You are on page 1of 17

Advances in Machine Learning

Dr.Jay B.Simha
CTO, ABIBA Systems
What is Machine learning

How can you classify two shapes based


on X and Y measurements?
Y

X
What is Machine learning

If Y < Y1 (below red line) it is ellipse,


else it is triangle :
• All triangles are classified
correctly
• 80% of ellipses are classified
correctly
Y

X
What is Machine learning

If mX + nY < some value (below black


line) it is ellipse, else it is triangle :
• All triangles are classified
correctly
• All ellipses are classified
correctly
Y

X
What is Machine learning

If Y > Y1 it is triangle
• All triangles are classified
correctly
If Y < Y1 and X < X1 it is ellipse
Y = Y1 • All ellipses are classified
correctly
If Y < Y1 and X > X1 it is square
• All squares are classified
Y

correctly
X = X1

X
What is Machine learning

Y = Y1
Y

X = X1
X
X
Linear model Non Linear model
What is Machine learning

feedback performance standard


critic percepts

ENVIRONMENT
changes
learning performance
element element actions
knowledge
learning goals

problem
generator
Machine learning - Past
X2

• Supervised: Classification.

Decision Trees Naïve Bayes

KNN SVM
X1

Multi Layer Classification


Perceptron
Perceptron
Machine learning - Past
X2

• Supervised: Regression.

Linear Regression

Neural Networks

X1
Decision Tree Regression
Regression
Machine learning - Present

• Fuzzy Modeling
• Clustering - System Identification
• TSK models - Analog ouput
• Combs models - linear growth
• Neural networks
• Auto encoders - Dimension reduction
• Deep Belief Networks
• Deep Neural networks
• Convolutional networks
• Deep teaching - GAN
Machine learning - Fuzzy Modeling

• Fuzzy Logic
• Lingistic variables
• Paralle rules
• Approximate reasoning
Machine learning - Fuzzy Modeling

• Fuzzy Logic
– Lingistic variables
– Paralle rules
– Approximate reasoning
Machine learning - Combs Fuzzy Modeling

– A AND B = A is TRUE AND B is TRUE


– A AND B = A OR B (When both are true)
– This logic can be extended to N - numbe rof
variables
– The numbe rof rules = Numbe rof Variables !!!
Machine learning - Deep Learning AE

Application

Encoder Decoder

Noisy Denoised
Latent space
Input Input
representation
Machine learning - Deep Belief Networks

Restricted
Boltzman
Machine (RBM)
DBN - Stacked RBM
Machine learning - Deep Learning CNN
THANK YOU

You might also like