MiniRocket¶
MiniRocket transforms input time series using a small, fixed set of convolutional kernels. MiniRocket uses PPV pooling to compute a single feature for each of the resulting feature maps (i.e., the proportion of positive values). The transformed features are used to train a linear classifier.
Dempster A, Schmidt DF, Webb GI (2020) MiniRocket: A Very Fast (Almost) Deterministic Transform for Time Series Classification arXiv:2012.08791
1 Univariate Time Series¶
1.1 Imports¶
Import example data, MiniRocket, RidgeClassifierCV
(scikit-learn), and NumPy.
Note: MiniRocket and MiniRocketMultivariate are compiled by Numba on import. The compiled functions are cached, so this should only happen once (i.e., the first time you import MiniRocket or MiniRocketMultivariate).
[1]:
# !pip install --upgrade numba
[2]:
import numpy as np
from sklearn.linear_model import RidgeClassifierCV
from sklearn.pipeline import make_pipeline
from sktime.datasets import load_arrow_head # univariate dataset
from sktime.datasets.base import load_basic_motions # multivariate dataset
from sktime.transformations.panel.rocket import MiniRocket, MiniRocketMultivariate
1.2 Load the Training Data¶
For more details on the data set, see the univariate time series classification notebook.
Note: Input time series must be at least of length 9. Pad shorter time series using, e.g., PaddingTransformer
(sktime.transformers.panel.padder
).
[3]:
X_train, y_train = load_arrow_head(split="train", return_X_y=True)
1.3 Initialise MiniRocket and Transform the Training Data¶
[4]:
minirocket = MiniRocket() # by default, MiniRocket uses ~10,000 kernels
minirocket.fit(X_train)
X_train_transform = minirocket.transform(X_train)
1.4 Fit a Classifier¶
We suggest using RidgeClassifierCV
(scikit-learn) for smaller datasets (fewer than ~10,000 training examples), and using logistic regression trained using stochastic gradient descent for larger datasets.
Note: For larger datasets, this means integrating MiniRocket with stochastic gradient descent such that the transform is performed per minibatch, not simply substituting RidgeClassifierCV
for, e.g., LogisticRegression
.
[5]:
classifier = RidgeClassifierCV(alphas=np.logspace(-3, 3, 10), normalize=True)
classifier.fit(X_train_transform, y_train)
[5]:
RidgeClassifierCV(alphas=array([1.00000000e-03, 4.64158883e-03, 2.15443469e-02, 1.00000000e-01,
4.64158883e-01, 2.15443469e+00, 1.00000000e+01, 4.64158883e+01,
2.15443469e+02, 1.00000000e+03]),
normalize=True)
1.5 Load and Transform the Test Data¶
[6]:
X_test, y_test = load_arrow_head(split="test", return_X_y=True)
X_test_transform = minirocket.transform(X_test)
2 Multivariate Time Series¶
We can use the multivariate version of MiniRocket for multivariate time series input.
2.1 Imports¶
Import MiniRocketMultivariate.
Note: MiniRocketMultivariate compiles via Numba on import.
[8]:
# (above)
2.2 Load the Training Data¶
Note: Input time series must be at least of length 9. Pad shorter time series using, e.g., PaddingTransformer
(sktime.transformers.panel.padder
).
[9]:
X_train, y_train = load_basic_motions(split="train", return_X_y=True)
2.3 Initialise MiniRocket and Transform the Training Data¶
[10]:
minirocket_multi = MiniRocketMultivariate()
minirocket_multi.fit(X_train)
X_train_transform = minirocket_multi.transform(X_train)
2.4 Fit a Classifier¶
[11]:
classifier = RidgeClassifierCV(alphas=np.logspace(-3, 3, 10), normalize=True)
classifier.fit(X_train_transform, y_train)
[11]:
RidgeClassifierCV(alphas=array([1.00000000e-03, 4.64158883e-03, 2.15443469e-02, 1.00000000e-01,
4.64158883e-01, 2.15443469e+00, 1.00000000e+01, 4.64158883e+01,
2.15443469e+02, 1.00000000e+03]),
normalize=True)
2.5 Load and Transform the Test Data¶
[12]:
X_test, y_test = load_basic_motions(split="test", return_X_y=True)
X_test_transform = minirocket_multi.transform(X_test)
3 Pipeline Example¶
We can use MiniRocket together with RidgeClassifierCV
(or another classifier) in a pipeline. We can then use the pipeline like a self-contained classifier, with a single call to fit
, and without having to separately transform the data, etc.
3.1 Imports¶
[14]:
# (above)
3.2 Initialise the Pipeline¶
[15]:
minirocket_pipeline = make_pipeline(
MiniRocket(), RidgeClassifierCV(alphas=np.logspace(-3, 3, 10), normalize=True)
)
3.3 Load and Fit the Training Data¶
Note: Input time series must be at least of length 9. Pad shorter time series using, e.g., PaddingTransformer
(sktime.transformers.panel.padder
).
[16]:
X_train, y_train = load_arrow_head(split="train", return_X_y=True)
# it is necessary to pass y_train to the pipeline
# y_train is not used for the transform, but it is used by the classifier
minirocket_pipeline.fit(X_train, y_train)
[16]:
Pipeline(steps=[('minirocket', MiniRocket()),
('ridgeclassifiercv',
RidgeClassifierCV(alphas=array([1.00000000e-03, 4.64158883e-03, 2.15443469e-02, 1.00000000e-01,
4.64158883e-01, 2.15443469e+00, 1.00000000e+01, 4.64158883e+01,
2.15443469e+02, 1.00000000e+03]),
normalize=True))])
3.4 Load and Classify the Test Data¶
[17]:
X_test, y_test = load_arrow_head(split="test", return_X_y=True)
minirocket_pipeline.score(X_test, y_test)
[17]:
0.8685714285714285
Generated by nbsphinx. The Jupyter notebook can be found here.