Principle of parsimony: Simplest possible explanation

Defining Doagnosis

Diagnosis: To determins what is wrong with a malfunctioning device or the mapping of the data space to the hypothesis space

Data and Hypothesis Spaces

data spaces

Heuristic Classification

Mapping Data to Hypothesis

Mapping needs to take place from data-space to hypothesis space, and additionally from the hypothesis space to the data space


1) One data point, multiple hypothesis 2) One hypothesis may explain multiple data points 3) Multiple hypothesis multiple sets of data 4) Mutually exclusive hypothesis 5) Interacting data points (data points cancelling each other out)

Two Views of Diagnosis


Mapping data onto hypothsis


Rule: Cause -> Effect

Rule: Bob hates Joe Cause: Bob walks in Effect: Joe leaves

Fundamental Forms of Inference

Choosing a Hypothesis

1) Hypothesis must cover as much of the data as possible 2) The smallest number of hypothesis out to be used, Principle of parsimony 3) Some hypothesis may be more likely than others