Instance-based learning: Difference between revisions

Revision as of 18:46, 16 June 2017

Latest revision as of 17:47, 25 June 2025

(11 intermediate revisions by 8 users not shown)

Line 1:

In [[machine learning]], '''instance-based learning''' (sometimes called '''memory-based learning'''<ref>{{cite book |author1=Walter Daelemans |authorlink1=Walter Daelemans |author2=Antal van den Bosch |authorlink2=Antal van den Bosch |year=2005 |title=Memory-Based Language Processing |publisher=Cambridge University Press}}</ref>) is a family of learning algorithms that, instead of performing explicit generalization, comparescompare new problem instances with instances seen in training, which have been stored in memory. Because computation is postponed until a new instance is observed, these algorithms are sometimes referred to as "lazy."<ref name="mitchell"></ref>

It is called instance-based because it constructs hypotheses directly from the training instances themselves.<ref name='aima733'>[[Stuart J. Russell|Stuart Russell]] and [[Peter Norvig]] (2003). ''[[Artificial Intelligence: A Modern Approach]]'', second edition, p. 733. Prentice Hall. {{ISBN|0-13-080302-2}}</ref>

This means that the hypothesis complexity can grow with the data:<ref name='aima733'/> in the worst case, a hypothesis is a list of ''n'' training items and the computational complexity of [[Classification (machine learning)|classifying]] a single new instance is [[Big O notation|''O'']](''n''). One advantage that instance-based learning has over other methods of machine learning is its ability to adapt its model to previously unseen data. Instance-based learners may simply store a new instance or throw an old instance away.

Examples of instance-based learning algorithmalgorithms are the [[k-nearest neighborneighbors algorithm|''k''-nearest neighborneighbors algorithm]], [[kernel method|kernel machines]] and [[Radial basis function network|RBF networks]].<ref name="mitchell">{{cite book |author=Tom Mitchell |title=Machine Learning |year=1997 |publisher=McGraw-Hill}}</ref>{{rp|ch. 8}} These store (a subset of) their training set; when predicting a value/class for a new instance, they compute distances or similarities between this instance and the training instances to make a decision.

To battle the memory complexity of storing all training instances, as well as the risk of [[overfitting]] to noise in the training set, ''instance reduction'' algorithms have been proposed.<ref>{{cite journal |title=Reduction techniques for instance-based learning algorithms |author1=D. Randall Wilson |author2=Tony R. Martinez |journal=[[Machine Learning (journal)|Machine Learning]] |publisher=Kluwer |year=2000}}</ref>

Gagliardi<ref name=Gagliardi2011>{{cite journal|last=Gagliardi|first=F|title=Instance-based classifiers applied to medical databases: Diagnosis and knowledge extraction|journal=Artificial Intelligence in Medicine|year=2011|volume=52|issue=3|pages=123–139|doi=10.1016/j.artmed.2011.04.002|url=http://dx.doi.org/10.1016/j.artmed.2011.04.002}}</ref> applies this family of classifiers in medical field as second-opinion [[Clinical decision support system|diagnostic tools]] and as tools for the knowledge extraction phase in the process of [[knowledge discovery in databases]].

One of these classifiers (called ''Prototype exemplar learning classifier'' ([[PEL-C]]) is able to extract a mixture of abstracted prototypical cases (that are [[syndrome]]s) and selected atypical clinical cases.

==See also==

Line 19 ⟶ 17:

[[Category:Machine learning]]

{{AI-stub}}

{{machine-learning-stub}}