HKUST Institutional Repository
(5.016 recursos)
Repository of Hong Kong University of Science and Technology. Managed by the HKUST Library.
Mostrando recursos 1 - 20 de 116
1.
A computationa theory of decision networks - Zhang, Nevin Lianwen
This thesis is about how to represent and solve decision problems in Bayesian decision theory (e.g. Fishburn 1988). A general representation named decision networks is proposed based on influence diagrams (Howard and Matheson 1984). This new representation incorporates the idea, from Markov decision process (e.g. Denardo 1982), that a decision may be conditionally independent of certain pieces of available information. It also allows multiple cooperative agents and facilitates the exploitation of separability in the utility function. Decision networks inherit the advantages of both influence diagrams and Markov decision processes.
Both Influence diagrams and finite stage Markov decision processes are stepwise-solvable, in...
2.
Context constraints for automated compositional reachability analysis - Cheung, Shing-Chi; Kramer, Jeff
Behaviour analysis of complex concurrent systems has led to the search for enhanced reachability analysis techniques which support modularity and which control the state explosion problem. While modularity has been achieved, state explosion is still a problem. Indeed, this problem may even be exacerbated as a locally minimised subsystem may contain many states and transitions forbidden by its environment or context. Context constraints, specified as interface processes, are restrictions imposed by the environment on subsystem behaviour. Recent research has suggested that the state explosion problem can be effectively controlled if context constraints are incorporated in compositional reachability analysis (CRA). Although...
3.
Checking subsystem safety properties in compositional reachability analysis - Cheung, Shing-Chi; Kramer, Jeff
The software architecture of a distributed system can be described as a hierarchical composition of subsystems, with interacting processes as the leaves of the hierarchy. Process behaviour can be specified using finite-state machines. A global state machine describing the overall system behaviour can be constructed using compositional reachability analysis techniques. These techniques compose the global state machine of a system from its component processes in stages, based on the specified hierarchy. The key to the success of these analysis techniques is to employ a modular software architecture and hide as many internal actions as possible in each subsystem. A subsystem...
4.
Integrated behaviour analysis of distributed systems - Cheung, Shing-Chi; Kramer, Jeff
Behaviour analysis is useful at all stages in the design and maintenance of well-behaved distributed systems. Dataflow and reachability analyses are two orthogonal but complementary behaviour analysis techniques. Individually, each of these techniques may be inadequate for the analysis of large-scale distributed systems. On the one hand, dataflow analysis algorithms, while tractable, may not be sufficiently accurate to provide meaningful detection of errors. On the other hand, reachability analysis, while providing exhaustive analysis, may be computationally too expensive for complex systems. In this paper, we present a method which integrates a dataflow and a reachability analysis technique to provide a...
5.
A delay scheduling strategy for dynamic load balancing schedulers - Hui, Chi-Chung; Chanson, Samuel T.
This paper investigates the issue of practical load balancing in the highly dynamic environ-ment of local area networks (LANs). Most existing dynamic load balancing techniques dispatch the jobs immediately upon arrival irrespective of the overall loading of the LAN. This may lead to system saturation and thrashing. Moreover, these schemes focus on steady state system throughput without considering the behavior in the transient periods. As a result, they may not deliver satisfactory performance in practice where jobs may arrive in batches and system loading may fluctuate widely. To tackle the above problems, the delay scheduling strategy is proposed to dynamically...
6.
SNR : sensitive DCA in multicell environment - Choy, Manhoi
A number of dynamic channel allocation (DCA) protocols are proposed in this paper for a typical multicell environment. These protocols make use of the receiving signal strength and signal to noise ratio (SNR) to assign channels without violating the minimum requirements for reliable communications. A corresponding model based on SNR is developed to study these protocols. Significant improvement to the overall system performance is obtained when compared with existing protocols and models that are based on the concept of minimum reuse distance.
7.
