Learning Prototypical Concept Descriptions
نویسندگان
چکیده
We describe a new representation for learning concepts that di ers from the traditional decision tree and rule approach. This representation, called prototypical concept descriptions, can represent several prototypes for a concept. We also describe PL, our algorithm for learning these prototypes, and demonstrate that prototypical concept descriptions can, in some situations, classify more accurately than standard Machine Learning algorithms. More importantly, we show that they yield more stable descriptions when applied in noisy and dynamic situations.
منابع مشابه
Family patterns of reaction to a child with a learning disability: a mediational perspective.
The impact of a child's learning disability on the family system is explored with regard to the general consequences for the system and to the parents' abilities to react to the problems generated for the child and family. Feuerstein's concept of mediated learning experience is proposed as a useful way of understanding the process issues, to assess parental reactions, and to guide more producti...
متن کاملHealth System Resilience: What Are We Talking About? A Scoping Review Mapping Characteristics and Keywords
Background Health systems are based on 6 functions that need to work together at all times to effectively deliver safe and quality health services. These functions are vulnerable to shocks and changes; if a health system is unable to withstand the pressure from a shock, it may cease to function or collapse. The concept of resilience has been introduced with the goal of strengthening healt...
متن کاملHYDRA-MM: Learning Multiple Descriptions to Improve Classi cation Accuracy
For learning tasks with few examples, greater classi cation accuracy can be achieved by learning several concept descriptions for each class in the data and producing a classi cation that combines evidence from multiple descriptions. Stochastic (randomized) search can be used to generate many concept descriptions for each class. Here we use a tractable approximation to the optimal Bayesian meth...
متن کاملOn Learning Multiple Descriptions of a Concept
In sparse data environments, greater classi cation accuracy can be achieved by learning several concept descriptions of the data and combining their classi cations. Stochastic search is a general tool which can be used to generate many good concept descriptions (rule sets) for each class in the data. Bayesian probability theory o ers an optimal strategy for combining classi cations of the indiv...
متن کاملHydra-mm: Learning Multiple Descriptions to Improve Classification Accuracy
For learning tasks with few examples, greater classi cation accuracy can be achieved by learning several concept descriptions for each class in the data and producing a classi cation that combines evidence from multiple descriptions. Stochastic (randomized) search can be used to generate many concept descriptions for each class. Here we use a tractable approximation to the optimal Bayesian meth...
متن کامل