Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 13

Lecture 1: Building the Framework

Introduction to
Machine Learning
“Everything should be made as simple as possible, but not simpler”
Machine Learning : Seems to be all around us
What is Machine Learning ? Why study it ?

Is this machine
learning?

Input Output
(3 + 4) (7)
What is Machine Learning ? Why study it ? Is this machine
learning?

IF ( MSG CONTAINS
“OFFER”) { SPAM } ELSE
{INBOX}
What is Machine Learning ? Why study it ?
Machine Learning is man’s pursuit of transferring his own (learning) abilities
to machine using science and technology through the language of Math
How do we exactly transfer this learning ability ?
“A smart person learns from his mistakes (PAST) , but a truly wise person learns from the mistakes
of others”

DAY, TIME,
WEATHER,
CITY

WORDS IN
MAIL - OFFER,
SUBJECT

DETECTION OF
EYES, NOSE,
MOUTH

DIFFERENT
A
DAT MEDICAL
PARAMETERS-
BLOOD,SUGAR
Building the ML Framework

IMAGES

TABULAR
TEXT
DATA

AUDIO
DAY, TIME,
WEATHER, CITY

WORDS IN MAIL
- OFFER,
SUBJECT

DETECTION OF
EYES, NOSE, Usually the physics of the phenomena is not
known to us particularly well OR It is hard to
MOUTH

DIFFERENT formalize the learning process


MEDICAL
PARAMETERS-
BLOOD,SUGAR
Building the ML Framework: Next?

Well how do we make the


machine learn the
phenomena using the data

Learning
function
function
Input
Machine Learning Program
Computer Program Input Code function
Output

Output Machine
Code
A simple example: Have we been doing
Machine Learning since a long time ?

Will India win the How long will it take me


How many runs will next match ? to reach Westend
Virat Kohli score in the Mall ?
next match ?

How many runs will


Zak Crawley score in Will it rain today or not ?
the next match ?
Who are the top 10
batsmen in the
world ?
Will this movie be a
hit or not ?
Building the ML Framework: Run machine VK

30, 40, 100, 0, 10, ?

Methods for predicting the next score

M1 : Take average of last two innings

M2 : Take average till the latest innings

M3 : { 1. Add 10 ; 2. Add 60; {if score>=100; set


0;} repeat }

M4 : Predict the last score


Building the ML Framework: Asking the right questions

What should be output : the average,


median, 6th quantile,etc. ?

Different error metrics may have different


optimal solutions

Its most important to FIRST fix the error


function
The BareBone Framework : Learnings from VK

Feature
Engineering
Our MAGIC bag of
FUNCTIONS :
Naive Bayes
Random Forests
Neural Network
...
Thank You !!!

“Everything should be made as simple as possible, but not simpler”

You might also like