v0.3.0 - Initial TensorFlow support
This release adds initial support for generating TensorFlow examples. The provided code is an updated version of that used in our publication, Scalable and accurate deep learning with electronic health records. Additional usage examples and sample models will be added over the coming months.
The support for generating custom protos has been much improved in this release. The C++ implementation now comes with a JSON parser and printer, which supports profiles with inlined extensions.
Features
- Code to convert FHIR bundles to TensorFlow SequenceExamples
- Google-defined extensions for labels used in example generation
- C++ JSON parser / printer (in addition to the existing Java JSON parser/printer)
- Generation of DataTypes, Extensions and Resources from StructureDefinitions
- Expanded support for profiles and slicing
- Generated protos for the US Core Implementation Guide
- Updated examples