Dear Colleagues,

We encourage submissions of manuscripts for the following special 

issue of the International Journal of Approximate Reasoning (IJAR) on

Robustness in Probabilistic Graphical Models.

Probabilistic Graphical Models (PGM) have become the de facto framework

for representing and manipulating probabilistic knowledge in the Machine

Learning and Artificial Intelligence communities. The specification of

the numerical parameters of these models relies either on estimates

obtained from data or on subjective knowledge elicited from experts. In

either case, such parameters are prone to imprecisions and inaccuracies

resulting from noisy, incomplete or scarce data, poor human judgment,

unaccounted factors, or a combination of these situations. Such issues

can be linked to a lack of model robustness, and addressing them is

crucial if such models are to be used to support scientific theories or

to automate tasks in a reliable manner.

This special issue aims at collecting papers concerned with achieving

robustness in probabilistic graphical models. Topics of interest

include, but are not limited to:

- Methods for sensitivity analysis in PGMs; 

- Empirical analysis of PGMs' robustness;

- Non-trivial robustness issues in PGMs (e.g. adversarial examples);

- Design of inference, learning, or decision making techniques that are

robust to perturbations and/or outliers in data, in elicited knowledge,

or in model parameters;

- Robust analysis and design of robustness measurements;

- Extensions of PGMs that account for imprecision and incompleteness in

data and knowledge;

- Reliable qualitative learning and reasoning;

- Robust treatment of missing data;

- Computational complexity of learning, inference and/or decision making 

with robust models.

Submissions should be carried out through the EVISE system:

selecting "VSI:Robustness in PGMs". Submissions will follow the

standard reviewing process of IJAR. In order to guarantee an efficient

process, articles will enter the revision process as soon as they are

submitted, and be published in a Virtual Issue. The latter entails that

publication will occur as soon as an acceptance decision is reached.

Submissions should comply with the high-quality standards of IJAR, and

will be assessed by their originality, relevance and contribution to the

topic of this special issue. While there is no limit for the length of

manuscripts, typical articles contain between 20-40 pages in IJAR

format.

Submission deadline: October 1st, 2018. Since reviewing starts as soon

as a paper is submitted, we encourage early submissions.

Publication: papers will be virtually published as soon as accepted.

Denis Deratani Maua', University of Sao Paulo, Brazil, denis.maua@usp.br

Cassio P. de Campos, Utrecht University, The Netherlands, c.decampos@uu.nl