I noticed that there seems to be an overflow happening with `numpy.linalg.matrix_power` when using large exponents. In my opinion, this should at least raise an `OverflowError`. ``` np.version.git_revision Out[8]: '68538b74483009c2c2d1644ef00397014f95a696' np.version.full_version == '1.6.1' In [10]: A Out[10]: array([[1, 2], [1, 0]]) In [11]: np.linalg.matrix_power(A, 100) Out[11]: array([[-6148914691236517205, -6148914691236517206], [ 6148914691236517205, 6148914691236517206]]) ```