MTF some arrays with no constant error!

[cssaheel]

MTF some arrays with no constant error!

[johannfaouzi]

Hi,

The error comes from this private method:

pyts/pyts/preprocessing/discretizer.py

Lines 138 to 140 in 10b7a7a

	def _check_constant(self, X):
	if np.any(np.max(X, axis=1) - np.min(X, axis=1) == 0):
	raise ValueError("At least one sample is constant.")

If this error is raised, it means that there is at least one sample (i.e. one time series) that is constant. The MarkovTransitionField algorithm discretizes time series into n_bins bins, which would lead to a single bin if a time series is constant. It may be better to raise a warning rather than an error.

If you remove the constant time series from your 2D-array, this error will no longer be raised. You could find the corresponding indices using np.where(np.var(x_train_np, axis=1) == 0)[0].

I hope that this helps a bit.

[wwjd1234]

I have this issue as well when padding my time series with zeros so that they are all the same length