Issue with ITD method

[ivan-marroquin]

Describe the bug
Thanks for such a great package!

I am using Python 3.12.0 with scipy 1.15.2, PySDKit 0.4.17, and numpy 2.1.3. Python running on Ubuntu 22.04

When I run the ITD method on my dataset, I get the following error message:

LK1 (48,) vs LK2 (48,)
[-1.20752663e-02 -1.09578711e-02 -9.88145499e-03 ... -1.11700587e+03
-1.11700565e+03 -5.40769799e-03]
LK1 (14,) vs LK2 (12,)
Traceback (most recent call last):
File "", line 1, in
File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/pysdkit/_itd/itd.py", line 140, in fit_transform
L1, H1 = self.itd_baseline_extract(signal)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/pysdkit/_itd/itd.py", line 109, in itd_baseline_extract
LK = np.vstack((LK1, LK2)).T
^^^^^^^^^^^^^^^^^^^^^
File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/numpy/_core/shape_base.py", line 291, in vstack
return _nx.concatenate(arrs, 0, dtype=dtype, casting=casting)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: all the input array dimensions except for the concatenation axis must match exactly, but along dimension 1, the array at index 0 has size 14 and the array at index 1 has size 12

I attached to this ticket, the data used to reproduce the encountered issue.

Thanks for your help
Ivan

To Reproduce
import numpy as np
from pysdkit import ITD
trace= np.load('trace.npy')

itd_gral= ITD(N_max= 3)
imf_gral= itd_gral.fit_transform(trace)

Expected behavior
I expect the ITD method to generate 3 decompositions per analyzed signal

trace.zip

[wwhenxuan]

Describe the bug Thanks for such a great package!

I am using Python 3.12.0 with scipy 1.15.2, PySDKit 0.4.17, and numpy 2.1.3. Python running on Ubuntu 22.04

When I run the ITD method on my dataset, I get the following error message:

LK1 (48,) vs LK2 (48,) [-1.20752663e-02 -1.09578711e-02 -9.88145499e-03 ... -1.11700587e+03 -1.11700565e+03 -5.40769799e-03] LK1 (14,) vs LK2 (12,) Traceback (most recent call last): File "", line 1, in File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/pysdkit/_itd/itd.py", line 140, in fit_transform L1, H1 = self.itd_baseline_extract(signal) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/pysdkit/_itd/itd.py", line 109, in itd_baseline_extract LK = np.vstack((LK1, LK2)).T ^^^^^^^^^^^^^^^^^^^^^ File "/home/ivan_phd/python/python-3.12.0/lib/python3.12/site-packages/numpy/_core/shape_base.py", line 291, in vstack return _nx.concatenate(arrs, 0, dtype=dtype, casting=casting) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ValueError: all the input array dimensions except for the concatenation axis must match exactly, but along dimension 1, the array at index 0 has size 14 and the array at index 1 has size 12

I attached to this ticket, the data used to reproduce the encountered issue.

Thanks for your help Ivan

To Reproduce import numpy as np from pysdkit import ITD trace= np.load('trace.npy')

itd_gral= ITD(N_max= 3) imf_gral= itd_gral.fit_transform(trace)

Expected behavior I expect the ITD method to generate 3 decompositions per analyzed signal

trace.zip

@ivan-marroquin

First of all, thank you for your willingness to use PySDKit and your recognition of our work. The success of PySDKit is inseparable from the support of all users.

Thank you for your valuable questions. I tested the ITD algorithm with a univariate signal, but found that the algorithm currently works properly. The shape of the NumPy array I input is (seq_len, ) where seq_len represents the length of the signal. But I will still conduct further tests, such as when the array length is too long or has extreme values.

Thank you for the data you provided. I visualized and analyzed it and found that the signal was too simple. I speculated that the ITD algorithm could not recognize the possible time scale information, so an error occurred. I will further check and modify it. Thank you for your valuable question.

The signal you provided has a small fluctuation degree and can almost be regarded as an intrinsic mode function. But if you want to analyze it, I will recommend two algorithms, EMD and VMD. The specific code is as follows:

from pysdkit import EMD
from pysdkit.plot import plot_IMFs

emd = EMD()
imfs = emd(data)

plot_IMFs(data, imfs, save_figure=True, dpi=400)

Thank you for your valuable questions and suggestions to make PySDKit better.

I wish you a happy day 🥰
Whenxuan

[ivan-marroquin]

Hi @wwhenxuan

Thanks for suggesting testing the EMD and VMD algorithms with my dataset.

I wish you a wonderful day.

Ivan

[ivan-marroquin]

Hi @wwhenxuan

I thought you may be interested. I mentioned your package on a post on LinkedIn.

https://www.linkedin.com/posts/ivan-d-marroquin_signalprocessing-machinelearning-python-activity-7336775140703092736-4e-L?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAJ8htIB3O6pOKctydOoNc9UQqWDTbwImg4