A theory of the visual motion coding in the primary visual cortex - Li, Zhaoping
This paper demonstrates that much of the visual motion coding in the primary visual cortex can be understood from a theory of efficient motion coding in the multiscale representation. The theory predicts that cortical cells can have a spectrum of directional indices, are tuned to different directions of motion, and their receptive fields (RF) can be spatio-temporally separable or inseparable. The predictions also include the following correlations between the motion coding and spatial, chromatic, and stereo codings: the preferred speed is larger when the cell receptive field sizes are larger: the color channel prefers smaller speed than the luminance channel,...
8.
The management of interdependent asynchronous transactions in heterogeneous database environments - Do, Lyman; Drew, Pamela
Today, enterprises maintain many, disparate information sources over which complex business procedures and applications are executed. The informal and ad hoc characteristics of these environments make the information in them very prone to inconsistency. Yet, the flexibility of application execution given to different parts of an organization is desirable. This paper introduces a new mechanism in which the execution of asynchronous, pre-existing, yet related, applications can be harnessed. A multidatabase framework that supports the concurrent execution of these applications with heterogeneous, distributed transactions is presented. The types of inconsistency that can arise in this framework are categorized and methods to...
9.
Han character font sharing across incompatible bitmap file formats - Pong, Man-Chi; Lee, Fung Fung; Yeung, Ricky; Zhang, Yongguang
We identify the problem of sharing Han character font across incompatible bitmap file formats and discuss the related issues. The ideal solution should meet the requirements including independence of character code schemes, such as GB2312-80, Big5, and Unicode, no conversion of existing bitmap font files, storage efficiency, flexibility, extensibility, platform independence, as well as simplicity. Basic principles of how to achieve the above requirements are described. As a realization of the ideas, a new Han character bitmap font specification format, called HBF, has been developed and is presented here. In fact, the HBF approach allows the sharing of fixed-pitch fonts...
10.
Transient responses in dynamical neural models - Stiber, Michael; Segundo, Jose P.
We consider the input/output behavior of a realistic dynamical neural model in comparison to those typically used in artificial neural networks. We have found that such models duplicate well those behaviors seen in living neurons, displaying a range of behaviors commonly seen in a wide variety of nonlinear dynamical systems. This is not captured well by weighted sum/monotonic transfer function models. An example of the consequences of nonlinear dynamics in neural responses is presented for monotonically changing input transients.
11.
A corrigendum for the optimized-IPIA - Kean, Alex; Tsiknis, George
In a paper published earlier by this Journal (Kean and Tsiknis, 1990), we presented an incremental algorithm for generating prime implicants/implicates, called IPIA. In an attempt to improve IPIA some optimizations were also presented. It was pointed out to us that one of these improvements, is in conflict with the others and produces incorrect results. This corrigendum provides a solution to this problem.
12.
Checking and correcting safety properties using compositional reachability analysis - Cheung, Shing-Chi
The software architecture of a distributed program can be represented by a hierarchical composition of subsystems, with interacting processes at the leaves of the hierarchy. Compositional reachability analysis (CRA) is a promising state reduction technique which can be automated and used in stages to derive the overall behaviour of a distributed program based on its architecture. CRA is particularly suitable for the analysis of programs which are subject to evolutionary change. When a program evolves, only the behaviours of those subsystems affected by the change need be re-evaluated. The technique however has a limitation. The properties available for analysis are...
13.
A probabilistic query language - Wuthrich, Beat
We define a new query language for data and knowledge bases. This language is able to deal with weighted or uncertain base and derived information and reduces to a known query anguage in case that all information is deterministic. We show that this language is probabilistic in the sense of axiomatic probability theory. The novelty is that this language is non-propositional, able to deal with function symbols and offering extensibility through user defined built-in functions. Issues of knowledge acquisition are also addressed. We show where weighted base information can come from and how uncertain derived information might be assimilated.
14.
