“Privacy concerns have extended beyond the accessibility of data sets since the development of machine learning algorithms. While it is common to assume that a model that generalizes well protects privacy, out-of-distribution predictions could reveal sensitive information. That being said, because algorithms may learn and record personal information, their deployment may be as informative as releases of the actual training data sets, with respect to personally identifying information…”
My partner and I investigated the performance and usability of the recently released TensorFlow Privacy package using a large, sensitive edX dataset. Read more about our project below, and in our repository.
Karina Huang 
You must be logged in to post a comment.