Title: Neural%20Networks%20in%20Social%20Networks
1Neural Networks in Social Networks
Student Nikolic Filip nf143006m_at_student.etf.rs
Professor Veljko Milutinovic Assistant Bojan
Furlan
2Tasks
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Classification friend into group (family,
friend, colleague) - Recognition of friends
- Find new friend
- Find all friends
- Automatic add friend with new profile
3Existing algorithms
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Facebook-People You May Know
- LinkedIn-connection you may know
- Decision Tree
- DataBase Query
People You May Know looks at, among other things,
your current friend list and their friends, your
education info and your work info. If you are
already friends on Facebook with some people
from your last job, for example, you may find
some more of your former coworkers
Anything you do on LinkedIn site is tracked. For
every action, you get points. 5 order LinkedIn
suggestion effects.
4What is Neural Networks?
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Analogy -Neuron in the brain
5What is Neural Networks?
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Neuron model Logistic unit
input
output
hidden
family
address
neighbords
faculty
colleagues
work
black list
surname birthday education where are you
from hobby
friends from university childhood friends sports
friends ex-girlfriend/boyfriend
6What is Neural Networks?
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Neuron model Simple example-AND
bias
-30
20
- family
surname
where are you from
20
if
if
0 0
0 1
1 0
1 1
0 (-30)
0 (-10)
0 (-10)
1 (10)
7Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Pick a parameters of network architecture
Layer 1
Layer 2
Class 1
Input 1
Class 2
Input 2
Class 3
Input 3
Class 4
8Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Randomly initialize weights
Layer 1
Layer 2
Class 1
Input 1
Class 2
Input 2
Class 3
Input 3
Class 4
9Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Forward propagation by Training Set
Layer 1
Layer 2
Class 1
Input 1
Class 2
Input 2
Class 3
Input 3
Class 4
Training Set
In1 In2 In3 Out1 Out2 Out3 Out4
0 0 0 1 0 0 0
0 0 1 0 1 0 0
0 1 0 0 0 1 0
0 1 1 0 0 0 1
10Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
Layer 1
Layer 2
Class 1
Input 1
Class 2
Input 2
Class 3
Input 3
Class 4
Training Set
In1 In2 In3 Out1 Out2 Out3 Out4
0 0 0 1 0 0 0
0 0 1 0 1 0 0
0 1 0 0 0 1 0
0 1 1 0 0 0 1
11Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Loop through training set
Layer 1
Layer 2
Class 1
Input 1
Class 2
Input 2
Class 3
Input 3
Class 4
Training Set
In1 In2 In3 Out1 Out2 Out3 Out4
0 0 0 1 0 0 0
0 0 1 0 1 0 0
0 1 0 0 0 1 0
0 1 1 0 0 0 1
12Advantages
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- A neural network can perform tasks that a linear
program cannot. - When an element of the neural network fails, it
can continue without any problem by their
parallel nature. - A neural network learns and does not need to be
reprogrammed. - It can be implemented in any application and
without any problem. - Neural networks are the closest thing to having
an actual human operate a system (i.e., they can
"learn") - Easy implementation in parallel calculation
process.
13Disadvantages
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- The neural network needs training to operate.
- The operation of neural networks is limited to
the training process. - The architecture of a neural network is
different from the architecture of
microprocessors therefore, needs to be
emulated. - Neural networks are difficult to design.
- Requires high processing time for large neural
networks.
14Conclusion
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion
- Neural Networks are an imitation of the
biological neural networks, but much simpler
ones. - Social Networks behave as Neural Networks.
- Neural networks can imitatecomplex mathematical
functions, biological functions, and sociological
behavior. - If given sufficient training setThe enemy of my
enemy is my friend. - Literaturehttps//class.coursera.org/ml-007/lect
ure/indexhttp//home.etf.bg.ac.rs/vm/os/dmsw/osd
mswpreddatamining.html - QA