Shared workspaces for group collaboration: an experiment usingInternet and UNIX interprocess communications

The system described is intended to link institutions (such as universities or industrial research organizations) that have most of their computers connected to local area networks (LANs). The main objective of the research is to bring users of remote systems together to collaborate in reviewing and editing documents containing text, graphs and image objects. The term workspace is used to denote a collection of such objects belonging to some application, and the software tools needed to access these objects. The major components of the system and the user interface are presented. A prototype implemented in C under 4.3BSD UNIX is discussed. The system is then related to and contrasted with other work on group collaboration. The suitability and shortcomings of UNIX as an operating system for building group collaboration software tools are assessed     

A refined cut-through buffer management scheme for layered protocolstacks

Avoiding recopying of user data as it is passed between protocol layers helps break the protocol-execution bottleneck. The authors use a use a simple pointer-variable updating technique in the sending phase of their refined buffer cut-through scheme. Binary tree traversal is not required. Time is saved in allocating (and later deallocating) buffers for storing usually short data transfer protocol headers. Their scheme saves working buffer space for segmenting and linking of protocol data units. Because a simple linked list is constructed in the receiving phase, significant working buffer space can be saved. Each node of the linked list points to a user data segment. During the linked-list traversal, user data is copied directly to the user process when a node is visited     

Class decomposition for GA-based classifier agents--a Pitt approach.

This paper proposes a class decomposition approach to improve the performance of GA-based classifier agents. This approach partitions a classification problem into several class modules in the output domain, and each module is responsible for solving a fraction of the original problem. These modules are trained in parallel and independently, and results obtained from them are integrated to form the final solution by resolving conflicts. Benchmark classification data sets are used to evaluate the proposed approaches. The experiment results show that class decomposition can help achieve higher classification rate with training time reduced.     

An incremental approach to genetic-algorithms-based classification.

Incremental learning has been widely addressed in the machine learning literature to cope with learning tasks where the learning environment is ever changing or training samples become available over time. However, most research work explores incremental learning with statistical algorithms or neural networks, rather than evolutionary algorithms. The work in this paper employs genetic algorithms (GAs) as basic learning algorithms for incremental learning within one or more classifier agents in a multiagent environment. Four new approaches with different initialization schemes are proposed. They keep the old solutions and use an "integration" operation to integrate them with new elements to accommodate new attributes, while biased mutation and crossover operations are adopted to further evolve a reinforced solution. The simulation results on benchmark classification data sets show that the proposed approaches can deal with the arrival of new input attributes and integrate them with the original input space. It is also shown that the proposed approaches can be successfully used for incremental learning and improve classification rates as compared to the retraining GA. Possible applications for continuous incremental training and feature selection are also discussed.     

Randomness quality of permuted pseudorandom binary sequences

This paper uses the DIEHARD statistical test suite to test the randomness quality of “permuted” versions of maximum length sequences generated by linear finite state machines (LFSM) such as cellular automata and linear feedback shift registers. Analysis shows that permuted sequences can be equivalently generated by using time-varying transformations derived from the original LFSM. Based on the above, we suggest the permuted transformation sequence scheme. Experimental results show that DIEHARD results are improved with respect to the original non-permuted sequences—up to seven more tests can be passed (total of 19 tests). Furthermore, a permutation vector is used to generate cyclically distinct permuted sequences and each sequence has a desirable maximum length period of 2ⁿ − 1.     

Reduced Pattern Training Based on Task Decomposition Using Pattern Distributor

Task decomposition with pattern distributor (PD) is a new task decomposition method for multilayered feedforward neural networks (NNs). Pattern distributor network is proposed that implements this new task decomposition method. We propose a theoretical model to analyze the performance of pattern distributor network. A method named reduced pattern training (RPT) is also introduced, aiming to improve the performance of pattern distribution. Our analysis and the experimental results show that RPT improves the performance of pattern distributor network significantly. The distributor module's classification accuracy dominates the whole network's performance. Two combination methods, namely, crosstalk-based combination and genetic-algorithm (GA)-based combination, are presented to find suitable grouping for the distributor module. Experimental results show that this new method can reduce training time and improve network generalization accuracy when compared to a conventional method such as constructive backpropagation or a task decomposition method such as output parallelism (OP).     

