nm

SOM utils

To install: pip install nm

Overview

The nm package provides a Python implementation of Self-Organizing Maps (SOMs), a type of unsupervised learning neural network that is used to produce a low-dimensional (typically two-dimensional), discretized representation of the input space of the training samples. This implementation includes various functionalities such as data normalization, initialization methods, training processes, and visualization tools.

Main Features

• Normalization: Support for data normalization which is crucial for effective SOM training.
• Initialization Methods: Includes PCA-based and random initialization to start the SOM training.
• Training: Supports both batch and sequential training methods.
• Visualization: Functions to visualize the SOM's codebooks and the mapping from data space to SOM space.
• Projection: Ability to project new data onto the trained SOM.
• Clustering: Utilize the trained SOM for clustering data.
• Utility Functions: Additional functions for data denormalization, finding k-nearest nodes, etc.

Usage

Initialization

To create a SOM instance:

from nm import SOM
som = SOM(name="My_SOM", Data=my_data, mapsize=[20, 30], norm_method='var', initmethod='pca', neigh='Gaussian')

Training

To train the SOM:

som.train(n_job=1, shared_memory='no', verbose='on')

Visualization

To visualize the trained SOM:

som.view_map(what='codebook', which_dim='all', pack='Yes', text_size=10, save='No', grid='Yes', text='Yes')

Project Data

To project new data onto the trained SOM and find the best matching units (BMUs):

new_data_projection = som.project_data(new_data)

Clustering

To perform clustering on the map nodes:

cluster_labels = som.cluster(method='Kmeans', n_clusters=8)

Prediction

To predict using the trained SOM:

predicted_values = som.predict(X_test)

Documentation

Each function in the SOM class is well documented with docstrings, providing details on the parameters, the expected inputs, and outputs. Users are encouraged to refer to these docstrings for more detailed usage instructions.

Installation

This package can be installed using pip:

pip install nm

Requirements

This package depends on several external libraries including:

• NumPy
• Matplotlib
• Scikit-learn
• Joblib
• NumExpr
• SciPy

Ensure these are installed in your Python environment to use nm package effectively.