Sampling-based approach for quantittive quantitative verification of Deep Neural Nets. We propose a new attack agnostic metric adversarial hardness to capture the model's robustness: https://arxiv.org/pdf/2002.06864.pdf Authors: Teodora Baluta, Zheng Leong Chua, Kuldeep S. Meel and Prateek Saxena.
Our paper on classification rules in relaxed logical form is accepted in ECAI-2020. Authors: Bishwamittra Ghosh, Dmitry Malioutov, and Kuldeep S. Meel
Dec 23, 2019
Our paper on Symmetry breaking and model counting is accepted to TACAS 2020. Authors: Wenxi Wang, Muhammad Usman, Alyas Almaawi, Kaiyuan Wang, Kuldeep S. Meel, and Sarfraz Khurshid
Dec 7, 2019
Two of our group's papers are accepted to accepted as poster presentations with a spotlight talk at StarAI 2020 workshop in AAAI 2020: Our AIES-19 paper on incremental classification rule learning and our CCS-19 paper on quantiatative verification for binarized neural networks.
Nov 11, 2019
Our paper on MaxSAT-based formulation for group testing is accepted in AAAI 2020. Authors: Lorenzo Ciampiconi, Bishwamittra Ghosh, Jonathan Scarlett and Kuldeep S. Meel
Our NPAQ framework focused on providing PAC guarantees for verification of Neural Networks is accepted to CCS-19 Quoting reviewer: “This work is pioneering a new technique to solve an incredibly challenging problem, and it shows that smaller problem can be solved. I can live with that, future work can improve computational efficiency.” Authors: Teodora Baluta, Shiqi Shen, Shweta Shinde, Kuldeep S. Meel, Prateek Saxena
Jun 23, 2019
Our paper on interpretable rules expressed as relaxed-CNF is accepted at IJCAI workshop on XAI (Explainable Artificial Intelligence) and DSO (Data Science meets Optimization), 2019. Authors: Bishwamittra Ghosh, Dmitry Malioutov, Kuldeep S. Meel.
May 15, 2019
Kuldeep recieved notification of the award of NRF Fellowship for AI for the project: Provably Verified and Explainable Probabilistic Reasoning.
May 9, 2019
Two papers accepted to IJCAI.The first paper explores the phase transition behavior of conjunction of cardinality and XOR constraints. Authors: Yash Pote, Saurabh Joshi, Kuldeep Meel. The second paper describes a radically new approach to exact counting wherein we compute estimates that are probabilistically exact! Authors: Shubham Sharma, Kuldeep Meel. Combined with our invited paper on #DNF , this makes 3 papers that we will be presenting at IJCAI.
Apr 22, 2019
Two papers accepted to SAT 2019. The first paper introduces the first version of CrystalBall, a framework intended to allow gazing into the black box of SAT solving. Authors: Kuldeep, Mate Soos, Raghav Kulkarni. The second paper discusses how model counting can be used to analyze explanations provided by tools such as ANCHOR. Authors: Kuldeep, Nina Narodytska, Aditya Shrotri, Alexey Ignatiev, and Joao Marques Silva.