tensorflow/privacy
A TensorFlow library implementing optimizers and tools for training machine learning models with differential privacy guarantees.
Not currently ranked — collecting fresh signals.
star history
TensorFlow Privacy is a Python library that provides TensorFlow optimizers modified to support differential privacy training. It includes analysis tools for computing privacy guarantees (such as epsilon-delta bounds) and supports efficient per-example gradient clipping for Keras models. The library enables ML practitioners to train models while mathematically bounding the amount of information that can be learned about any individual training example.
Frequently asked
- What is tensorflow/privacy?
- A TensorFlow library implementing optimizers and tools for training machine learning models with differential privacy guarantees.
- Is privacy open source?
- Yes — tensorflow/privacy is open source, released under the Apache-2.0 license.
- What language is privacy written in?
- tensorflow/privacy is primarily written in Python.
- How popular is privacy?
- tensorflow/privacy has 2k stars on GitHub.
- Where can I find privacy?
- tensorflow/privacy is on GitHub at https://github.com/tensorflow/privacy.