eonu / sequentia

Currently if a single sequence is provided to predict() in HMMClassifier, it is converted into a singleton list:

sequentia/lib/sequentia/classifiers/hmm/hmm_classifier.py

Lines 118 to 119 in b857c72

    
           # Convert single observation sequence to a singleton list 
        
           X = [X] if isinstance(X, np.ndarray) else X

When returning the prediction, we assume that if X has one element, the output should not be returned as a singleton.

sequentia/lib/sequentia/classifiers/hmm/hmm_classifier.py

Lines 133 to 136 in b857c72

    
           if len(X) == 1: 
        
               return (labels.item(), scores.flatten()) if return_scores else labels.item() 
        
           else: 
        
               return (labels, scores) if return_scores else labels

This can be unintuitive as if the user actually provides a singleton sequence as input, they will receive a prediction value that is not returned as a singleton.

Fixed in #200.

	# Convert single observation sequence to a singleton list
	X = [X] if isinstance(X, np.ndarray) else X

	if len(X) == 1:
	return (labels.item(), scores.flatten()) if return_scores else labels.item()
	else:
	return (labels, scores) if return_scores else labels

Don't treat single sequence input as singleton for HMMClassifier