simu_bfem: Experimental setting of the chapter BFEM

In FisherEM: The FisherEM Algorithm to Simultaneously Cluster and Visualize High-Dimensional Data

Description

Experimental setting of the chapter BFEM

Usage

simu_bfem(n, which = "Chang1983", ...)

Arguments

n	Number of observations
which	Type of simulation, either: "Chang1983" - Simulate the dataset of Chang's (1983) paper : a mixture of 2 Gaussian with in dimension p=15. "section4.2" - Experimental setting of Section 4.2: DLM model in dimension p with d=2 and K=3, with noisy dimensions. "section4.3" - Experimental setting of Section 4.3: Same as '"section4.2"' except the noise is expressed in term of signal-to-noise ration (decibels).
...	Additional param controlling the simulation p - The desired observed space dimension, the latent dimension is kept fixed to d=2 and noisy Gaussian dimensions are added (useless for '"Chang1983"') noise (for '"section4.2"' only) - Variance of the noise snr (for '"section4.3"' only) - Signal-to-noise ratio (in decibels) representing the ratio of signal and noise variances in logarithmic scale. The greater snr, the smaller noise variance.

Value

A list with slots

• Y - The simulated data.

• cls - The true clustering.

Examples

n = 300

# Chang's 1983 setting
simu = simu_bfem(n = n, which = "Chang1983")

# Section 4.2 setting
p = 25
noise = 1
simu = simu_bfem(n, which = "section4.2", p = p, noise = noise)

# Section4.3 setting
snr = 3 # noise variance is 2 times smaller than that of the signal.
simu = simu_bfem(n, which = "section4.3", snr = 10)

FisherEM documentation built on Oct. 23, 2020, 8:08 p.m.