amaotone / pygtm

A python implementation of the Generative Topographic Mapping

PyGTM

A python implementation of Generative Topographic Mapping.

This is beta release. For example, this project has no test as you can see.

Requirements

• numpy
• scipy
• scikit-learn

Getting Started

To install PyGTM, use pip

$ pip install -U pygtm

The pygtm package inherits scikit-learn classes.

from pygtm import GTM
from sklearn.datasets import load_iris
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import make_pipeline

iris = load_iris()
model = make_pipeline(
    StandardScaler(),
    GTM(n_components=2)
)
embedding = model.fit_transform(iris.data)

References

• GTM: The Generative Topographic Mapping
• Development of the Generative Topographic Mapping

pygtm's Issues

Strange names of variables: n_rbfs, n_grids

When I set n_rbfs = 10, the real number of rbfs is set to be 11**2, and the same situation will occur for n_grids.

I think it is natural that #rbf is set to be n_rbf ** 2.

Add more example

• README.md
  • the result of transformation (iris)
• examples/
  • the heatmap of responsibility

utility of Line 40

The whole code is awesome and very useful, but what's the utility of this line?

pygtm/pygtm/gtm.py

Line 40 in 0bb84dd

np.fill_diagonal(inter_dist, np.inf)