SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
In this paper, we descrive our motivation and approach to devloping models and the neural network architecture. Initial use of the artificial neural network for modeling the single plant process of transpiration is presented.
Explore the source record for details and available documents.
Air transport forms complex networks that can be measured in order to understand its structural characteristics and functional properties. Recent models for network growth (i.e., preferential attachment, etc.) remain stochastic and do not seek to understand other network-specific mechanisms that may account for their development in a more microscopic way. Air traffic is made up of many constituent airlines that are either privately or publicly owned and that operate their own networks. They follow more or less similar business policies each. The way these airline networks organize among themselves into distinct traffic distributions reveals complex interaction among them, which in turn can be aggregated into larger (macro-) traffic distributions. Our approach allows for a more deterministic methodology that will assess the impact of airline strategies on the distinct distributions for air traffic, particularly inside Europe. One key question this paper is seeking to answer is whether there are distinct patterns of preferential attachment for given classes of airline networks to distinct types of European airports. Conclusions about the advancing degree of concentration in this industry and the airline operators that accelerate this process can be drawn.
The use of an appropriate relationship model is critical for reliable prediction of future urban growth. Identification of proper variables and mathematic functions and determination of the weights or coefficients are the key tasks for building such a model. Although the conventional logistic regression model is appropriate for handing land use problems, it appears insufficient to address the issue of interdependency of the predictor variables. This study used an alternative approach to simulation and modeling urban growth using artificial neural networks. It developed an operational neural network model trained using a robust backpropagation method. The model was applied in the Myrtle Beach region of South Carolina, and tested with both global datasets and areal datasets to examine the strength of both regional models and areal models. The results indicate that the neural network model not only has many theoretic advantages over other conventional mathematic models in representing the complex urban systems, but also is practically superior to the logistic model in its capability to predict urban growth with better - accuracy and less variation. The neural network model is particularly effective in terms of successfully identifying urban patterns in the rural areas where the logistic model often falls short. It was also found from the area-based tests that there are significant intra-regional differentiations in urban growth with different rules and rates. This suggests that the global modeling approach, or one model for the entire region, may not be adequate for simulation of a urban growth at the regional scale. Future research should develop methods for identification and subdivision of these areas and use a set of area-based models to address the issues of multi-centered, intra- regionally differentiated urban growth.
We propose a modified diffusion-limited aggregation (DLA) model for the evolution of fluvial drainage networks. Random walkers are introduced randomly on a grid, and each two-dimensional random walk proceeds until the walker finds a drainage network on which to accrete. This model for headward growth of drainage networks generates drainage patterns remarkably similar to actual drainages. The model also predicts statistical features which agree with actual networks, including the frequency-order (bifurcation) ratio (R(sub b) = 3.98) and the stream length-order (R(sub r) = 2.09). Using the definition of network fractal dimension D = log R(sub b)/log R(sub r), we find that our DLA model gives D = 1.87, near the observed range of D approximately equal to 1.80 - 1.85.
Session MP2 includes short reports on: (1) Crew Regenerative Life Support in Long Duration Space Missions; (2) Bioconversion Systems for Food and Water on Long Term Space Missions; (3) Novel Laboratory Approaches to Multi-purpose Aquatic Biogenerative Closed-Loop Food Production Systems; and (4) Artificial Neural Network Derived Plant Growth Models.
Simulation models for LANs to be used in the proposed Space Station Training Facility, including models developed for Ethernet, Starlan, token ring, and Fiber Distributed Data Interface LANs, are discussed. Results of model runs for the basic configuration, for burst traffic, and for a growth scenario are presented. The PAWS and Network II.5 modeling tools are compared.
The pattern of erosion of a plateau along an escarpment may be modeled usng cluster growth techniques, recently popularized in models of drainage network evolution. If erosion on the scarp takes place in discrete events at rates subject to local substrate strength, the whole range of behavior is described by a combination of three cluster growth mechanisms: invasion percolation, Eden growth and diffusion-limited aggregation (DLA). These model the relative importance of preexisting substrate strength, background weathering, and seepage weathering and erosion respectively. The rate of seepage processes is determined by the efflux of groundwater at the plateau margin, which in turn is determined by the pressure field in the plateau aquifer. If this process acted alone, it would produce erosion patterns in the form of Laplacian fractals, with groundwater recharge from a distant source, or Poissionian fractals, with groundwater recharge uniform over the plateau. DLA is used to mimic the Laplacian or Poissonian potential field and the corresponding seepage growth process. The scaling structure of clusters grown by pure DLA, invasion percolation, or Eden growth is well known; this study presents a model which combines all three growth mechanisms for the first time. Mixed growth processes create clusters with different scaling properties and morphologies over distinct length scale ranges, and this is demonstrable in natural examples of plateau erosion.
