Unsupervised Information Extraction and Retrieval

Unsupervised Characteristics

Basic model of unsupervised learning is:

• Hebb's rule: The connectivity between two connected neurons is greater if both are activated and less if they are not.

• Side interaction: Every neuron has an excited connection with near neighbours and inhibited with the far ones, this model imitates the brain.

Learning and training is based on the next concepts:

• Self-adaptive: the algorithm gets adaptation through the modification of its parameters to solve the problem.

• Competition: Based in encouraging the winners against the losers.

Advantages and disadvantages of unsupervised learning (UL) are:

• UL may find an innovative solution.

• UL may not find a good solution.

• It is difficult to create models and select the parameters.

This site is developed for the assignement Information and Retrieval Organization of the subject Information Retrieval and Access over Computer Engineering of Carlos III University of Madrid.

Available versions:

Download the DOC article

Download the PDF article