Rao, R. B., Greiner, R., and Hancock, T. (1994). Exploiting the absence of irrelevant information. In AAAI Fall Symposium on `Relevance', New Orleans. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Learning to Classify Incomplete Examples - Schuurmans, Greiner (1993) (6 citations) Self-citation (Greiner) (Correct)
....of an attribute is correlated with its value. For example ex inmates are less likely to answer questions of the form have you ever been in prison . Also, in medical databases missing values may actually provide clues about the subsequent classification (Porter, Bareiss and Holte, 1990; Rao, Greiner and Hancock, 1994). Here, this arbitrary blocking model is able to capture correlations between hidden attributes and their values, other attributes, or even concept membership. We define a learning context (fi; by the type of blocking process fi, and type of training examples , and consider various contexts ....
....problems. For example, missing attribute values in medical databases typically provide useful information namely that the missing attributes are irrelevant to the classification given the known attributes (Porter, Bareiss and Holte, 1990) which could be exploited by a learning system (Rao, Greiner and Hancock, 1994). Notice that fi I is overly restrictive and fi A is too underconstrained to adequately model this situation. We are currently investigating alternative blocking models that (we hope) lead to better empirical learning performance in such domains. Other interesting research directions involve ....
Rao, R. B., Greiner, R., and Hancock, T. (1994). Exploiting the absence of irrelevant information. In AAAI Fall Symposium on `Relevance', New Orleans.
Research Summary - Greiner Self-citation (Greiner) (Correct)
....run, along with the diagnosis reached leaving the other attribute values blank. Afterwards, the learner s task is to reconstruct that tree, using only these sparsely filled records. Here, the omitted values are omitted because their values are irrelevant to the classification. The papers [22, 11, 12] prove that that this (ir)relevance information significantly simplifies the learning task, as it allows a learner to probably approximately correct PAC learn arbitrary decision trees, or even DNF formulae two well studied classes not known to be PAC learnable in the standard model We ....
R. Bharat Rao, Russell Greiner, and Thomas Hancock. Exploiting the absence of irrelevant information. In AAAI Fall Symposium on `Relevance', New Orleans, 1994. ($GreinerFTP/superfluous.ps).
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC