Package src :: Package stats306b :: Package lecture7 :: Module nongaussian

Module nongaussian

This module tries to illustrate the heuristic in ICA that "most" projections are non-Gaussian, so to find non-Gaussian independent components, one should look for non-Gaussian components.

Three examples:

exponential, uniform, Bernoulli

Functions

[hide private]

exponential(size=(1), length=30)
Generate an array of Exponential(1) with shape source code

uniform(size=(1), length=30)
Generate an array of Exponential(1) with shape source code

bernoulli(size=(1), length=30, p=0.5)
Generate an array of Exponential(1) with shape source code

spherical(length=30)
Generate a random vector uniformly on the sphere.

source code

uniform_proj(dbn=<function exponential at 0xa93adbc>, size=(1000), length=30)
Generate a histogram of a uniformly chosen projection from one of the examples above. source code

coordinate_proj(dbn=<function exponential at 0xa93adbc>, size=(1000), length=30, delta=0.01)
Generate a histogram of a projection chosen to be near a coordinate axis (the first). source code