Incremental Linear Discriminant Analysis Using Sufficient Spanning Sets and Its Applications

This paper presents an incremental learning solution for Linear Discriminant Analysis (LDA) and its applications to object recognition problems. We apply the sufficient spanning set approximation in three steps i.e. update for the total scatter matrix, between-class scatter matrix and the projected data matrix, which leads an online solution which closely agrees with the batch solution in accuracy while significantly reducing the computational complexity. The algorithm yields an efficient solution to incremental LDA even when the number of classes as well as the set size is large. The incremental LDA method has been also shown useful for semi-supervised online learning. Label propagation is done by integrating the incremental LDA into an EM framework. The method has been demonstrated in the task of merging large datasets which were collected during MPEG standardization for face image retrieval, face authentication using the BANCA dataset, and object categorisation using the Caltech101 dataset.     

Coding of watershed and river hierarchy to support GIS-based hydrological analyses at different scales

A procedure is presented to automatically assign a modified Pfafstetter code to a stream network and its associated watersheds. The first step is to build a topological network according to the object-oriented Arc Hydro framework definition in ArcGIS. These network relations are then mapped into a code that fully represents the stream hierarchy of a basin in a single attribute. Its use does not depend on ArcGIS object orientation and facilitates simple hierarchical aggregation and querying using only the attribute tables of shapefile versions of the stream network and watershed datasets. The code was implemented for the whole of Austria to support the flexible presentation and user-defined aggregation of water balance results in the digital version of the Hydrological Atlas of Austria.     

Evaluation of six night vision enhancement systems: qualitative and quantitative support for intelligent image processing

OBJECTIVE: An evaluation study was conducted to answer the question of which system properties of night vision enhancement systems (NVESs) provide a benefit for drivers without increasing their workload. BACKGROUND: Different infrared sensor, image processing, and display technologies can be integrated into an NVES to support nighttime driving. Because each of these components has its specific strengths and weaknesses, careful testing is required to determine their best combination. METHOD: Six prototypical systems were assessed in two steps. First, a heuristic evaluation with experts from ergonomics, perception, and traffic psychology was conducted. It produced a broad overview of possible effects of system properties on driving. Based on these results, an experimental field study with 15 experienced drivers was performed. Criteria used to evaluate the development potential of the six prototypes were the usability dimensions of effectiveness, efficiency, and user satisfaction (International Organization for Standardization, 1998). RESULTS: Results showed that the intelligibility of information, the easiness with which obstacles could be located in the environment, and the position of the display presenting the output of the system were of crucial importance for the usability of the NVES and its acceptance. Conclusion: All relevant requirements are met best by NVESs that are positioned at an unobtrusive location and are equipped with functions for the automatic identification of objects and for event-based warnings. APPLICATION: These design recommendations and the presented approach to evaluate the systems can be directly incorporated into the development process of future NVESs.     

Constructive Membership Predicates as Index Types

In the constructive setting, membership predicates over recursive types are inhabited by terms indexing the elements that satisfy the criteria for membership. In this paper, we motivate and explore this idea in the concrete setting of lists and trees. We show that the inhabitants of membership predicates are precisely the inhabitants of a generic shape type. We show that membership of x (of type T) in structure S, (x_TS) can not, in general, index all parts of a structure S and we generalize to a form ρS where ρ is a predicate over S. Under this scheme, (λx.True)S is the set of all indexes into S, but we show that not all subsets of indexes are expressible by strictly local predicates. Accordingly, we extend our membership predicates to predicates that retain state “from above” as well as allow “looking below”. Predicates of this form are complete in the sense that they can express every subset of indexes in S. These ideas are motivated by experience programming in Nuprl's constructive type theory and examining the constructive content of mechanically checked formal proofs involving membership predicates.     

Obituary

RILEM Bulletin

Obituary     

Inverse random under sampling for class imbalance problem and its application to multi-label classification

In this paper, a novel inverse random under sampling (IRUS) method is proposed for the class imbalance problem. The main idea is to severely under sample the majority class thus creating a large number of distinct training sets. For each training set we then find a decision boundary which separates the minority class from the majority class. By combining the multiple designs through fusion, we construct a composite boundary between the majority class and the minority class. The proposed methodology is applied on 22 UCI data sets and experimental results indicate a significant increase in performance when compared with many existing class-imbalance learning methods. We also present promising results for multi-label classification, a challenging research problem in many modern applications such as music, text and image categorization.     

Consensus in the presence of mortal Byzantine faulty processes

We consider the problem of reaching agreement in distributed systems in which some processes may deviate from their prescribed behavior before they eventually crash. We call this failure model “mortal Byzantine”. After discussing some application examples where this model is justified, we provide matching upper and lower bounds on the number of faulty processes, and on the required number of rounds in synchronous systems. We then continue our study by varying different system parameters. On the one hand, we consider the failure model under weaker timing assumptions, namely for partially synchronous systems and asynchronous systems with unreliable failure detectors. On the other hand, we vary the failure model in that we limit the occurrences of faulty steps that actually lead to a crash in synchronous systems.