Scribd
Upload
47 views39 pages

Introduction To Machine Learning: Pekka Parviainen

This document provides an introduction and agenda for a machine learning lecture. The key points are: - The lecture will cover the basic concepts of machine learning and how to apply, implement, and evaluate machine learning algorithms. - The goals are for students to understand machine learning approaches, develop and apply algorithms to real data, and have an overview of machine learning uses. - Topics will include supervised learning, unsupervised learning, reinforcement learning, and modeling techniques like linear models, decision trees, and neural networks. - The agenda outlines lectures, exercises, projects, exams, and prerequisites to help students learn machine learning.

Uploaded by

Val
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
47 views39 pages

Introduction To Machine Learning: Pekka Parviainen

This document provides an introduction and agenda for a machine learning lecture. The key points are: - The lecture will cover the basic concepts of machine learning and how to apply, implement, and evaluate machine learning algorithms. - The goals are for students to understand machine learning approaches, develop and apply algorithms to real data, and have an overview of machine learning uses. - Topics will include supervised learning, unsupervised learning, reinforcement learning, and modeling techniques like linear models, decision trees, and neural networks. - The agenda outlines lectures, exercises, projects, exams, and prerequisites to help students learn machine learning.

Uploaded by

Val
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 39

Introduction to Machine Learning

Lecture 1

Pekka Parviainen

University of Bergen

23.8.2021
Programming task
I Task: write a program that recognizes handwritten digits
I How would you write the program specification? What is your
time estimate for this task? How many lines of code?
Machine learning paradigm
Machine learning approach

I Collect labelled examples and let the machine figure out the
rules
Today’s Agenda

I Practical matters
I Introduction to machine learning
Learning goals

At the end of the course the student should:


I understand the basic ideas of machine learning
I be able to compare modeling aspects of various machine
learning approaches
I develop and implement machine learning algorithms
I apply and evaluate machine learning algorithms on real data
sets
I have a good overview of how machine learning can be used in
different contexts in the society
Course team

Lecturer TA

Pekka Parviainen Pierre Gillot


Group leaders

Jacob Hay Eskil Hamre Isaksen Johanna Jøsang

Per Mjelde Magnus Øian


Lectures

I To give an overview of the topics


I On Monday 14:15-16:00 and Tuesday 10:15-12:00 via Zoom
I Check mittUiB for the lecture weeks
Weekly exercises

I To deepen understanding of the topics


I Initially 7 sessions (several groups may be cancelled due to
low attendance)
I Format: problem solving in small groups
I Voluntary
I If you have any symptoms of any sickness, do not go to a
physical exercise session
I Homework
I Mostly programming tasks
I Voluntary (returned via MittUiB to get feedback)
Projects

I To practice implementation of machine learning algorithms


and applying machine learning in real-life scenarios
I Deadlines: 24.9. and 22.10.
I Mandatory work
I Can be done in pairs
I Advice available in the exercise sessions
Exam

I Thursday, November 25th


I Check exact time and place two weeks before the exam
Learning material

I No course book
I Unless stated otherwise, you are expected master all topics
covered on lectures, exercises, and projects.
I Pointers to additional online material will be given
I Books that might be useful (available online):
I Bishop: Pattern recognition and machine learning
I Hastie et al.: The elements of statistical learning
Grading

I Exam 50%
I Projects (2) 50%
I To pass the course you have to
I pass the exam and
I pass each of the projects
Prerequisites

I Basic programming skills


I INF102 or equivalent
I All learning material will be in Python
I A Python crash course on Thursday 26.8. at 14:15 at
Allegaten 66, Auditorium B (All participants need to
register beforehand)
Prerequisites

I Linear algebra
I Eigenvalues and vectors
I Matrix inversion
I Calculus
I Derivatives (Gradient)
I Finding minima/maxima
I Probability and statistics
I Quiz about mathematical prerequisites in MittUiB
What will you learn during this course?

