Utilities¶
Linear Manifold Clustering Algorithm relies on multiple search and optimization methods:
-
mdl
(M, X; Pm = 32, Pd = 16, T = :Empirical, ɛ = 1e-4)¶ Performs calculation of the minimum description length for the linear manifold cluster.
Parameters: - M – Linear manifold cluster description as
Manifold
type instance. - X – Linear manifold cluster data as
Matrix
with points as its columns. - Pm – Precision encoding constant for the model, i.e. number of bits
required for encoding on element of the model description.
Default value is 32 which corresponds to
Float32
. - Pd – Precision encoding constant for the data.
- T – Type of a dataset encoding model as symbol:
:Gausian
,:Uniform
,:Empirical
. - ɛ – Error tolerance for bin quantization used in an empirical model encoding
Returns number of bits required to encode linear manifold cluster with the MDL schema.
- M – Linear manifold cluster description as
References
[1] |
|
[2] |
|