Scribd
Upload
116 views2 pages

Machine Learning With R - Chapter 1

This chapter introduces machine learning concepts. It discusses the learning process of moving from data to abstractions or concepts to generalization. Abstractions summarize data and can take the form of mathematical equations, diagrams, or rules. Models assume a structure in data can be uncovered. Generalization uses heuristics to generate results. Evaluation provides feedback to improve. The chapter also covers types of input data like structured vs unstructured, and machine learning algorithms like predictive models that estimate relationships to targets and descriptive models that summarize data patterns.

Uploaded by

ljing4
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
Download as txt, pdf, or txt
116 views2 pages

Machine Learning With R - Chapter 1

This chapter introduces machine learning concepts. It discusses the learning process of moving from data to abstractions or concepts to generalization. Abstractions summarize data and can take the form of mathematical equations, diagrams, or rules. Models assume a structure in data can be uncovered. Generalization uses heuristics to generate results. Evaluation provides feedback to improve. The chapter also covers types of input data like structured vs unstructured, and machine learning algorithms like predictive models that estimate relationships to targets and descriptive models that summarize data patterns.

Uploaded by

ljing4
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
Download as txt, pdf, or txt
Download as txt, pdf, or txt
You are on page 1/ 2

——————————————————————-

Machine Learning with R


——————————————————————-

CHAPTER 1 - INTRODUCING MACHINE LEARNING

*Learning Process

Data Storage Abstraction Generalization

Data -> Concepts -> Interferences -> Evaluation

# Example: Buying food (data) -> read recipe (concepts) -> cook meal
(generalization)

*Abstractions or Concepts

- Abstractions summarizes information collected from data

- Abstraction for humans are for example „pictures of pipes“ that represent a
real pipe.

- Abstraction for computers (but also for humans!) are mathematical


equations, relational diagrams, if/else rules, grouping of data as clusters.
Those are also called models.

- A model is not new data, but it assumes that there is a structure in the
data that can be unveiled.

# Example: You take the Covid data set, in order to see how the cases
are developing internationally on a daily scale. The graphs are the
concept.

*Generalization

- Heuristics are approximations used by the computer to generate a result or


action. For people those mental heuristics are the „gut-instincts“.

- People use availability heuristics, which is the estimate of likelihood of


an event based on how easily examples can be recalled (airplane vs. car
crash)

# Example: From a lot of graphs, you choose those which are the most
meaningful to you or most representative of the development.

*Evaluation

# Example: Let Hans or the BI Meeting members evaluate your work. It could
give you valuable insights and help you grow.

*Types of Input Data

- People are more comfortable consuming unstructured data


- Computers need structured data (similar examples have the same feature)

# Example: When listening to music you associate different emotions with the
sound. But when you have to process 10.000 hands from poker, you feel quite
lost at how the play style of someone differs from yours.

*Types of machine learning algorithms

- Predictive Model focus on a target feature and tries to see its


relationship to others. It does not necessarily need to predict futures. It
can also estimate the past (when was a baby born) or change gears (control
traffic). It is also called supervised learning, but it has nothing to do
with human supervision. The machine can „supervise“ or validate its results.

- Descriptive Model focus on all features, so that no feature is more


important than the others. summarizing datasets in a new or more
interesting way. It is more used in pattern discovery to identify association
within the data

- meta learners learn how to learn. Result of past learning enhance


additional learning

You might also like