Skip to main content
Join the official 2019 Python Developers SurveyStart the survey!

A Wasserstein Subsequence Kernel for Time Series

Project description

Wasserstein Time Series Kernel

Installation

The easiest way is to install WTK from the Python Package Index (PyPI) via

$ pip install wtk

Usage

The package provides functions to transform a set of n training time series and o test time series into an n x n distance matrix for training and an o x n distance matrix for testing. Additionally, we provide a way to run a grid search for a krein space SVM. krein_svm_grid_search runs a 5-fold cross-validation on the training set to determine the best hyperparameters. Then, its classification accuracy is computed on the test set.

from wtk import transform_to_dist_matrix
from wtk.utilities import get_ucr_dataset, krein_svm_grid_search

# Read UCR data
X_train, y_train, X_test, y_test = get_ucr_dataset('../data/UCR/raw_data/', 'DistalPhalanxTW')

# Compute wasserstein distance matrices with subsequent length k=10
D_train, D_test = transform_to_dist_matrix(X_train, X_test, 10)

# Run the grid search
svm_clf = krein_svm_grid_search(D_train, D_test, y_train, y_test)

Alternatively, you can get the kernel matrices computed from the distance matrices and train your own classifier.

from sklearn.svm import SVC
from wtk import get_kernel_matrix
from sklearn.metrics import accuracy_score

# Get the kernel matrices
K_train = get_kernel_matrix(D_train, psd=True, gamma=0.2)
K_test = get_kernel_matrix(D_test, psd=False, gamma=0.2)

# Train your classifier
clf = SVC(C=5, kernel='precomputed')
clf.fit(K_train, y_train)

y_pred = clf.predict(K_test)
print(accuracy_score(y_test, y_pred))

Help

If you have questions concerning WTK or you encounter problems when trying to build the tool under your own system, please open an issue in the issue tracker. Try to describe the issue in sufficient detail in order to make it possible for us to help you.

Contributors

WTK is developed and maintained by members of the Machine Learning and Computational Biology Lab of Prof. Dr. Karsten Borgwardt:

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for wtk, version 0.2.1.3
Filename, size File type Python version Upload date Hashes
Filename, size wtk-0.2.1.3-py3-none-any.whl (18.3 kB) File type Wheel Python version py3 Upload date Hashes View hashes
Filename, size wtk-0.2.1.3.tar.gz (6.5 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page