Image Processing Reference
In-Depth Information
12.4.1 User Interaction Probability
Let us assume that in a social network, there are
N
t
types of interaction between
N
u
users (e.g. chat, talk, like). A simple interaction probability
Ip
u
1
,
u
2
from user u
1
to
user
u
2
for a time period
t
could be defined as:
X
N
u
1
,
u
2
Nt
m
N
u
m
Ip
u
1
,
u
2
¼
w
m
ð
12
:
1
Þ
m¼
1
where
N
u
1
,
u
m
is the number of interactions of type
m
from user
u
1
to user
u
2
,
N
u
m
is the
number of interactions of type
m
from user
u
1
to any other user and
w
m
is a weight
factor showing the importance of interaction of type
m,
where:
X
Nt
w
m
¼
1
ð
12
:
2
Þ
m
¼
1
As many sessions could have more than two participants, it is:
X
Nu
N
u
m
6¼
N
u
1
,
u
k
m
ð
12
:
3
Þ
k
¼
1
This probability metric can determine the level of “connectedness” between
users
u
1
and
u
2
. It is worth noting that in general,
Ip
u
1
,
u
2
Ip
u
2
,
u
1
, as relationships
may not always be symmetrical, e.g. if
u
1
has established an immersive session, it is
very possible that also
u
2
will be connected. However,
u
2
being more social may
have sessions with other persons, without
u
1
participation. Moreover it is important
to note that interactions between more than two users have not been taken into
account. This is also an important factor, but it goes well beyond the scope of this
chapter.
6¼
12.4.2 User Mobility Probability
In order to put this approach into effective action, we need also to take into account
the mobility patterns of users. Users that connect to the 3D immersive network from
continuously changing network locations are not suitable candidates for this kind of
cluster-based optimization. This can be represented by a Connectivity Variation
probability a.k.a. how often a user is connected from the same location.
Let us assume that in a time period
t
, a user
u
1
is connected
N
u
1
times from
N
n
different geographical nodes/locations. The probability
Cp
u
1
,
n
1
of
u
1
to be
connected via node/location
n
1
is