The human visual system comprises layers of networks which sample, process, and code images. Understanding these networks is a valuable means of understanding human vision and of designing autonomous vision systems based on network processing. Ames Research Center has an ongoing program to develop computational models of such networks. The models predict human performance in detection of targets and in discrimination of displayed information. In addition, the models are artificial vision systems sharing properties with biological vision that has been tuned by evolution for high performance. Properties include variable density sampling, noise immunity, multi-resolution coding, and fault-tolerance. The research stresses analysis of noise in visual networks, including sampling, photon, and processing unit noises. Specific accomplishments include: models of sampling array growth with variable density and irregularity comparable to that of the retinal cone mosaic; noise models of networks with signal-dependent and independent noise; models of network connection development for preserving spatial registration and interpolation; multi-resolution encoding models based on hexagonal arrays (HOP transform); and mathematical procedures for simplifying analysis of large networks.
We present the current form of a provisional DNA sequence-based regulatory gene network that explains in outline how endomesodermal specification in the sea urchin embryo is controlled. The model of the network is in a continuous process of revision and growth as new genes are added and new experimental results become available; see http://www.its.caltech.edu/~mirsky/endomeso.htm (End-mes Gene Network Update) for the latest version. The network contains over 40 genes at present, many newly uncovered in the course of this work, and most encoding DNA-binding transcriptional regulatory factors. The architecture of the network was approached initially by construction of a logic model that integrated the extensive experimental evidence now available on endomesoderm specification. The internal linkages between genes in the network have been determined functionally, by measurement of the effects of regulatory perturbations on the expression of all relevant genes in the network. Five kinds of perturbation have been applied: (1) use of morpholino antisense oligonucleotides targeted to many of the key regulatory genes in the network; (2) transformation of other regulatory factors into dominant repressors by construction of Engrailed repressor domain fusions; (3) ectopic expression of given regulatory factors, from genetic expression constructs and from injected mRNAs; (4) blockade of the beta-catenin/Tcf pathway by introduction of mRNA encoding the intracellular domain of cadherin; and (5) blockade of the Notch signaling pathway by introduction of mRNA encoding the extracellular domain of the Notch receptor. The network model predicts the cis-regulatory inputs that link each gene into the network. Therefore, its architecture is testable by cis-regulatory analysis. Strongylocentrotus purpuratus and Lytechinus variegatus genomic BAC recombinants that include a large number of the genes in the network have been sequenced and annotated. Tests of the cis-regulatory predictions of the model are greatly facilitated by interspecific computational sequence comparison, which affords a rapid identification of likely cis-regulatory elements in advance of experimental analysis. The network specifies genomically encoded regulatory processes between early cleavage and gastrula stages. These control the specification of the micromere lineage and of the initial veg(2) endomesodermal domain; the blastula-stage separation of the central veg(2) mesodermal domain (i.e., the secondary mesenchyme progenitor field) from the peripheral veg(2) endodermal domain; the stabilization of specification state within these domains; and activation of some downstream differentiation genes. Each of the temporal-spatial phases of specification is represented in a subelement of the network model, that treats regulatory events within the relevant embryonic nuclei at particular stages. (c) 2002 Elsevier Science (USA).
Our focus is the evolution of business strategies and network structure decisions in the commercial passenger aviation industry. The paper reviews the growth of hub-and-spoke networks as the dominant business model following deregulation in the latter part of the 20 century, followed by the emergence of value-based airlines as a global phenomenon at the end of the century. The paper highlights the link between airline business strategies and network structures, and examines the resulting competition between divergent network structure business models. In this context we discuss issues of market structure stability and the role played by competition policy.
As Internet traffic continues to grow exponentially, developing a clearer understanding of, and appropriately measuring, network's performance is becoming ever more critical. An important challenge faced by the Information Resources Directorate (IRD) at the Johnson Space Center in this context remains not only monitoring and maintaining a secure network, but also better understanding the capacity and future growth potential boundaries of its network. This requires capacity planning which involves modeling and simulating different network alternatives, and incorporating changes in design as technologies, components, configurations, and applications change, to determine optimal solutions in light of IRD's goals, objectives and strategies. My primary task this summer was to address this need. I evaluated network-modeling tools from OPNET Technologies Inc. and Compuware Corporation. I generated a baseline model for Building 45 using both tools by importing "real" topology/traffic information using IRD's various network management tools. I compared each tool against the other in terms of the advantages and disadvantages of both tools to accomplish IRD's goals. I also prepared step-by-step "how to design a baseline model" tutorial for both OPNET and Compuware products.
The emergence of early metabolism - a network of catalyzed chemical reactions that supported self-maintenance, growth, reproduction and evolution of the ancestors of contemporary cells (protocells) was a critical, but still very poorly understood step on the path from inanimate to animate matter. Here, it is proposed and tested through mathematical modeling of biochemically plausible systems that the emergence of metabolism and its initial evolution towards higher complexity preceded the emergence of a genome. Even though the formation of protocellular metabolism was driven by non-genomic, highly stochastic processes the outcome was largely deterministic, strongly constrained by laws of chemistry. It is shown that such concepts as speciation and fitness to the environment, developed in the context of genomic evolution, also held in the absence of a genome.
