions
to
Partitions
–
Basic
Results
• Are
the
{Pi
|
i=1..k}
discovered
necessarily
vertex
disjoint
?
Ans
-‐
No
• Is
it
possible
to
obtain
{Pi
|
i=1..k}
that
are
vertex
disjoint
?
Ans
-‐
Yes
Simpli@ied
Versions
We
cannot
have
(fi,sj)
and
(fk,sl)
both
in
M
if
(i
<
k
and
j
>
l)
or
(i
>
k
and
j
<
l
)
because
the
two
edges
would
then
cross.
Simpli@ied
Versions
• For
the
problems
shown
in
the
previous
slide,
we
can
have
dynamic
programming
algorithms.
A
key
property
used
is
that
a
greedy
property
of
matching
a
separator
vertex
to
the
topmost
available
vertex.
• ObservaGons
–
• Greedy
strategies
may
work
for
general
case
• If
|F1|,|F2|
is
large,
we
may
not
find
an
efficient
algorithm
Additional
Constraints
on
Subgraph
Matchings
We
wish
to
have
an
algorithm
AlterMatching(G1,S1,M1)
which
outputs
a
matching
M1’
s.t.
–
• M1’
is
a
maximum
matching
• There
are
no
even
length
alternaGng
paths
from
any
vertex
v
in
F1’
to
any
vertex
s
in
S1
s1
s1
✔
f1 f2 v f2
s2
s2
For
biparGte
graphs,
it
is
possible
to
reduce
this
to
a
max-‐flow
problem.
Thus,
there
might
be
a
more
efficient
and
direct
way
to
achieve
these
condiGons
in
near
linear
Gme.
Nature
of
Augmenting
Paths
–
Vertex
Separators
• Paths
can
lie
only
to
one
side
as
all
verGces
in
S
are
free
S
• But,
no
augmenGng
path
possible
to
any
free
vertex
in
F1
or
F2
due
to
addiGonal
constraints
So,
yellow
path
not
possible
Nature
of
Augmenting
Paths
–
Edge
Separators
• Paths
can
cross
back
and
forth
between
G1
and
G2
• However,
no
augmenGng
path
possible
to
any
free
vertex
in
F1
or
F2
due
to
addiGonal
constraints
• So,
the
blue
path
is
not
allowed
• Thus,
any
augmenGng
path
is
between
two
verGces
of
S
(S
=
S1
U
S2)
Removing
Free
Vertices
• In
both
these
cases,
there
are
no
augmenGng
paths
induced
by
M
wrt
G
that
pass
through
F1
or
F2
• The
maximum
matching
of
G
\
{F1
U
F2}
is
the
same
as
maximum
matching
for
G
• Thus,
the
only
free
verGces
remaining
are
in
S,
S1,
S2
Bipartite
Graphs
and
Edge
Separators
• All
augmenGng
paths
are
from
a
vertex
in
S1
to
a
vertex
in
S2.
When
progressing
iteraGvely
from
the
topmost
free
vertex
in
S1,
a
greedy
property
that
is
valid
is
that
we
should
augment
along
the
’outermost’
augmenGng
path
Additional
Processing
and
Greedy
Strategies
• Unfortunately,
the
previous
condiGon
is
not
always
possible.