Scribd
Upload
142 views3 pages

Machine Learning Paradigms

This document outlines different machine learning paradigms based on three criteria: the amount of a priori knowledge provided, the role of a teacher, and what is learned by the system. It discusses categories ranging from rote learning with no prior knowledge to neural networks with little prior knowledge. Supervised learning involves a teacher providing labeled examples, while unsupervised learning has no teacher. Systems can learn prediction, concepts, explanations, or memorize exemplar cases.

Uploaded by

nooti
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
142 views3 pages

Machine Learning Paradigms

This document outlines different machine learning paradigms based on three criteria: the amount of a priori knowledge provided, the role of a teacher, and what is learned by the system. It discusses categories ranging from rote learning with no prior knowledge to neural networks with little prior knowledge. Supervised learning involves a teacher providing labeled examples, while unsupervised learning has no teacher. Systems can learn prediction, concepts, explanations, or memorize exemplar cases.

Uploaded by

nooti
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/ 3

Machine learning paradigms

How much a priori knowledge?

• Rote learning: storing data as it is. Example: database.


• Specifying parameters: all knowledge is already avail-
able in a parametric form. Example: a function in
programming.
• Knowledge acquisition: only structure of knowledge
is known (rules, frames etc.). Example: expert sys-
tem.
• Concept Learning: given a set of examples (training
data) create a description of this set in terms of a
particular language. A priori knowledge:
– Syntax of description language (syntactical bias)
– Description of the domain from which the exam-
ples are drawn – domain knowledge or semantic
bias.
– Search algorithm to find hypotheses – search bias.
• Neural networks: no (or very little) a priori knowl-
edge.

1
The role of the teacher

• Supervised learning: the system uses a teacher.


– Concept Learning: teacher provides labeled data
(pre-classified examples) to the system.
– Reinforcement learning: teacher provides an es-
timate of the quality of system’s response to the
data (e.g. positive/negative or scaled).
• Unsupervised learning: no teacher is available to the
system.
– Clustering: partitioning or conceptual, flat or hi-
erarchical.
– Finding regularities in data: Data Mining, Knowl-
edge discovery.

2
What does the system learn?

• Prediction: learning to predict values of unknown


function.
– Classification: binary function.
– Regression: continuous-valued function.
• Concept learning: the systems acquires descriptions
of concepts.
• Explanation-based learning: using traces (explana-
tions) of correct (or incorrect) performances the sys-
tem learns rules for more efficient performance of
unseen tasks.
• Case-based (exemplar-based) learning: the system
memorizes cases (exemplars) of correctly classified
data or correct performances and learns how to use
them (e.g. by making analogies) to process unseen
data.

You might also like