Neural Structured Learning v1.1.0
Release 1.1.0
Major Features and Improvements
-
Introduces
nsl.tools.build_graph, a function for graph building. -
Introduces
nsl.tools.pack_nbrs, a function to prepare input for
graph-based NSL. -
Adds
tf.estimator.Estimatorsupport for NSL. In particular, this release
introduces two new wrapper functions named
nsl.estimator.add_graph_regularizationand
nsl.estimator.add_adversarial_regularizationto wrap existing
tf.estimator.Estimator-based models with NSL. These APIs are currently
supported only for TF 1.x.
Bug Fixes and Other Changes
-
Adds version information to the NSL package, which can be queried as
nsl.__version__. -
Fixes loss computation with
Lossobjects inAdversarialRegularization. -
Adds a new parameter to
nsl.keras.adversarial_losswhich can be used to
pass additional arguments to the model. -
Fixes typos in documentation and notebooks.
-
Updates notebooks to use the release version of TF 2.0.
Thanks to our Contributors
This release contains contributions from many people at Google.