Sie sind auf Seite 1von 40

+

Machine Learning

Malik Jahan Khan


+
Why to take this course?

 By the end of the course, you should be able to know:

 Basic and intermediate level theoretical concepts

 Associated algorithms

 Implementation of the algorithms

 Potential application domains

 Basics of research happening in this field

Panoramic field of study in CS!


+
Theoretical concepts

 Basic mathematics (calculus / linear algebra / geometry,


probability)

 Design of algorithms (and their complexity)

 Intermediate level concepts involving these basics

 Supervised learning

 Unsupervised learning

 Semi-supervised learning
+
Algorithms

 Regression

 Classifiers
 Classical algorithms
 Decision trees
 CBR
 Neural networks
 Ensemble of classifiers

 Clustering
 Classical algorithms
 Density-based clustering
+
Implementation

 Mostly in Python
 TensorFlow
+
Application Domains

 Disease diagnosis in crops / plants

 Soil analysis to determine the nutrient deficiency

 Aerial image analysis to estimate agriculture yield

 Handwritten character recognition

 Diagnostics in variety of domains

 Planning / advisory services

 Predictions / forecasts
 Weather
 Financials
 Software engineering

 …
+
Research

 Identify high quality venues of ML research


 Conferences (ICML, NIPS, ICONIP...)
 Journals (PAMI, Applied Soft Computing, KBS...)

 Review literature

 Learn how to identify problems

 Learn how to address the identified problems

 Learn how to report your work

 Learn how to validate your work


+
Grading

 Three projects: 30%

 Midterm exam: 30%

 Final exam: 40%


+
Text Material

 Please see the course outline over LMS

 Some books will be shared over LMS

 Relevant readings will be shared over LMS


+
Learning Process

 Learner

 Trainer

 How do you learn a new concept?


 How to solve a set of equations?
 Who is the most popular professor: Dr. Israr, Dr. Noman or Dr.
Adnan?
+
+
+
+
+
+
+
+ Linear Regression

Supervised Learning
+
+
+

Gradient Descent
(Least Square)
+
+
+
+
+
+
+ Classification

Is it supervised learning?
+
+
+

Is it supervised learning?
+
What does this image contain?
+
What does this image contain?
+
What does this image contain?
+
What does this image contain?

Clustering: Unsupervised Learning


+
Nature of Available Data

 Features

 Will you take Machine Learning course as elective?


 YES
 NO

 What will matter in making the right decision?


 Content
 Alternatives
 Previous experience with the teacher
 Previous reviews of last year offering (N/A in this case)
 Workload
 Matching with your future plans…
+
How does ML fit in “Theory of
Computation”
 Solvable Problems vs Unsolvable Problems

 Easy vs Difficult Solvable Problems

 Exact Solutions vs Approximate Solutions

 What makes a solution attractive?


 Efficiency
 Correctness
 Which one would you prefer: efficient solution or correct solution?

 Depends on?
+
+
+
Example Projects

 Umair / Nishat

 Aaqib

 Fahad / Ammar

 Zeeshan / Waqar