Category: Non classé

We are happy to welcome two new PhD students this Autumn: Remi Desmartin, who will work with supervisors Komendantskaya, Stark and Passmore (from Imandra.ai) on neural network libraries in Imandra and Aina Centelles Tarres, who will work with supervisors Gabbay and Komendantskaya on nominal logic for neural network verification. Remi and Aina, have a great and productive time during your PhD studies!

LAIV PhD student, Daniel Kienitz, has just had his paper accepted at the Asian Conference on Computer Vision (ACCV 2022): D. Kienitz, E. Komendantskaya and M. Lones.  Comparing Complexities of Decision Boundaries for Robust Training: A Universal Approach. Many congratulations, Daniel!

• MSc dissertation marks have been released, and MSc students supervised by the LAIV team: Youssef, Henri-Louis, Assya, Samuel, Sinead, Haoran, all got excellent marks (one B and five As). Many congratulations to the students, but also their supervisors: Luca, Matthew, Marco, Natalia, Yuhui! Well done all. If curious to see the final titles and supervisory teams, check them here.
• in addition, Henri-Louis Boisvert got a prize for the best MSc dissertation award for his thesis “Evaluating the Performance of Different Reinforcement Learning Methods for Autonomous Racing“. Many congratulations Henri-Louis and Luca (the main supervisor)!

LAIV/DSG seminars are restarting this week, with Remi’s talk on neural net verification in Imandra. Join us online or F2F!

The Lab for AI and Verification is looking to fill one PhD post and one postdoctoral post. We are looking for candidates with solid knowledge of Theorem Proving and/or Functional/Logic programming, and enthusiasm to apply this knowledge in the domain of Artificial Intelligence.
The PhD post is for 4 years, starting in October 2022. It covers full stipend and PhD fees and is sponsored by the UKRI (ukri.org) and Schlumberger Cambridge (slb.com). The company will provide additional training and support during the PhD studies. This post needs to be filled in as  soon as possible.
We are also looking to employ a postdoctoral researcher for a  6-12 months project to formalise Criminal law in the Functional Language Catala. Formalising criminal law for autonomous cars is of particular interest. This project will be in collaboration with Jonathan Protzenko, Microsoft Research and the School of Law, Edinburgh University. This project has a flexible starting date. 

Please direct all queries to Ekaterina Komendantskaya (ek19@hw.ac.uk) 

In collaboration with IMANDRA (www.imandra.ai), we are developing a neural network verification library in the language Imandra. This week, this work was accepted for publication: R. Desmartin, G. Passmore E. Komendantskaya and M. Daggitt: CheckINN: Wide Range Neural Network Verification in Imandra. 24th International Symposium on Principles and Practice of Declarative Programming PPDP’22.

LAIV team (Katya, Kathrin, Filip, Marco, Natalia, Akilan) are running SPLV’22 this week. For web streaming and recordings visit SPLV main page: https://www.macs.hw.ac.uk/splv/splv-2022/

Many congratulations to LAIV PhD students and researchers, who will present their work at FOMLAS’22, as part of the FLOC’22 conference:

Matthew Daggitt, Wen Kokke, Robert Atkey, Luca Arnaboldi and Ekaterina Komendantskaya: Vehicle: A High-Level Language for Embedding Logical Specifications in Neural Networks

Natalia Ślusarz, Ekaterina Komendantskaya, Matthew Daggitt and Robert Stewart: Differentiable Logics for Neural Network Verification

Marco Casadio, Ekaterina Komendantskaya, Verena Rieser, Matthew Daggitt, Daniel Kienitz, Luca Arnaboldi and Wen Kokke: Why Robust Natural Language Understanding is a Challenge

Remi Desmartin, Grant Passmore and Ekaterina Komendantskaya: Neural Networks in Imandra: Matrix Representation as a Verification Choice

The School registration is now open, please attend and register at https://www.macs.hw.ac.uk/splv/splv22/ 

Many congratulations to Marco Casadio et al for having a paper accepted at the international conference on Computer-Aided Verification CAV’22 (Part of FLOC’22). Marco Casadio, Ekaterina Komendantskaya, Matthew L. Daggitt, Wen Kokke, Guy Katz, Guy Amir and Idan Refaeli: “Neural Network Robustness as a Verification Property: A Principled Case Study”