Workshop : Verification of Systems that Learn

Robotic and Autonomous systems (RAS) are moving outside of industrial settings and into areas where they will interact with the general public. A key component of many RAS is machine learning. The success of deep learning has achieved widespread publicity in the last few years, but many other forms of learning and adaptation exist. By their nature systems that learn pose significant challenges to the the verification community. Machine learning is of particular value in areas where developing a precise specification of desired behaviour is outside the scope of our current understanding of the world. For instance machine learning is widely deployed for image classification tasks – in these cases the specification is that the classifier should match the perception ability of a human which is not a formal specification in the usual sense. Even where formal properties can be provided the results of many machine learning systems (e.g., a set of weights in a neural network) are difficult to map onto these and to reason about in appropriate terms. As a result, while there is currently a great deal of interest in the verification of RAS, comparatively little attention has been paid to the verification of learning.

The aim of this workshop is to bring together researchers interested in the question of how systems that learn may be verified with the intention of building a UK community in the area.

Workshop Format

The workshop will be a part of the AISB Convention to be held at the University of Liverpool from  4th-6th April.

The workshop will take the form of a number of scientific presentations and posters, invited speakers and a possibly panel (depending on scheduling constraints).

Confirmed Keynote Speakers

Sandor M. Veres (University of Sheffield)

Contact Information

For more information contact Louise Dennis (L.A.Dennis@liverpool.ac.uk) or Alice Miller (Alice.Miller@glasgow.ac.uk)

Sponsored by the UK Network on the Verification and Validation of Autonomous Systems.

Advertisements