Neural networks are a very useful tool for analyzing and modeling complex real world systems. Applying neural network simulations to real world problems generally involves large amounts of data and massive amounts of computation. To efficiently handle the computational requirements of large problems, we have implemented at Los Alamos a highly efficient neural network compiler for serial computers, vector computers, vector parallel computers, and fine grain SIMD computers such as the CM-2 connection machine. This paper describes the mapping used by the compiler to implement feed-forward backpropagation neural networks for a SIMD (Single Instruction Multiple Data) architecture parallel computer. Thinking Machines Corporation has benchmarked our code at 1.3 billion interconnects per second (approximately 3 gigaflops) on a 64,000 processor CM-2 connection machine (Singer 1990). This mapping is applicable to other SIMD computers and can be implemented on MIMD computers such as the CM-5 connection machine. Our mapping has virtually no communications overhead with the exception of the communications required for a global summation across the processors (which has a sub-linear runtime growth on the order of O(log(number of processors)). We can efficiently model very large neural networks which have many neurons and interconnects and our mapping can extend to arbitrarily large networks (within memory limitations) by merging the memory space of separate processors with fast adjacent processor interprocessor communications. This paper will consider the simulation of only feed forward neural network although this method is extendable to recurrent networks.
The NASA Plant Water Management (PWM) technology demonstrations aboard ISS apply recent advances in microgravity capillary fluidics research towards the mundane yet problematic challenges of simply watering plants in space. Plant growth in a low-g environment is often hampered by inadequate aeration and over-saturation of the root zone. The present effort aims to exploit the passive capillary forces of poorly wetting liquids (i.e., contaminated water) within unique system geometries that effectively replace the role of gravity in providing sufficient aeration and hydration for simulated plants. Several flight demonstrations have been completed on ISS, including soil and hydroponic models in single and parallel channel networks. Two future demonstrations are still in work and plan to further develop the system to handle sustained plant growth with additional sensors to monitor the growth environment. The results to date demonstrate proof-of-concept, system stability, limits of operation, and more, for simulated plant models. Eventually, real plants will be incorporated into these systems and tested on orbit. The implications are discussed in relation to plant growth facilities for further near-term microgravity plant science research as well as for automated food production for long duration human exploration missions.
The past few years have witnessed a rapid growth of interest in a cluster of modes of modeling and computation which may be described collectively as soft computing. The distinguishing characteristic of soft computing is that its primary aims are to achieve tractability, robustness, low cost, and high MIQ (machine intelligence quotient) through an exploitation of the tolerance for imprecision and uncertainty. Thus, in soft computing what is usually sought is an approximate solution to a precisely formulated problem or, more typically, an approximate solution to an imprecisely formulated problem. A simple case in point is the problem of parking a car. Generally, humans can park a car rather easily because the final position of the car is not specified exactly. If it were specified to within, say, a few millimeters and a fraction of a degree, it would take hours or days of maneuvering and precise measurements of distance and angular position to solve the problem. What this simple example points to is the fact that, in general, high precision carries a high cost. The challenge, then, is to exploit the tolerance for imprecision by devising methods of computation which lead to an acceptable solution at low cost. By its nature, soft computing is much closer to human reasoning than the traditional modes of computation. At this juncture, the major components of soft computing are fuzzy logic (FL), neural network theory (NN), and probabilistic reasoning techniques (PR), including genetic algorithms, chaos theory, and part of learning theory. Increasingly, these techniques are used in combination to achieve significant improvement in performance and adaptability. Among the important application areas for soft computing are control systems, expert systems, data compression techniques, image processing, and decision support systems. It may be argued that it is soft computing, rather than the traditional hard computing, that should be viewed as the foundation for artificial intelligence. In the years ahead, this may well become a widely held position.
The Huntsville Operations Support Center (HOSC) is a distributed computer system used to provide real time data acquisition, analysis and display during NASA space missions and to perform simulation and study activities during non-mission times. The primary purpose is to provide a HOSC system simulation model that is used to investigate the effects of various HOSC system configurations. Such a model would be valuable in planning the future growth of HOSC and in ascertaining the effects of data rate variations, update table broadcasting and smart display terminal data requirements on the HOSC HYPERchannel network system. A simulation model was developed in PASCAL and results of the simulation model for various system configuraions were obtained. A tutorial of the model is presented and the results of simulation runs are presented. Some very high data rate situations were simulated to observe the effects of the HYPERchannel switch over from contention to priority mode under high channel loading.
The architecture of the C23A (Chambers de Culture Automatique en Atmosphere Artificielles) system for the controlled study of plant physiology is described. A modular plant growth chambers and associated instruments (I.R. CO2 analyser, Mass spectrometer and Chemical analyser); network of frontal processors controlling this apparatus; a central computer for the periodic control and the multiplex work of processors; and a network of terminal computers able to ask the data base for data processing and modeling are discussed. Examples of present results are given. A growth curve analysis study of CO2 and O2 gas exchanges of shoots and roots, and daily evolution of algal photosynthesis and of the pools of dissolved CO2 in sea water are discussed.