Managing time and uncertainty - Wuthrich, Beat
Similarities between the concepts of time and uncertainty are worked out. Using a functional and object-oriented query language we present techniques of how to enhance a query language to manage time and/or uncertainty. These results can be used to define semantics for syntactically small but extremely powerful languages enhanced with two dimensions: time or any dimension with a total ordering, and uncertainty which can be interpreted in probabilistic or any other way.
15.
An architecture for homogenizing federated databases - Karlapalem, Kamalakar; Li, Qing; Chung, Dak Shum
Many large organizations have a number of heterogeneous localized databases that are maintained and managed to support a set of local applications. One of the main difficulties in supporting global applications over a number of local-ized databases and migrating legacy ISs to modern computing environment is to cope with the heterogeneities of these systems. We propose a novel flexible architecture (HODFA) to dynamically connect these localized heterogeneous databases in forming a homogenized federated database system and to support the process of transforming a collection of heterogeneous information systems onto a homogeneous environment. We further develop an incremental method-ology of homogenization...
16.
Information filtering for planning in partially observable stochastic domains - Zhang, Nevin Lianwen; Boerlage, Brent
Partially observable decision processes (POMDP) can be used as a model for planning in stochastic domains. This paper considers the problem of computing an optimal policy for a finite horizon POMDP. The task is difficult because the decision at any time point depends upon information from all previous time points. We propose to filter out inconsistencies and insignificant details in the collection of information being passed from one time point to the next. This reduces the number of possible information states and hence speeds up computation. A bound on the sacrifice of optimality due to information filtering is given, which...
17.
Asynchronous consistency restoration under epsilon serializability - Drew, Pamela; Pu, Calton
Many database applications tolerate a certain amount of data inconsistency to gain increased concurrent processing and to accommodate real-world constraints. This paper describes how inconsistency can be modeled in a database and managed with consistency restoration methods. The correctness criterion for the maintenance of inconsistency is epsilon serializability (ESR). An informal notation to characterize inconsistency and several consistency restoration techniques are described.
18.
ASCOC: a recurrent neural network model for grammatical inference - Yeung, Dit-Yan; Yeung, Kei-Wai
Many real-world problems require the handling of temporally correlated events. Recently, some discrete-time recurrent neural network models have been proposed to solve various temporal sequence processing problems. One of the most popular models is Elman's _simple recurrent network_ (SRN), which has been used for grammatical inference (or formal language learning) problems. In this paper, some major drawbacks of SRN and its variants are identified and investigated in detail. Based on these observations, a new model called ASCOC, which stands for _auto-associative SRN with context-to-output connections_, is proposed to overcome the problems encountered by SRN. Two experiments in using the ASCOC...
19.
Auto-associative learning of on-line handwriting using recurrent neural networks - Yeung, Dit-Yan; Yeung, Kei-Wai
Traditionally, the parametric grammar-based approach to the modeling and recognition of temporal sequences using hidden Markov models (HMMs) involves a very crucial step, which requires human experts to determine a priori the appropriate model architecture to work on. This includes, among other things, determining the number of states in the (probabilistic) grammar and the (probabilistic) transitions between states. As a long-term effort, we attempt to develop a more domain-independent, principled approach to the modeling of grammatical structures in temporal sequences, without knowing in advance the topology of the underlying grammars. This is achieved through an unsupervised learning process. In particular,...
20.
Document ranking on weight-partitioned signature files - Lee, Dik Lun; Ren, Liming
A signature file organization, called the weight-partitioned signature file, for supporting document ranking is proposed.
It employs multiple signature files, each of which corresponds to one term frequency, to represent terms with different term frequencies. Words with the same term frequency in a document are grouped together and hashed into the signature file corresponding to that term frequency. This eliminates the need to explicitly record the term frequency for each word.
We investigate the effect of false drops on retrieval effectiveness if they are not eliminated in the search process.
We have shown that false drops introduce insignificant degradation on precision and recall...
1.
