In this paper we introduce indeed two alternative formulations of the theory of evidence by proving that both plausibility and commonality functions share the same combinatorial structure of sum function of belief functions, and computing their Moebius inverses called basic plausibility and commonality assignments. The equivalence of the associated formulations of the ToE is mirrored by the geometric congruence of the related simplices. Applications to the probabilistic approximation problem are briefly presented.
In this paper we build on previous work on the geometry of Dempster’s rule to investigate the geometric behaviour of various other combination rules, including Yager’s, Dubois’, and disjunctive combination, starting from the case of binary frames of discernment. Believability measures for unnormalised belief functions are also considered. A research programme to complete this analysis is outlined.
In this paper we propose a generalised maximum-entropy classification framework, in which the empirical expectation of the feature functions is bounded by the lower and upper expectations associated with the lower and upper probabilities associated with a belief measure. This generalised setting permits a more cautious appreciation of the information content of a training set. We analytically derive the KarushKuhn-Tucker conditions for the generalised max-entropy classifier in the case in which a Shannon-like entropy is adopted.
In this paper, motivated by the treatment of conditional constraints in the data association problem, we state and prove the generalisation of the law of total probability to belief functions, as finite random sets. Our results apply to the case in which Dempster’s conditioning is employed. We show that the solution to the resulting total belief problem is in general not unique, whereas it is unique when the a-priori belief function is Bayesian. Examples and case studies underpin the theoretical contributions. Finally, our results are compared to previous related work on the generalisation of Jeffrey’s rule by Spies and Smets.
In this paper we prove that a recent Bayesian approximation of belief functions, the relative belief of singletons, meets a number of properties with respect to Dempster’s rule of combination which mirrors those satisfied by the relative plausibility of singletons. In particular, its operator commutes with Dempster’s sum of plausibility functions, while perfectly representing a plausibility function when combined through Dempster’s rule. This suggests a classification of all Bayesian approximations into two families according to the operator they relate to.
Currently all video search engines are text-based, i.e. they search for the text labels associated with any video to retrieve the desired ones. However, this can lead to incorrect or inaccurate results, as labelling or annotating a video is mainly done manually. Consequently, many false positive results are generated during video searches by mislabelled videos.To solve this problem we need to improve the process of video annotation. This can be achieved by automatic annotation of videos based on their actual content, rather than text labels or tags. To accomplish this we need to enable computers to extract video “storylines”, composed by the events or actions taking place in each video. This has the potential to save time and provide better results for online video searches, as well as improve event detection in real-world surveillance footage. The project aims to facilitate Probabilistic Semantic Search and Query Answering by annotating videos in the way described, through machine lea
Adequate testing of AI applications is essential to ensure their quality. However, it is often prohibitively difficult to generate realistic test cases or to check software correctness. This paper proposes a new method called datamorphic testing, which consists of three components: a set of seed test cases, a set of datamorphisms for transforming test cases, and a set of metamorphisms for checking test results. With an example of face recognition application, the paper demonstrates how to develop datamorphic test frameworks, and illustrates how to perform testing in various strategies, and validates the approach using an experiment with four real industrial applications of face recognition.
Current state-of-the-art action detection systems are tailored for offline batch-processing applications. However, for online applications like human-robot interaction, current systems fall short. In this work, we introduce a real-time and online joint-labelling and association algorithm for action detection that can incrementally construct space-time action tubes on the most challenging untrimmed action videos in which different action categories occur concurrently. In contrast to previous methods, we solve the linking, action labelling and temporal localization problems jointly in a single pass. We demonstrate superior online association accuracy and speed (1.8ms per frame) as compared to the current state-of-the-art offline and online systems.