We present an automatic iterative abstraction-refinement methodology in which the initial abstract model is generated by an automatic analysis of the control structures in the program to be verified. Abstract models may admit erroneous (or "spurious") counterexamples. We devise new symbolic techniques which analyze such counterexamples and refine the abstract model correspondingly.

these abstracts from the Eighteenth Annual Scientific Meeting may not present completed work nor were they formally peer-reviewed for technical content. Individuals wishing to reference or quote from these abstracts in whole or part should obtain the author's permission. Abstracts were optically scanned and then edited for entry into this compilation. But since the process is not perfect, errors may have been introduced, for which we apologize.

Neural networks have traditionally been applied to recognition problems, and most learning algorithms are tailored to those problems. We discuss the requirements of learning for generalization, where the traditional methods based on gradient descent have limited success. We present a new stochastic learning algorithm based on simulated annealing in weight space. We verify the convergence properties and feasibility of the algorithm. We also describe an implementation of the algorithm and validation experiments. 1. Introduction Neural networks are being applied to a wide variety of applications from speech generation[1], to handwriting recognition[2]. Last decade has seen great advances in design of neural networks for a class of problems called recognition problems, and in design of learning algorithms[3-5, 5-7]. The learning of weights for neural network for many recognition problem is no longer a difficult task. However, designing a neural network for generalization problem is ...

Abstract. In this paper, we revisit a general class of multi-criteria multi-constrained network design problems and attempt to solve, in a novel way, with Evolutionary Algorithms (EAs). A major challenge to solving such problems is to capture possibly all the (representative) equivalent and diverse solutions. In this work, we formulate, without loss of generality, a bi-criteria bi- constrained communication network topological design problem. Two of the primary objectives to be optimized are network delay and cost subject to satisfaction of reliability and flowconstraints. This is a NP-hard problem so we use a hybrid approach (for initialization of the population) along with EA. Furthermore, the twoobjective optimal solution front is not known a priori. Therefore, we use a multiobjective EA which produces diverse solution space and monitors convergence; the EA has been demonstrated to work effectively across complex problems of unknown nature. We tested this approach for designing networks of different sizes and found that the approach scales well with larger networks. Results thus obtained are compared with those obtained by two traditional approaches namely, the exhaustive search and branch exchange heuristics. 1

Abstract. Network-on-chip-based multiprocessor systems-onchip are considered as future embedded systems platforms. One of the steps in mapping an application onto such a parallel platform involves scheduling the communication on the network-on-chip. This paper presents different scheduling strategies that minimize resource usage by exploiting all scheduling freedom offered by networks-on-chip. Our experiments show that resource-utilization is improved when compared to existing techniques. 1.

Many numerical methods require the evaluation of Jacobians for vector functions given as evaluation procedures. If known, the sparsity pattern of these first derivative matrices can be used to evaluate, store, and manipulate them more efficiently. Especially on discretizations of differential equations and other large scale problems, detecting and specifying the pattern of potentially nonzero entries may be a laborious and error prone task.

1 In this paper, we consider the problem of space-time coding over Rician channels in the presence of spatio-temporal correlation. We derive an upper bound on the pairwise word error probability of space-time codes and use it as a basis for a unified approach to analysis and design of space-time codes over any flat, Rician or Rayleigh, block-fading channel. Based on the statistical properties of a Rician channel, the high SNR behavior of the bound is either exponential or rational. In the former case, design criteria for space-time codes are based on a Euclidean-distance-like measure. However, in the latter case, design criteria are based on a rank criterion together with a coding gain criterion. The analysis of the bound shows that the performance of rank-deficient codes can be highly degraded in the presence of spatio-temporal correlation, but the performance degradation is not severe for full-rank codes. These codes exhibit robustness against the channel correlation profile. Specifically, for channels with a nonsingular covariance matrix, the asymptotic performance of a full-rank code is proportional to its performance in the i.i.d. case. In the case that the only constraint on the codewords is a maximum energy constraint, we show

Detecting structural similarities between XML documents has been the subject of several recent work, and the proposed algorithms mostly use tree edit distance between the corresponding trees of XML documents. However, evaluating a tree edit distance is computationally expensive and does not easily scale up to large collections. We show in this paper that a tree edit distance computation often is not necessary and can be avoided. In particular, we propose a concise structural summary of XML documents and show that a comparison based on this summary is both fast and effective. Our experimental evaluation shows that this method does an excellent job of grouping documents generated by the same DTD, outperforming some of the previously proposed solutions based on a tree comparison. Furthermore, the time complexity of the algorithm is linear on the size of the structural description. 1

As technology advances and demand for higher performance increases hardware designs are becoming more and more sophisticated. A typical chip design may contain over ten million switching devices. Since the systems become more and more complex, detecting design errors for systems of such scale becomes extremely difficult. Formal verification methodologies can potentially catch subtle design errors. However, many state-of-the-art formal verification tools suffer from the state explosion problem. This thesis explores abstraction techniques to avoid the state explosion problem. In our methodology, atomic formulas extracted from an SMV-like concurrent program are used to construct abstraction functions. The initial abstract structure is built by using existential abstraction techniques. When the model checker disproves a universal property on the abstract structure, it generates a counterexample. However, this abstract counterexample might be spurious because abstraction is not complete. We provide a new symbolic algorithm to determine whether an abstract counterexample is spurious. When a counterexample is identified to be spurious, the algorithm will compute the shortest prefix of the abstract counterexample that does not correspond to an actual trace in the concrete model. The last abstract state in this prefix is split into less abstract states so that the spurious counterexample is eliminated. Thus, a more refined abstraction function is obtained. It is usually desirable to obtain the coarsest refinement which eliminates the counterexample because this corresponds to the smallest abstract model that avoids the spurious counterexample. We prove, however, that finding the coarsest refinement is NP-hard. Because of this, we use a polynomial-time algorithm which gives a su...

Network topology design problems find application in several real life scenarios. However, most designs in the past either optimize for a single criterion like delay or assume simplistic traffic models like Poisson. Such assumptions make the solutions inapplicable in the practical world. In this paper, we formulate and solve a multiobjective network topology design problem for a realistic Internet traffic model which is assumed to be self similar. We optimize for the average packet delivery delay and network layout cost to construct realistic network topologies. We present a multiobjective evolutionary algorithm (MOEA) to obtain the diverse near-optimal network topologies. For fair comparison, we design a multiobjective deterministic heuristic based on branch exchange – we call the heuristic Pareto Branch Exchange (PBE). We empirically show that the MOEA used performs well for real networks of various sizes, and generated topologies are quite different with significantly larger delays for the self similar traffic model.