Program
AAAI 2019 Spring Symposium
Verification of Neural Networks (VNN19)
Monday, March 25
9:15 am - 9:30 am Symposium opening: Clark Barrett, Alessio Lomuscio
9:30 am - 10:30 am Invited talk: Suman Jana. Scalable Training of Verifiably Robust Neural Networks. [slides]
10:30 am - 11:00 am Coffee Break
11:00 am - 12:30 pm Analysis of Neural Networks.
11:00 Changliu Liu, Tomer Arnon, Christopher Lazarus, Clark Barrett and Mykel Kochenderfer. Algorithms for Verifying Deep Neural Networks. [slides]
11:30 Weiming Xiang, Hoang Dung Tran and Taylor T Johnson. Specification-Guided Safety Verification for Feedforward Neural Networks. [slides]
12:00 Qinglong Wang, Kaixuan Zhang, Xue Liu and C. Lee Giles. Verification of Recurrent Neural Networks Through Rule Extraction.
12:30 pm - 2:00 pm Lunch
2:00 pm - 3:30 pm Robustness Analysis.
2:00 Armando Tacchella. There is plenty of room at the bottom. [slides]
2:30 Tommaso Dreossi, Shromona Ghosh, Sanjit A. Seshia and Alberto Sangiovanni-Vincentelli. A General Formalization of Robustness for Deep Neural Networks. [slides: keynote pdf]
3:00 Sumeet Singh, Gamaleldin Elsayed, Yinlam Chow, Mohammad Ghavamzadeh and Marco Pavone. Designing Robust Margin Neural Networks using Convex Relaxations.
3:30 pm - 4:00 pm Coffee Break
4:00 pm - 5:30 pm Panel Discussion. Artur Garcez, Suman Jana, Nina Narodytska [slides], Michael Warren, Iain Whiteside.
6:00 pm - 7:00 pm Reception.
Tuesday, March 26
9:30 am - 10:30 am Invited talk: Krishnamurthy Dvijotham. Verification of neural networks: Tightness, Scalability and Broad Applicability. [slides]
10:30 am - 11:00 am Coffee Break
11:00 am - 12:30 pm Closed-loop systems 1.
11:00 Francesco Leofante, Nina Narodytska, Luca Pulina and Armando Tacchella. Reasoning about Neural Networks: a Taxonomy of Current Approaches. [slides]
11:30 Weiming Xiang, Xiaodong Yang, Hoang Dung Tran and Taylor T Johnson. Reachability Analysis and Safety Verification for Neural Network Control Systems. [slides]
12:00 Kyle Julian, Shivam Sharma, Jean-Baptiste Jeannin and Mykel Kochenderfer. Verifying Aircraft Collision Avoidance Neural Networks Through Linear Approximations of Safe Regions. [slides]
12:30 pm - 2:00 pm Lunch
2:00 pm - 3:30 pm Closed-loop systems 2.
2:00 Michael Akitunde, Andrea Kevorchian, Alessio Lomuscio, Lalit Maganti, Edoardo Pirovano. Verification of RNN-based neural systems. [slides]
2:30 Radoslav Ivanov, James Weimer, Rajeev Alur, George Pappas and Insup Lee. Verisig: Verifying safety properties of closed-loop systems with neural network controllers. [slides]
3:00 Andy Shih, Adnan Darwiche and Arthur Choi. Verifying Binarized Neural Networks by Local Automaton Learning. [slides]
3:30 pm - 4:00 pm Coffee Break
4:00 pm - 5:30 pm Tools.
4:00 Guy Katz, Derek Huang, Duligur Ibeling, Kyle Julian, Christopher Lazarus, Rachel Lim, Parth Shah, Shantanu Thakoor, Haoze Wu, Aleksandar Zeljic, David Dill, Mykel Kochenderfer, Clark Barrett. The Marabou Framework for Verification and Analysis of Deep Neural Networks. [slides]
4:30 Souradeep Dutta, Taisa Kushner, Susmit Jha, Sriram Sankaranarayanan, Natarajan Shankar and Ashish Tiwari. Sherlock : A Tool for Verification of Deep Neural Networks. [slides]
5:00 Huan Zhang, Pengchuan Zhang and Cho-Jui Hsieh. RecurJac: An Efficient Recursive Algorithm for Bounding Jacobian Matrix of Neural Networks and Its Applications.
6:00 pm - 7:00 pm Joint plenary session. [slides]
Wednesday, March 27
9:00 am - 10:30 am Other approaches.
9:00 Arthur Choi, Weijia Shi, Andy Shih and Adnan Darwiche. Compiling Neural Networks into Tractable Boolean Circuits. [slides]
9:30 Nathanaël Fijalkow. Global Correctness for Neural Networks Image Classifiers. [slides]
10:00 Michael Warren and Chris Serrano. Introspection Learning. [slides]
10:30 am - 11:00 am Coffee Break
11:00 am - 12:00 pm Symposium discussion on benchmark formats.
12.00 pm Symposium close.