# Distributions Relations Probability Distributions Probability distributions are an

• Slides: 11

Distributions & Relations

Probability Distributions Probability distributions are an important source of technical knowledge. They need not follow simple forms like Normal or Poisson.

Distribution Learn Run The network structure supporting a distribution on the variable Intensity. The same structure supports learning and activation - the distribution is filled by setting its state to Learn, activated by setting its state to Run - the direction of information flow reverses.

Contents of Distribution The bins are ranges, which can vary in size to adapt to the shape of the distribution - narrow where there are many instances, wide where there are few, nonexistent where there is none

Distributions are Functions Distributions need to change as the circumstances that generated them change. That is, distributions need to be functional on other variables through correlations.

Correlations A RELATION operator connects variables with distributions together. The operator also learns from its activation, a state pin switching from Learn to Run

Contents of RELATION The RELATION operator stores a multi-dimensional map of all the distributions on all the dimensions. Cutting the range on one dimension will affect the distributions on the other dimensions - the range need not be contiguous.

Analytic and Experiential Structure A variable can have a PLUS operator on one side and a RELATION on the other.

Statistical Operators EXTRACT operators are provided to do the usual statistical things with distributions and correlations - maximum, mean, etc. The RANDOM operator will pluck a value from the range of the variable based on its current distribution.

Related Operators The Productmap operator links a set of products with their attributes - constraining to one product produces its attributes - constraining the attributes produces a set of products which have those attributes

Learn on the Run By shifting states, the system can respond based on what it knows, then use the result to update the values in its memory - the distributions and relations