Introduction to Machine Learning with Find-S

29
Introduction to Machine Learning with Find-S By: Girish Ch. Bharti Sr. Software Consultant Knoldus Software LLP

Transcript of Introduction to Machine Learning with Find-S

Page 1: Introduction to Machine Learning with Find-S

Introduction to MachineLearning with Find-S

By: Girish Ch. Bharti

Sr. Software Consultant Knoldus Software LLP

Page 2: Introduction to Machine Learning with Find-S

Agenda

● A brief about machine learning

● What is inductive learning?

● What is concept learning

● Inductive bias

● Conjunctive concepts

● Find-S Algo

● Demo

Page 3: Introduction to Machine Learning with Find-S

Machine learning

Page 4: Introduction to Machine Learning with Find-S

Machine learning

Page 5: Introduction to Machine Learning with Find-S

Machine learning, AI & Deep learning

Page 6: Introduction to Machine Learning with Find-S

Applications of Machine learning

Page 7: Introduction to Machine Learning with Find-S

Machine learning (2017)

Page 8: Introduction to Machine Learning with Find-S

Machine learning workflow

Page 9: Introduction to Machine Learning with Find-S

Learning

● What is learning?

● Types of learning

● Passive learning (From teachers)Rote Learning (Memorization)Inductive learning (Learning through examples)

Page 10: Introduction to Machine Learning with Find-S

What is inductive learning?

● Learning by examples

● Ex:

-> Rote Learning:

2,4,6,8,10,12,14,16,18,20

-> Inductive learning:

2 * 1 = 2

2 * 2 = 4

2 * 3 = 6

2 * 4 = ?

Page 11: Introduction to Machine Learning with Find-S

First machine learning algo(Checkers , Early 60s)

Page 12: Introduction to Machine Learning with Find-S

What is machine learning?

Examples ->

Experience ->

Prediction

Page 13: Introduction to Machine Learning with Find-S

Machine learning problems● Identify apple and orange

Page 14: Introduction to Machine Learning with Find-S

Machine learning

● Features:

➢ Shape

➢ Color

➢ Taste

➢ Hardness

----------------------------

Weight?

Page 15: Introduction to Machine Learning with Find-S

Notion of data

● Data

● Learning

● Predictions

Page 16: Introduction to Machine Learning with Find-S

What is concept learning?

● “Problem of searching through a predefined space of potential hypotheses for the hypothesis that best fits the training examples.”

Page 17: Introduction to Machine Learning with Find-S

Concept Learning

Concept Data Object

C X Label

x -Belogs to a concept c 1, +1, true

-Not belogs to a concept 0, -1, false

Note: Concept is not tied to data

Page 18: Introduction to Machine Learning with Find-S

Concept Learning

● Data objectx = ( circular, large, smooth, dark)

● Concept SpaceX = All possible values of feature valuesSize: => (No of features) power (possible values)=> 4 * 4 = 16

● Features

● Shape –> oval, circular

● Size -> large, small

● Color -> dark, light

● Surface -> smooth, irregular

Page 19: Introduction to Machine Learning with Find-S

Concept space

Page 20: Introduction to Machine Learning with Find-S

Conjuctive Concept

● X -> (cir, sm, dark, irr)

● X1 -> (1, 0, 1, 0)

● X2 -> (0, 0, 1, 1)

----------------------- h -> (0, 0, 1, 0)

● Circular: 1, Oval: 0

● Large: 1, Small: 0

● Dark: 1, Light: 0

● Smooth: 1, Irregular: 0

Page 21: Introduction to Machine Learning with Find-S

Conjuctive concept

● 1. Circular ^ Smooth

● 2. Large ^ Light ^ Irregular

● 3. Circular ^ Large ^ Dark ^ Smooth (Will lable only one as positive)

● 4. Oval (Will label 8 as positive)

Page 22: Introduction to Machine Learning with Find-S

Notations

● Circular ^ ? ^ ? ^ Smooth ---- (People in middle)

● ? ^ ? ^ ? ^ ? ---- (Accepts all)

● ⵁ (Fi) ---- (Rejects All)

● Total number of conjuctive concepts:

(3 * 3 * 3 * 3) + 1 = 82

If we have d binary features: ((3 power d) + 1 ) => Conjuctive concepts

Page 23: Introduction to Machine Learning with Find-S

Finding malignant tumor from MRI scan

Page 24: Introduction to Machine Learning with Find-S

Finding malignant tumor from MRI scan

● Attributes:

➢ Shape -> Circular, Oval

➢ Size -> Large, Small

➢ Color -> Light, Dark

➢ Surface -> Smooth, Irregular

● Concept: Malignent Tumor

Page 25: Introduction to Machine Learning with Find-S

Find S Algorighm

1) Start with h = ⵁ

2) Use next input {x, c(x)}

3) If c(x) = 0, go to step 2

4) h <- h ⴷ x (Pairwise AND)

5) If more examples : Go to step 2

6) Stop

Page 26: Introduction to Machine Learning with Find-S

Disadvantages of Find S

● Not accurate

● Do not consider negative examples

● Error prone

● Does not cove desjunctive concepts

Page 28: Introduction to Machine Learning with Find-S

References

● https://blog.knoldus.com/2017/08/16/concept-learning-the-stepping-stone-towards-machine-learning-with-find-s/

● https://blog.knoldus.com/2017/11/01/concept-learning-find-s-implementation-with-scala/

● Wikipedia

● Tom Michel

Page 29: Introduction to Machine Learning with Find-S

Thanks