Neural%20Networks%20in%20Social%20Networks

1
Neural Networks in Social Networks
Student Nikolic Filip nf143006m_at_student.etf.rs
Professor Veljko Milutinovic Assistant Bojan
Furlan
2
Tasks
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

3
Existing 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.
4
What is Neural Networks?
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion

• Analogy -Neuron in the brain

5
What 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
6
What 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)
7
Training 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
8
Training 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
9
Training 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
10
Training a Neural Network
Data Mining Neural Networks
School of Electrical Engineering, Belgrade
Introducing What isNeuralNetworks? Training F
eature Conclusion

• Backward propagation

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
11
Training 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
12
Advantages
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.

13
Disadvantages
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.

14
Conclusion
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