Lyra

The Lyra research project is a long-term research effort to enhance the understanding and reliabilty of data science software. It aims ad developing new practical and accessible analyses and tools to reason about and provide rigorous guarantees of the behavior of data analytics, big data, machine learning, and deep learning applications.

Lyra

Lyra is an umbrella project including the following focused research projects:

  • Libra (focused on fairness-aware training and certification of machine learning models)

  • Sedano (focused on designing and developing static analyses for data science notebooks)

Completed Projects

Publications

. Verifying Attention Robustness of Deep Neural Networks against Semantic Perturbations. In NFM, 2023.

PDF Project HAL Springer

. Verifying Attention Robustness of Deep Neural Networks against Semantic Perturbations. In APSEC, 2022.

Project IEEE

. Verifying Attention Robustness of Deep Neural Networks against Semantic Perturbations. CoRR abs/2207.05902, 2022.

PDF Project arXiv

. A Review of Formal Methods applied to Machine Learning. CoRR abs/2104.02466, 2021.

PDF Project arXiv HAL

. MaxSMT-Based Type Inference for Python 3. In CAV, 2018.

PDF Code Project Artifact BibTeX Springer

Talks

Interpretability-Aware Verification of Machine Learning Software
Thursday, February 9, 2023 2:00 PM
Static Analysis for Data Scientists
Friday, July 8, 2022 1:30 PM
Static Analysis for Data Scientists
Tuesday, June 14, 2022 1:30 PM
Static Analysis for Data Scientists
Friday, May 20, 2022 2:00 PM
Formal Methods for Robust Artificial Intelligence: State of the Art
Wednesday, January 13, 2021
Static Analysis for Data Science
Monday, November 2, 2020 10:00 AM
A Guided Tour of a Static Analyzer for Data Science Software
Monday, July 20, 2020 7:15 AM
Static Analysis of Data Science Software
Wednesday, October 9, 2019 2:00 PM
What Programs Want: Automatic Inference of Input Data Specifications
Tuesday, April 2, 2019 11:30 AM
An Abstract Interpretation Framework for Input Data Usage
Monday, October 2, 2017 5:00 PM
An Abstract Interpretation Framework for Input Data Usage
Tuesday, September 12, 2017 3:30 PM