cms

In the first part of this talk, I would like to introduce my research group at my home university. The research in Software Engineering Group of Nanjing University focuses on formal modeling and verification, program analysis and testing, and the research work also covers areas such as aspect-oriented programming, model-driven architecture, service-oriented computing, and so on. Some selected research results and tools will be briefly introduced in this part of the talk.

Web services bridge the gap between heterogeneous systems. By providing formal support to web services, we can use existing research results to improve the understanding and increase the confidence to the services. In our work, we give a Petri-net based model to BPEL web services. Using this model, we perform verification on BPEL services against scenario-based specifications (such as message sequence charts and UML Sequence Diagrams). In addition to verification, we also focus on web service behavior extraction. Third party web services cannot always fit the customers' requirement, some critical behavior may not be guaranteed to appear in each of the service invocation. The behavior extraction is targeted to provide a wrapper service to ensure that the required behavior occurs in every execution of the target service. In the second part of the talk, the idea and approach of the behavior verification and extractions will be presented.

In (universal) algebra, G. Birkhoff proved his famous variety and completeness theorems in the 1930's which are results about equational logic. In the 1990's, a theory of coalgebras was developed which allows to reason abstractly about "systems", for example various kinds of transition systems and automata. Soon Birkhoff's theorems could be dualised, and one obtains a coequational logic which is about properties of such systems instead of properties of algebras.

In this talk we first review the notion of a cofree coalgebra (on the category of sets and functions); then we see how coequations and coequational logic are defined. In the main part, we present simple deduction systems for coequations that are sound and complete for a wide range of system types: we start with systems of polynomial type and extend to a subclass of accessible types using a representation technique for coequations.

A recurrent challenge that appears in enterprises is the need to enhance the functionality of their software eco-system by making some of the existing applications to interoperate with others. In the literature, this problem is known as Enterprise Application Integration (EAI) and is all about making two or more existing applications, that belongs to the same enterprise, to synchronize their data or to create new functionalities on top of them; in either case, the software that implements the integration is called the integration solution. This seminar will give an introduction on a DSL called Guaraná to design EAI solutions.

Local search algorithms represent a traditional heuristic approach to intractable computational problems. We addressed the inconvenience that a typical local search is not concerned with having a global vision over its own evolution, over its trajectory through the search space. We present two position-guided algorithms that work on top of a local search process - i.e. Tabu Search (TS). The goal is to guide the local search toward certain targeted regions of the search space. These techniques are focused on problems for which one can define a distance measure between candidate solutions.

The first algorithm (TS-Div) records its own trajectory and “pays attention” not to visit the same search space regions (spheres) repeatedly. The second algorithm (TS-Int) makes deep investigations in a “limited perimeter” around a given candidate solution. TS-Int employs a breath-first-search routine to enumerate all spheres from this “limited perimeter,” and each of these spheres is thoroughly explored by numerous independent TS processes. We experimentally observed that if such a “limited perimeter” contains a global optimum, TS-Int does not fail in eventually finding it. TS-Div ensures diversity, TS-Int enforces intensification, and together they reached very good results on a competitive problem (graph colouring). Details on the numerical experiments are available in a recent paper to be published by Computers & Operations Research.

The Two Dimensional Range Minimum Query (\mbox{2D-RMQ}) problem is to preprocess a static two dimensional array~$A$ of size $m\times n$, where~$m\le n$ such that subsequent queries asking for the index of the minimum element in a rectangular range within~$A$ can be answered efficiently. We show that every algorithm enabled to access~$A$ during the query and using~$O(N/c)$ bits additional space, requires query time~$\Omega(c)$ for $N=m\cdot n$ and any value of~\mbox{$1 \le c \le N$}. In particular this lower bound holds for the 1D-RMQ problem. We complement this lower bound with an algorithm that with~$O(N)$ preprocessing time and~$O(N/c)$ bits additional space achieves~$O(c\log2 c)$ query time. For~$c=1$, this is the first optimal algorithm using $O(N)$ bits additional space with $O(1)$ query time. We also consider the problem in the Encoding model where the query algorithms use an encoding data structure to solve the problem without utilizing~$A$. For this model, we present an~$O(1)$ query time algorithm using~$O(mn\log n)$ bits for the encoding data structure. We also give an alternative proof for the formerly known~$\Omega(mn\log m)$ bits for the size of the encoding data structure.