This code implements the Hierarchical Nested Segmentation model (HNS) from:

Hierarchical Span-Based Conditional Random Fields for Labeling and Segmenting Events in Wearable Sensor Data Streams. Roy Adams, Nazir Saleheen, Edison Thomas, Abhinav Parate, Santosh Kumar, and Benjamin Marlin. International Conference on Machine Learning, 2016.

This code requires a number of python modules listed in requirements.txt. To install these requirements using pip, call

pip install -r requirements.txt

Additionally, portions of this code is written in cython and must be compiled. To complie this code exectute,

python setup.py build_ext --inplace

The core of the code is implemented in the hierarchical_nested_segmentation class. This class implements the scikit-learn fit/predict interface.

from hierarchical_nested_segmentation import HierarchicalNestedSegmentation as hns
import numpy as np

X,Y = np.load("../data/test_data.npy")
hns = HNS(n_x0_features=10,n_x1_features=10)
hns.fit(X,Y)
Y_pred = hns.predict(X)