journal article
LitStream Collection
Home
doi: 10.1177/003754979105600403pmid: N/A
This column focuses on a single topic within the field of computer simulation. All the material comes directly from the electronic news group (bulletin board) called Simulation Digest. The Digest can be obtained over INTERNET by accessing the news group 'comp.simulation' or by sending e-mail to: 'simulation [email protected]'
doi: 10.1177/003754979105600406pmid: N/A
We simulate five neural networks on a vector multiprocessor. The training time can be reduced significantly especially when the training data size is large. These five neural networks are: 1) the ART1 network, 2) the ART 2 network, 3) the feedforward network, 4) the recurrent network, and 5) the Hopfield network. The training algorithms are programmed in such a way to best utilize 1) the inherent parallelism in neural computing, and 2) the vector and concur rent operations available on the parallel machine. To prove the correctness of parallelized training algorithms, each neural network is trained to perform a specific function. ART 1 and ART 2 are trained to recognize binary and analog patterns. The feedforward network is trained to perform the Fourier transform, the recurrent network is trained to predict the solution of a delay differential equation, the Hopfield network is trained to solve the traveling salesman problem. The machine we experiment with is the Alliant FX/80.
Abdel-Hamid, Tarek K.; Leidy, Frank H.
doi: 10.1177/003754979105600407pmid: N/A
Increasingly software quality assurance (QA) is being recognized as a critical factor in the successful development of software systems. However, because the utilization of QA tools and techniques adds significantly to the cost of developing software, the cost-effectiveness of QA has been a significant concern to the software quality manager. As of yet this concern has not been adequately addressed in the literature.The objective of this research effort is to develop an expert/simulator to support the QA effort allocation decision. The expert system module derives QA effort allocation schemes, feeds them into the simulation model for testing, and uses the feedback from the simulation results to improve upon the efficiency of the QA effort distribution.In a case study involv ing a NASA software project, the model accurately replicated the project's dynamic behavior. The expertlsimulator was then capable of deriving a more cost effective QA policy that achieved a 26% reduction in total project cost.
Forouzbakhsh, Farshid; Deiters, Robert M.; Kermanshahi, Bahman S.
doi: 10.1177/003754979105600408pmid: N/A
A computer algorithm for the optimal scheduling of generators in a power system is presented and tested. The algorithm, based on goal programming, automatically and dynami cally schedules the output of each generator in the system for optimal operation. The optimal operation can take into consideration multiple objectives such as economy, security, and reduction of pollution as well as practical constraints.To validate and test the algorithm, an example system of 5 generators, 10 busses, and 11 transmission lines is optimized for two objec tives: minimal generation cost and minimal emission of nitrous oxides (NOx). Hourly changes in total power demand in the range of 90% to 110% are considered together with a constraint of maximum permissible total NOx emission. Other practical equality and inequality constraints are incorporated into the optimiza tion algorithm.The simulation results demonstrate that the outputs of the generators can be changed smoothly and dynamically. Furthermore, using the algorithm, computer control is practicable either by direct on-line optimization or by using the feasible operation region generated as a small data-base by off-line computation.
doi: 10.1177/003754979105600409pmid: N/A
Researchers have felt that expert perfor mance must also rest on knowledge of deep models which relate underlying causal variables to observable facts. Simulation based upon causal knowledge is an impor tant method to infer possible consequences from given situations. This paper presents a rule-based causal simulation system called CAUSIM, which basically offers two kinds of simulation: backward simulation and forward simulation. Backward simulation is used to infer the instant behavior of specific attributes, whereas forward simulation is taken to arrive at possible overall scenarios. In addition, CAUSIM invokes constraint rules which describe incompatible behavior and values among related variables before applying simulation rules in order to obviate the inconsistencies between the simulation result and existing facts. The strength of CAUSIM lies in the capability of perform ing both qualitative and quantitative causal simulation in an integrated environment.
doi: 10.1177/003754979105600410pmid: N/A
This paper is intended to break the current research impasse on issues of mechanized thought. It bypasses the untractable recursive question: Is the human brain powerful enough to understand its own working? focusing on a more tractable:Can we model, and maybe induce the process of thought? Given that this modeling effort is a particular thought process, this second question is also recur sive in nature and more general: it permits us not to consider the limitations of our brains due to their biological implementa tion.In this article I model the concepts of complexity, error, surprise, intuition, thought and insight. I also analyze the concepts of consciousness and feelings. I describe the features any thinking system must have. It appears that humans might be able to create such systems in future.
Showing 1 to 10 of 11 Articles