diff options
author | brett koonce <koonce@hello.com> | 2018-06-14 21:10:49 -0700 |
---|---|---|
committer | Shanqing Cai <cais@google.com> | 2018-06-15 00:10:49 -0400 |
commit | 71ad57040b6303d2944989c2f78fa35d2a3ff103 (patch) | |
tree | 6d3e649713b93a0b0e552611f243b390dc07b772 /tensorflow/contrib/constrained_optimization | |
parent | c4bc35950e23a5c35acfce9e30897bc37ce5c8b5 (diff) |
contrib: autograph/constrained_optimization: minor spelling tweaks (#20044)
Diffstat (limited to 'tensorflow/contrib/constrained_optimization')
-rw-r--r-- | tensorflow/contrib/constrained_optimization/README.md | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/contrib/constrained_optimization/README.md b/tensorflow/contrib/constrained_optimization/README.md index c65a150464..cb1dd7d836 100644 --- a/tensorflow/contrib/constrained_optimization/README.md +++ b/tensorflow/contrib/constrained_optimization/README.md @@ -46,7 +46,7 @@ document. Imagine that we want to constrain the recall of a binary classifier to be at least 90%. Since the recall is proportional to the number of true positive classifications, which itself is a sum of indicator functions, this constraint -is non-differentible, and therefore cannot be used in a problem that will be +is non-differentiable, and therefore cannot be used in a problem that will be optimized using a (stochastic) gradient-based algorithm. For this and similar problems, TFCO supports so-called *proxy constraints*, |