Calister Nnona

Code

Software that accompanies the research, released so the results can be reproduced and the methods deployed. Everything is on GitHub.

Repositories

cortec-framework
Reference implementations of CoRTeC for deployment: the differentially private release, generation through a frozen language model, the utility transmission bound, and a hybrid variant that pairs the release with a marginal-based generator. Python, Apache-2.0.
cortec
The papers, figures and experiment harness behind the CoRTeC results, for reproducing every table in the paper and the technical report.
mlops-pipeline
One reusable training pipeline, from dataset download through preprocessing, training, evaluation and model persistence, driven from a single call. Exercised on tabular regression, spam classification, KDD network-anomaly detection and Malimg malware-image classification. scikit-learn and PyTorch.