A Julia implementation of the Rocket method of using random feature kernels for time series classification.

This project is not programmed by the original authors of the original paper; please see the Attribution section for more details on the original paper and software.

Documentation	Coverage	CI Status	Releases
Dependents	Issues	JuliaHub Status	Downloads

• Table of Contents
• Usage
• Attribution
  • Authors
  • Icons

For detailed usage instructions, please see the Documentation.

To use the package, you must:

1. Load Rocketeer,
2. Create a RocketModule object (with optionally specified input_length and n_kernels hyperparameters),
3. apply_kernels to your dataset to extract the Rocket features,
4. Optionally save_rocket and load_rocket if you intend to utilize the exact same kernels in future experiments.

For example:

# Load the module
using Rocketeer

# Set some parameters of the example
filepath = "my_rocket"  # Point to a save file
input_length = 10       # The length of the input window
n_kernels = 200         # The number of kernels to generate

# Create a rocket module
my_rocket = RocketModule(input_length, n_kernels)

# Save it for future use
save_rocket(my_rocket, filepath)

# Load the module back into a new object
my_new_rocket = load_rocket(filepath)

# Create some random data
X = rand(input_length)

# Apply the kernels to get features
features = apply_kernels(my_new_rocket, X)

This Julia package is programmed by:

• Sasha Petrenko petrenkos@mst.edu @AP6YC

The original paper is authored by:

• Angus Dempster
• Francois Petitjean
• Geoff Webb

The links for the original work are:

• Papers:
  • Journal article (DOI)
  • Preprint
• Software:
  • rocket (Python)

The bibtex entry for the original paper is:

@article{dempster_etal_2020,
    author  = {Dempster, Angus and Petitjean, Francois and Webb, Geoffrey I},
    title   = {ROCKET: Exceptionally fast and accurate time classification using random convolutional kernels},
    year    = {2020},
    journal = {Data Mining and Knowledge Discovery},
    doi     = {https://doi.org/10.1007/s10618-020-00701-z}
}

The icon used for the project logo is from the following:

• Space shuttle icons created by Freepik - Flaticon (startup_1067357)