Slides used at IEEE Computational Intelligence Society, Bangalore Chapter:
Winter School On Emerging Topics in Computational Intelligence -Theory and Applications
An experimental study in using natural admixture as an alternative for chemic...
Computational Intelligence and Applications
1. Computational Intelligence and
Applications
IEEE Computational Intelligence
Society Bangalore Chapter
Winter
School On
Emerging Topics in
Computational Intelligence -Theory and Applications
S Chetan Kumar
Co-founder AiKaan
2. Topics covered
● ML is cutting-edge of AI
● DL is cutting-edge of cutting-edge
● Is tensorflow good playground for ANN?
● CI and AI will lead to GI ?
● Biologically motivated learnings are needed to solve
real world problems !!
Confused !!
3.
4. Back propagated RNN with Bayesian
optimization can prevent
Long Short-Term memory issues of
gradient descent
Explain me in simple terms !!
5. General Intelligence: to perform intellectual task that a human can
Artifical Intelligence
My long-term goal is to
reach General Intelligence
Com
putational Intelligence
6. CI vs AI
Computational Intelligence Artificial Intelligence
Soft Computing techniques Hard computing techniques
Follows fuzzy logic Follows binary logic
Nature inspired models Based on mathematical
models
Can work inexact and
incomplete data
Not very effective
Probabilistic results Deterministic results
7. Computational and Artificial Intelligence
Computational Intelligence
Artificial
Intelligence
Fuzzy logic and others
8. Principles of Computational Intelligence
Fuzzy Logic
Probabilistic
model
Learning
theory
Evolutionary
computing
Artificial
Intelligence
Hybrid Techniques
9. Artificial Intelligence
● Soft computing technique
● Machines trying to achieve general intelligence
● Machine learning is one of the technique
● Knowledge based system is one another
● ML has become more popular
12. AI, ML, DL
Deep learning
Feature/
Representation
learning
Machine learning
Artificial Intelligence
Machine learning
Feature/
Representational learning
Deep Learning
13. Machine learning
● Basic machine learning
Eg Logistic regression
● Feature or Representational learning
If there objects to be classified, which feature of the
object should I use to classify
Eg. Shallow auto encoders
● Deep Learning
Hierarchical representational learning
Use feature learning as one of the inner layer in a
multilayer perceptrons
15. Fuzzy Logic
● Multi valued logic
An adjective !, how pretty the girl is
● Many applications
facial pattern recognition,
air conditioners, washing machines,
antiskid braking systems, transmission systems,
vacuum cleaners,
17. Evolutionary Computing
● Choose a set of solution for a problem
● Pass them through a performance testing
(survival track)
● Best performing solutions reproduce (select
fittest)
● Add random mutation
18. What can CI take up ?
● Mundane cognitive & intellectual tasks
Like evolution, repetitive work, slow change
● Creative cognitive & intellectual tasks
Like mutation, new genesis
● CI or machines can take up mundane tasks
Remember how mechanical mundane tasks are done
by machines
19. Few Applications of CI
● Negotiation and Bargaining
● Judgmental transactions
Judgmental insurance claim settlements
● Power Grid management
● Self operated factories
● Detection Fake News
Generation is already done :-)
● Autonomous Transporting systems
● Self operated networks !!
20. Fake News
● It is lot easier to create news !
And much easier to create a fake one!!
● Fake news can create havoc
● Fake news detection needs correlation of data
from multiple source
● Looking at the sentiments
● Looking at environment/reaction
23. Autonomous transport
● Do we really need a car ?
I mean driver or pilot or captain
● Our transport systems (right from home till
destination) must be autonomous system
Err.. not like this :-)
24. Self operated networks
● Just plug in devices(equipments) and networks
must be formed
● Should provide services as per application needs
● Should identify faults in network
● Must repair faults
● Must optimize it self