You are on page 1of 8

Header1

Machine Learning (ML) Workshop to Start Your


Career as a Data Scientist

Dr. Muhammad Jawad


Assistant Professor
Department of Electrical and Computer Engineering
COMSATS University Islamabad, Lahore Campus
Land Line: +92 (111) 001-007 ext. 151
Mobile: +92 (331) 9694221
E-mail: mjawad@cuilahore.edu.pk

1 2

Outline Machine Learning Skill Track


 Machine Learning (ML) engineer is an expert on
 Python Installation, Basics, Codes and Datasets using data to training models. The models are then
 Data Preprocessing used to automate processes like image
 Regression: (Simple Linear Regression, Multiple Linear classification, speech recognition, and market
Regression, Polynomial Regression, Support Vector Machine forecasting.
(SVM), Decision Tree Regression, Random Forest Regression)
Classification: (k-NN, SVM, Kernel SVM, Naïve Bayes, Decision

Tree, Random Forest)


Introduction to ANN and CNN

Dimensionality Reduction

Principle Component Analysis

Model Selection and Boosting

3 4

3 4

Footer1 1
Header1

Market Opportunity What You can Learn Yourself After This Workshop
 Machine learning enabled solutions are being • Master Machine Learning on Python
Make accurate predictions
significantly adopted by organizations worldwide •

• Handle specific topics like Reinforcement Learning, NLP and Deep


to enhance customer experience, ROI, and to gain a Learning
competitive edge in business operations. • Apply the most advanced Data Visualization techniques with Seaborn
Moreover, in the coming years, applications of and Matplotlib
Build an army of powerful Machine Learning models and know how to
machine learning in various industry verticals is •
combine them to solve any problem
expected to rise exponentially. • Understand the underlying theory and mathematics behind Artificial
Neural Networks
• Train Artificial Neural Networks (ANNs) using back propagation and
gradient descent methods
• Apply ANNs to predict house prices given parameters such as area,
number of rooms, etc.
• Develop a fraud detection classifier using Machine Learning
Techniques
• Master Python and Scikit-Learn for Data Science and Machine Learning
5 6

5 6

Ten Examples that Excite you to Learn ML Ten Examples that Excite you to Learn ML

 Video Game Playing Device  Speech to Text or Voice to


 Facebook Face Recognition  Virtual Reality Headset
(Kinect) Text Recognition

 Use Action Recognition using ML Algorithm called


Random Forest

7 8

7 8

Footer1 2
Header1

Ten Examples that Excite you to Learn ML Ten Examples that Excite you to Learn ML

Robot Dogs (Learn Facebook Ads


 Recommender Systems
how to walk and (Know about you
behave like real dogs better than you)
using Reinforcement
Learning)
Application in Medical Science to Save People Life

9 10

9 10

Ten Examples that Excite you to Learn ML ML vs. DL vs. AI – What’s the Difference
Use in Space to Recognize To Explore Unknown
Certain Areas of the World Territories, such a MARS

11 12

11 12

Footer1 3
Header1

ML vs. DL vs. AI – What’s the Difference ML vs. DL vs. AI – What’s the Difference

13 14

13 14

ML vs. DL – What’s the Difference Regression Types


 Difference between difference regression types:
Simple Linear Regression
Multiple Linear Regression
Polynomial Regression

15 16

15 16

Footer1 4
Header1

Simple Linear Regression Example of Simple Linear Regression


 Simple Linear Regression  If we want to predict the salary of a person based on his
We predict the value of one variable ‘Y’ based on number of years of experience
another variable ‘X’. If no. of years are more, more will be the salary of the
person.
We can use a linear relationship with following Eqn.
 Why we called it Simple? 𝑌 = 𝑚𝑋 + 𝐶 OR
Because it find the relationship between two variable only. Salary = m * (#years of experience) + C
 Why we called it Linear? GOAL:
Because if the independent variable increases/decreases, Identify the best values of ‘m (slope of the line)’
then the dependent variable will increase/ decrease in linear and ‘C (y-intercept)’ that best fits the data points.
fashion. Let:

Now, this equation can be used to measure the salary of any newly hired employee

17 18

17 18

Multiple Linear Regression Polynomial Regression


 To examine the relationship between more than two  To find the relationship between independent variable ‘x’
variable. and dependent variable ‘y’ as an nth degree polynomial of
We can use a linear relationship with following Eqn. ‘x’, such as:

We have n independent variables and one dependent  For salary example, if the data looks like this:
variable.
Linear Line fitting results in terrible fit.
 Let assume same salary example and we want to add
another parameter: #years of education.

19 20

19 20

Footer1 5
Header1

Polynomial Regression ML is the Future


 If we increase the order of the equation, then we ill get  We have a cloud that shows data is every where.
much better fit.

 Complexity of the model increases (as no. of parameters


goes up) but we capture much more complex relationships.  There is a term called ‘data exhaust’ that means: “wherever
you go you leave data every where, such as text, face
book, city visit (location through GPS or mobile antenna)”.
 Data is accumulating exponentially, growing all the time.

21 22

21 22

ML is the Future ML is the Future


 Since the Dawn of time….  Now we have 1000 KB for a whole book = 1 megabyte (MB).
 Up until 2005…..

 Humans had created ….

 130 Exabytes of data

 Example:

 Letter A is taking one byte of data  Now, if we multiple it with another 1000 we get 1 Gigabyte (GB) that
 One page of book having 1000 letters will take 1 Kbyte. can fit one human genome (that take 75MB). You can also encode
whole human in 1 GB (only DNA).

23 24

23 24

Footer1 6
Header1

ML is the Future ML is the Future


 Now, if we multiply 1GB with another 1000 we get = 1 Terabyte (TB)  Now, if we multiply 1PB with another 1000 we get = 1 Exabyte (EB)
 Example: HD video camera and record a person life for 80 years
(every minute, every second) that ill fit into 1TB.

 Now, if we multiply 1TB with another 1000 we get = 1 Petabyte (PB)  From Dawn till 2005 – 130 Exabytes
 Example: Amazon Rain Forest:  Till 2010 – 1,200 Exabytes
1.4 billion acres of trees, nearly 500 trees/ acre, so a total of 700  Till 2015 – 7,900 Exabytes
billion trees in amazon rain forest. Now hypothetically, if you chop  Till 2020 – 40,900 Exabytes
all tree and convert them into papers and fill every paper with
 Future Growth Prediction:
letters are text, then you can get 1-2 petabyte.

25 26

25 26

ML is the Future Google Drive Link for Resources


 Our Capacity to process the data in the world as a data  PDF notes:
scientist is:

 Machine Learning
 Datasets and Codes:
 https://drive.google.com/drive/folders/1ndQTL-
Data Machines are using in real- onJZmyTzp5orgocCOrP3irZ5jR?usp=sharing
world
Data Scientist Capacity to process
data

 So, potential to explore this unused data is only by using


machine learning algorithms to step up to that challenge.
 Therefore, what you will see, will be the cutting edge in this
field of data science.

27 28

27 28

Footer1 7
Header1

Presentation of ML folder, Collaboratory, Jupyter


notebook and Spyder

29

29

Footer1 8

You might also like