Selected Newsgroup Message

Dejanews Thread

From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: Methods of machine learning ?
Date: 18 Nov 1998 00:00:00 GMT
Message-ID: <3652f8fc.0@news3.ibm.net>
References: <72s716$tu8$1@nnrp1.dejanews.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 18 Nov 1998 16:42:36 GMT, 200.229.240.55
Organization: SilWis
Newsgroups: comp.ai

jazf@my-dejanews.com wrote in message <72s716$tu8$1@nnrp1.dejanews.com>...
>I am currently researching a report on the developmentof artificial
>intelligence.  I am looking for a summary of themethods and techniques used
>in machine learning. Cananyone assist ?
>

You didn't say the size of the summary you wanted, so I thought you may
like this very reduced, quick-and-dirty one:

Concept Learning
From sets of positive and negative examples, derive general categories
Examples: Candidate-elimination algorithm,
          Unbiased Inductive Learner

Decision Trees
Methods to approximate functions over discrete values with good
resistance to noise and able to learn disjunctive expressions
Examples: ID3, C4.5

Artificial Neural Networks
Learning of real-valued or discrete-valued functions from examples
Also vector-valued functions.
Examples: Perceptrons, Feedforward, Backpropagation, Recurrent Nets, etc

Bayesian Learning
Probabilistic approach inferring distributions from observed data
Examples: Brute-force Bayes Concept learning,
          Gibbs algorithm, naive Bayes Classifier,
          Bayesian Belief Network

Instance-Based learning
Training examples are stored and generalization occurs only when
new instances must be classified
Examples: k-Nearest Neighbor, Locally weighted regression,
          Radial Basis Functions (also a method of ANNs)

Learning Sets of Rules
Procedures for learning sets of if-then rules
Examples: FOIL, FOCL, ILP

Reinforcement Learning
Learning is done by performing actions in one environment and
receiving, usually by a trainer, a reward or punishment based
on the appropriateness of the action taken.

Other Methods
Genetic Algorithms: Learning based on simulation of evolutionary processes
EBL: Explanation-Based Learning
KBANN: Knowledge-Based Artificial Neural Networks
and others...

Sergio Navega.


Back to Menu of Messages               Sergio's Homepage

Any Comments? snavega@attglobal.net