Open MatthewMiddlehurst opened 1 week ago
note that multirocketmultivariate is fine. Inappropriate squeezing I guess. I can fix in #1711, I'll add a test for all base collection transformers
from aeon.transformations.collection.convolution_based import MultiRocket, MultiRocketMultivariate
import numpy as np
)
X = np.random.random((40,1,20))
mr = MultiRocketMultivariate()
mr.fit(X)
print("Finished fit for multirocket multivariate")
X2 = np.random.random((1,20))
Xt = mr.transform(X2)
print(Xt)
mr = MultiRocket()
mr.fit(X)
print("Finished fit for multirocket")
X2 = np.random.random((1,20))
Xt = mr.transform(X2)
print(Xt)
Describe the bug
From #1696
The
MultiRocket
transform does not accept collections of size 1 when transforming. Discovered through theArsenal
classifier.We should add to the general testing a line to predict cases like this. Should be relatively easy and cheap.
Steps/Code to reproduce the bug
Expected results
transformer transforms the single case without exception
Actual results
Versions
N/A