I Formulate real-world problems as machine learning problems


I Some basic machine learning tools and how to choose a
proper one for the problem at hand
I How to evaluate a machine learning solution
I Programming + mathematical understanding
Background poll
What is machine learning?

I “Field of study that gives computers the ability to learn


without being explicitly programmed” (Arthur Samuel, 1959)
I A well-defined machine learning problem (Mitchell):
I A computer program is said to learn from experience E with
respect to some class of tasks T and performance measure P,
if its performance in tasks in T , as measured by P, improves
with experience E
I Three important concepts:
I Task: What is the problem that the program is solving?
I Performance measure: How is performance of the program
(when solving the given task) measured?
I Experience: What is the data (examples) that the program is
using to improve its performance?
Digit recognition revisited

The postal office wants to develop an automatic system for sorting


letters. As a subroutine, we need a program that recognizes
handwritten digits.
I Task: Given a 28 × 28 pixel grayscale image of a handwritten
digit, return the label (digit)
I Performance measure: Proportion of correctly labelled
examples in a new data set
I Experience: Pairs of images and labels
Subfields of machine learning

I Supervised learning
I Unsupervised learning
I Reinforcement learning
Supervised learning

I Task: predict labels of new, unseen objects


I Performance measure: e.g., prediction accuracy
I Experience: object-label pairs
Let’s practice

I Task: Classify images into two classes (sininen and punainen)


I Performance measure: Proportion of correctly classified
images on unseen data
I Training data: Labeled images
Training data

Sininen Sininen Sininen

Punainen Punainen Punainen


Test data

I Guess the label:


Test data

I Guess the label:


Unsupervised learning
I Learn a hidden structure from unlabeled data
I Typical tasks:
I Clustering
I Dimensionality reduction
Let’s practice
I Clustering: partition objects into groups based on their
similarity

(a) (b) (c) (d)

(e) (f) (g) (h)


Reinforcement learning

I An agent tries to maximize a reward by taking actions in an


environment
I No object/label pairs, only feedback from the actions

I For example, chess AI


Machine learning in practice: Learning a classifier

I Goal: accurate predictions on new objects (generalization)


I Workflow
I Collect data
I Choose candidate models
I Learn
I Evaluate
Collect data

I Task: Based on coordinates, try to guess whether a point is a


red circle or blue triangle
Choose candidate models
Learn

I Find a decision boundary


I Decision boundary is boundary between regions corresponding
to different classes
Evaluate

I Validation data (unseen examples)


Evaluate

I Too complex models tend to overfit


Learning = Representation + evaluation + optimization

I Representation
I What is the set of candidate models (hypothesis space)?
I Evaluation
I How to compare different models?
I Optimization
I How to find a good model?
Machine learning vs. other fields

I Artificial intelligence: ”study of intelligent agents”


I Machine learning is typically classified as a subfield of artificial
intelligence
I Lots (most?) of the recent success of AI can be contributed to
machine learning methods
I Statistics
I Use lots of the same methods
I Many machine learning methods are originated from statistics
I Statistics has branched off from mathematics, machine
learning from computer science
I Statistics more concerned on theoretical aspects
I Machine learning more practically oriented
I Hard to draw a line between them
I For example, the department of mathematics has a course
“Statistical learning” which covers most of the topics of this
course
Course topics

I Supervised learning
I Linear models
I Decision trees
I Nearest neighbors
I Neural networks (including deep learning)
I Support vector machines
I Probabilistic methods
I Ensemble learning
I Unsupervised learning
I Dimensionality reduction
I Clustering
I Model selection and evaluation
Summary

I Machine learning is about programs that improve their


performance based on data
I Goal is to generalize
I Several types of learning
I Supervised learning
I Unsupervised learning
I Reinforcement learning
Further readings

I A Few Useful Things to Know about Machine Learning


https://homes.cs.washington.edu/~pedrod/papers/
cacm12.pdf

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy