WS2014/2015, 10.12.2014
Intelligent Systems
Introduction to Machine Learning
Carsten Rother, Dmitrij Schlesinger
10.12.2014
Intelligent Systems: Introduction 2
Machine Learning everywhere
Recommender Systems
Slide credits: Bernt Schiele, Stefan Roth, Peter Gehler …
10.12.2014
Intelligent Systems: Introduction 3
Machine Learning everywhere
AdWords predictions
10.12.2014
Intelligent Systems: Introduction 4
Machine Learning everywhere
Ranking web queries
10.12.2014
Intelligent Systems: Introduction 5
Machine Learning everywhere
Image segmentation
10.12.2014
Intelligent Systems: Introduction 6
Machine Learning everywhere
Body part detection on Kinect
10.12.2014
Intelligent Systems: Introduction 7
What is Machine Learning
Semantics and Structure Database of Text
Images and Videos
Symbolic Data and Measurements 2D, 3D microscopy
y = f(x; w)
10.12.2014
Intelligent Systems: Introduction 8
What is Machine Learning
Goal of machine learning:
Machines that learn to perform a task from experience We can formalize this as:
is called output variable, the input variable and
the model parameters
learn... adjust the parameter
… to perform …
... a task ... the function
... from experience using a training dataset , where of either or
y = f(x; w)
x y
w
w
f
L = (x1, y1), (x2, y2) . . . (xl, yl) L = (x1, x2 . . . xl) L
10.12.2014 Intelligent Systems: Introduction
Performance: “99% correct classification”
- of what ?
- i.e. on speech recognition (correct words, characters, speakers identification etc.)
- over which dataset ?
- remember on “precision/recall” from the previous lectures
9
What does it mean “to perform” ?
“The car drives without human intervention 99%
of the time on country roads”
Is 99% good enough ?
1% false alarm for 300.000 passengers at airport Frankfurt ?
10.12.2014 Intelligent Systems: Introduction
• Supervised learning: there is a completely labelled (annotated) dataset
• Unsupervised learning: no annotations at all
• Reinforcement learning: there is a reward for correct action/recognition Some variants:
• Semi-Supervised learning: dataset is partially labelled
• Transductive learning: no model, just fund the “true” answer
• On-line learning: the training samples are not available at a time
• Large-scale learning: lots of data
• Active learning: how to ask a “teacher” for an annotation ?
10
Different learning scenarios
10.12.2014 Intelligent Systems: Introduction
Training:
training data Testing:
Testing data (different from training)
11
General Paradigm (supervised learning)
L = (x1, y1), (x2, y2) . . . (xl, yl)
) Unknown parameter ✓ Learn
) Predict
0,1,2 … prediction
10.12.2014 Intelligent Systems: Introduction
Assume, it is necessary to place a (predefined) number of radio stations in order to supply the signal of the best quality.
We should:
• Assign each house to a stations
• find the best station positions (in order to minimize e.g. the total distance)
If the assignments are known, it would be a supervised learning. Now however, we have to find both classification (house assignments) and unknown parameters (station positions) simultaneously.
12
Unsupervised learning
10.12.2014 Intelligent Systems: Introduction
Given a situation find an action to maximize a reward function No optimal output action is presented
Each action yields a reward (which may have to be approximated) Exploration: try out new actions
Exploitation: use known actions that yield high rewards
Find a good trade-off between exploration and exploitation
Examples: learning robot motor primitives: ball in a cup, elevator scheduling, Playing Backgammon/Go …
13
Reinforcement Learning
10.12.2014 Intelligent Systems: Introduction
10.12: Introduction to Machine Learning, Probability Theory 17.12: Decision Making, Statistical Learning (Schlesinger) 07.01: Directed Graphical Models (Rother)
14.01: Undirected Graphical Models (Rother) 21.01: Neural Networks (Schlesinger)
28.01: Clustering (Rother) 04.02: Wrap-up (Rother)
First exercise: 15.12 about Probability Theory (Heidrich, Groth)
Course homepage: http://www.inf.tu-dresden.de/index.php?node_id=3486&ln=de
14