[agents] AAAI Symposium on verification of neural networks - Expressions of Interest

[Alessio Lomuscio]

Dear Colleague

We are considering to submit a proposal for a AAAI spring symposium on 
verification of Neural Networks to be held at Stanford University one 
day between the 25th and the 27th of March 2019
(https://aaai.org/Symposia/Spring/sss19.php).

The ambition of the symposium is to bring together researchers 
interested in the topic in a thematic one day event to compare 
approaches and ideas on the subject.

Should you have an interest on the topic and you believe you are likely 
to attend should the workshop go ahead, can you please let us know? A 
simple "yes" reply to this email is sufficient.

If there is sufficient interest and the symposium is approved we will 
form a PC and proceed with the arrangements.

Kind regards

Clark Barrett (Stanford University, USA)
Alessio Lomuscio (Imperial College London, UK)


====================


https://aaai.org/Symposia/Spring/sss19.php


Preliminary Call for Papers

AAAI Spring Symposium on Verification of Neural Networks (VNN18)
Stanford University
25 March 2019 (TBC)

General Information
The 2018 AAAI Spring Symposium on Verification of Neural Networks 
(VNN18) aims to bring together researchers interested in methods and 
tools providing guarantees about the behaviours of neural networks and 
systems that include them.

Introduction
Methods based on machine learning are increasingly being deployed for a 
wide range of problems, including recommender systems, machine vision, 
autonomous driving, and beyond. While machine learning has made 
tremendous contributions to such applications, concerns remain about the 
lack of methods and tools available to provide formal guarantees about 
the behaviours of the resulting systems.

In particular, for data-driven methods to be usable in safety-critical 
applications, including autonomous systems, robotics, cybersecurity, and 
cyber-physical systems, it is paramount that the behaviours generated by 
neural networks are well-understood and can be predicted at design time. 
In the case of systems that are learning at run-time it is desirable 
that any change to the underlying system respects a given 
safety-envelope for the system.

While the literature on verification of traditionally designed systems 
is wide and successful, there has been a lack of results and efforts in 
this area until recently. The symposium intends to bring together 
researchers working on a range of techniques for the verification of 
neural networks, ranging from formal methods to optimisation and 
testing. The key objectives include: presentation of recent work in the 
area; discussion of key difficulties; collecting community benchmarks; 
and fostering collaboration.

One challenge for this research this area is that results are being 
published in several research communities, including formal 
verification, security and privacy, systems, and AI. One of the 
objectives of having a AAAI symposium is to help bridge these 
interdisciplinary divides to form a cross-cutting community interested 
in the verification and validation of systems based on machine learning.
The symposium workshop will include invited speakers, contributed 
papers, demonstrations, breakaway sessions, and panel sessions.

Topics of interest
The topics covered by the symposium include, but are not limited to, the 
following:

Formal specifications for neural networks and systems based on them;
SAT-based and SMT-based methods for the verification of machine learning 
systems;
Testing approaches to neural networks;
Optimisation-based methods for the verification of neural networks;
Mixed-integer Linear Programming methods for the verification of neural 
networks;
Statistical approaches to the verification of neural networks.

Submissions
We will consider two types of submissions: previously published papers 
and novel contributions. Each submission will be clearly identified as 
belonging to one of these categories. Submissions of previously 
published papers pertaining to the topic above will be lightly reviewed 
by PC members and the PC chairs. Submissions of novel material will be 
reviewed following conference standards.

Tentative Key Dates
15 February 2019: Submission deadline.
5 March 2019: Acceptance notification.
25 March 2019: Symposium

Proceedings
Compatible with any copyright restriction, we will aim to collect the 
accepted papers as informal proceedings to be made available on the day 
of the symposium and on the symposium web page. We will consider a 
special issue in a journal should there be sufficient interest.

Program Chairs

Clark Barrett, Stanford University, USA
Alessio Lomuscio, Imperial College London, UK

Programme Committee

TBC