search

facebookresearch / pysparnn

Approximate Nearest Neighbor Search for Sparse Data in Python!
Other
915 stars 146 forks source link

RuntimeWarning: divide by zero encountered in true_divide #17

Open yannDroy opened 6 years ago

yannDroy commented 6 years ago 
RuntimeWarning: divide by zero encountered in true_divide
    magnitude = 1.0 / (a_root_sum_square * self.matrix_root_sum_square)

is there a way to get rid of that warning ? Thanks

VictorDDT commented 5 years ago

The problem appears here in "matrix_distance.py": magnitude = 1.0 / (a_root_sum_square * self.matrix_root_sum_square)

Here is a simple debug which shows that the problem comes due to "0" in self.matrix_root_sum_square

a_root_sum_square <class 'numpy.ndarray'> (1000, 1) np.unique: [1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]

self.matrix_root_sum_square <class 'numpy.ndarray'> (916,) np.unique: [0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.]

*a_root_sum_square self.matrix_root_sum_square* <class 'numpy.ndarray'> (1000, 916) np.unique: [0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.] /pysparnn/matrix_distance.py:213: RuntimeWarning: divide by zero encountered in true_divide magnitude = 1.0 / (a_root_sum_square self.matrix_root_sum_square)