A golden-block-based self-refining scheme for repetitive patterned wafer inspections

Abstract. This paper presents a novel technique for detecting possible defects in two-dimensional wafer images with repetitive patterns using prior knowledge. The technique has a learning ability that can create a golden-block database from the wafer image itself, then modify and refine its content when used in further inspections. The extracted building block is stored as a golden block for the detected pattern. When new wafer images with the same periodical pattern arrive, we do not have to recalculate their periods and building blocks. A new building block can be derived directly from the existing golden block after eliminating alignment differences. If the newly derived building block has better quality than the stored golden block, then the golden block is replaced with the new building block. With the proposed algorithm, our implementation shows that a significant amount of processing time is saved. Also, the storage overhead of golden templates is reduced significantly by storing golden blocks only. Received: 21 February 2001 / Accepted: 21 April 2002 Correspondence to: S.-U. Guan     

An evolutionary approach to the design of controllable cellular automata structure for random number generation

Cellular automata (CA) has been used in pseudorandom number generation for over a decade. Recent studies show that two-dimensional (2-D) CA pseudorandom number generators (PRNGs) may generate better random sequences than conventional one-dimensional (1-D) CA PRNGs, but they are more complex to implement in hardware than 1-D CA PRNGs. In this paper, we propose a new class of 1-D CA - controllable cellular automata (CCA)-without much deviation from the structural simplicity of conventional 1-D CA. We first give a general definition of CCA and then introduce two types of CCA: CCA0 and CCA2. Our initial study shows that these two CCA PRNGs have better randomness quality than conventional 1-D CA PRNGs, but that their randomness is affected by their structures. To find good CCA0/CCA2 structures for pseudorandom number generation, we evolve them using evolutionary multiobjective optimization techniques. Three different algorithms are presented. One makes use of an aggregation function; the other two are based on the vector-evaluated genetic algorithm. Evolution results show that these three algorithms all perform well. Applying a set of randomness tests on the evolved CCA PRNGs, we demonstrate that their randomness is better than that of 1-D CA PRNGs and can be comparable to that of 2-D CA PRNGs.     

SAFE: secure agent roaming for e-commerce

The development of the Internet has made a powerful impact on the concept of commerce. E-commerce, a new way to conduct business, is gaining more and more popularity. Despite its rapid growth, there are limitations that hinder the expansion of e-commerce. The primary concern for most people when talking about online shopping is security. Due to the open nature of the Internet, personal financial details necessary for online shopping can be stolen if sufficient security mechanism is not put in place. How to provide the necessary assurance of security to consumers remains a question mark despite various past efforts. Another concern is the lack of intelligence. The Internet is an ocean of information depository. It is rich in content but lacks the necessary intelligent tools to help one locate the correct piece of information. Intelligent agent, a piece of software that can act on behalf of its owner intelligently, is designed to fill this gap. However, no matter how intelligent an agent is, if it remains on its owner's machine and does not have any roaming capability, its functionality is limited. With the roaming capability, more security concerns arise. In response to these concerns, SAFE, secure roaming agent for e-commerce, is designed to provide secure roaming capability to intelligent agents (Guan and Yang, 1999).     

Incremental multiple objective genetic algorithms

This paper presents a new genetic algorithm approach to multiobjective optimization problems--incremental multiple objective genetic algorithms (IMOGA). Different from conventional MOGA methods, it takes each objective into consideration incrementally. The whole evolution is divided into as many phases as the number of objectives, and one more objective is considered in each phase. Each phase is composed of two stages. First, an independent population is evolved to optimize one specific objective. Second, the better-performing individuals from the single-objecive population evolved in the above stage and the multiobjective population evolved in the last phase are joined together by the operation of integration. The resulting population then becomes an initial multiobjective population, to which a multiobjective evolution based on the incremented objective set is applied. The experiment results show that, in most problems, the performance of IMOGA is better than that of three other MOGAs, NSGA-II, SPEA, and PAES. IMOGA can find more solutions during the same time span, and the quality of solutions is better.