Write a critical report - in the scientiﬁc article style - on one of the topics listed below.
• Bayesian network inference
This topic deals with a framework for dealing with probabilistic information that has been applied in a variety of domains, primarily when dealing with large datasets involving uncertainty. Among the examples are ways of restoring a noisy image, retrieving details of corrupted transmission, determining proﬁle of a user responding to an advertisement, mapping genes onto a chromosome and so on. In your article you should consider, among others, conditional probability tables, network-independent exact and approximate algorithms, prediction and diagnostic tasks, Markovian assumptions, network construction and sensitivity analysis.
• Software certiﬁcation
Software is one of the most widely used artefacts affecting most aspects of our lives. Yes, from the engineering perspective, it is frequently a weak artefact that does not provide the warranty expected of other engineering products. The presence of the weaknesses not only makes the software unreliable but also introduces vulnerabilities leading to loss of privacy and security. In your article consider metrics (including reliability and security) and system software (including operating system, compiler, drivers) among other relevant topics. Consider modular reasoning and separation logic as methodologies involved in making program veriﬁcation scale. Also discuss certiﬁed compilation and lightweight formal methods.
• Private statistical databases
In analysis of private data, it is not sufﬁcient to ensure authenticated access or to validate the purpose for which information retrieval is required, in order to preserve privacy. For instance, if direct query about an individual X is disallowed, then the following may provide the information. If it is known that X is in a database, then a query to ﬁnd how many have a condition y combined with a query of how many, not named X, have a condition y, would provide information about X. Encryption, clearly, does not help. Consider ’input perturbation, randomizing output, auxiliary information and differential privacy, among other relevant topics.