text
stringlengths 0
89.3k
|
---|
tween the present post being visited at time twith the users
|
interests vector and the considered post to be accessed in time
|
t 1 with the users interests vector Formula 1 shows
|
how the information scent is calculated when considering to
|
move from the current post located on the edge eito one of
|
its neighbors located on the edge ej
|
InfoScent ej Simej ISimei I 1
|
Where
|
Iis the users interests vector
|
eiis the current post
|
ejNi with Nibeing the set of adjacent edges to ei
|
ieeis neighborhood
|
Simrepresents the cosine similarity between two
|
vectors
|
The main role of the information scent is to guide the for
|
aging a positive value indicates that we are approaching a
|
relevant post in the social graph whereas a negative value
|
indicates the oppositeD Trail surfing path
|
While foraging food animals follow a certain path that al
|
lows them to reach food sources in an optimal way according
|
to the OFT Web users have a similar behavior as they visit
|
Web pages one at a time until reaching a relevant page that
|
satisfies their information needs constructing there a surfing
|
path composed of one or more Web pages
|
The information foraging process starts from an initial post
|
and progresses through each step attempting to reach a post
|
with more relevant information than its predecessor A surf
|
ing path is built for this purpose starting with an initial
|
contentsharing edge and then being enriched by adding fur
|
ther edges to create a chain of related posts This means
|
that at each step of the foraging process the system should
|
choose one contentsharing edge to visit from the neighbor
|
hood of the current post This choice is made based on For
|
mula 2
|
Peiej
|
0 ifInfoScent ej0
|
InfoScent ejP
|
elNpiInfoScent elotherwise
|
2
|
where
|
Peiejis the likelihood of selecting the edge ej
|
among the reachable edges from the current edge ei
|
Npiis the set of adjacent contentsharing edges of the
|
edge eiwith a positive information scent value ie
|
elNpiInfoScent el0
|
IV Adapted elephant herding optimization to
|
information foraging
|
The social structure of elephants is complex varying by gen
|
der and population dynamics Adult females form a matriar
|
chal societies while adult males are usually solitary 29 30
|
A herd structure is similar to concentric rings with the in
|
nermost circle comprising a family unit of related female
|
adults A family unit is formed by the eldest most domi
|
nant female called the matriarch as well as her adult daugh
|
ters their calves and a number of juveniles The male calves
|
leave the herd when reaching adulthood generally between
|
the age of 12 and 15 From this stable core the groupings
|
widen to include less familiar individuals A clan is formed
|
when elephants gather in large groups consisting of different
|
herds The functioning of the elephants society is illustrated
|
by Figure 4
|
We decided to combine the information foraging theory with
|
EHO in order to efficiently identify useful information in
|
large social graphs Each elephant will look for relevant so
|
cial posts by browsing a section of the graph in this manner
|
constructing surfing paths that lead to relevant information
|
The elephants perform the foraging while taking advantage
|
of the hierarchy and organization of their society This will
|
allow them to collaborate and find relevant information in a
|
more effective and efficient way
|
In this section we adapt the basic Elephant Herding Opti
|
mization algorithm to information foraging Note that the
|
original EHO was developed to address continuous prob
|
lems whereas information foraging is a discrete combinaEnhanced Elephant Herding Optimization for Large Scale Information Access on Social Media 32
|
Figure 4 Elephants society structure
|
torial problem Further improvements to basic EHO will be
|
introduced in sections V and VI
|
A Generating the elephant population and assigning the po
|
sitions
|
To create an elephant population with pclans we first gener
|
atepdifferent elephants with respect to distClan which rep
|
resents the shortest minimal distance between clans The rest
|
of the elephants in each clan are then formed using the posi
|
tions of the pelephants with respect to distElephant which
|
represents the maximum distance between elephants in the
|
same clan Figure 5 depicts a population of three clans scat
|
tered throughout a social graph
|
Figure 5 A population of elephants distributed over a social
|
graph
|
We assign mdifferent positions to a social graph with m
|
contentsharing edges one for each edge These positions
|
are represented by integer values in the interval 1 m Each
|
elephant jbelonging to clan ciis identified by a unique po
|
sition denoted by the xcij An elephants position at time t
|
is the position of the edge it is visiting at that timeB Surfing paths construction
|
Each elephant is assigned the duty of building a surfing path
|
leading to relevant information during one iteration of the
|
EHO algorithm This is accomplished using Algorithm 1
|
Algorithm 1 Building a surfing path
|
Input xcij elephants position I users interests G so
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.