BUGFIX: sample_stats.percentile: Fractions for interpolation were done in
authorIan Langmore <langmore@google.com>
Fri, 30 Mar 2018 02:29:21 +0000 (19:29 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Fri, 30 Mar 2018 02:33:48 +0000 (19:33 -0700)
commit9d8201cad77f6892c8da059bf23348c7eb16e3be
treeb174f1bebf0c3e4de81bb709163a378da1bb5f5a
parent8712a7fa24da7803a79c10501f8fd65d358a62b5
BUGFIX:  sample_stats.percentile:  Fractions for interpolation were done in
float, which caused the error d - 1 == d for large 'd'.  Fix is to do in
double.  Also clipping index values to [0,..., d - 1] in case double isn't enough for some huge array
PiperOrigin-RevId: 191023164
tensorflow/contrib/distributions/python/kernel_tests/sample_stats_test.py
tensorflow/contrib/distributions/python/ops/sample_stats.py