Robert T. Miyamoto

Turning pictures into numbers

We present a study of complex visualization usage by expert meteorological forecasters. We performed a protocol analysis and examined the types of visualizations they examined. We present evidence for how experts are able to make use of complex visualizations. Our findings suggest that users of complex visualizations create qualitative mental models from which they can then generate quantitative information. In order to build their qualitative mental models, forecasters integrated information across multiple visualizations and extracted primarily qualitative information from visualizations in a goal-directed manner. We discuss both theoretical and practical implications of this study.

Inversion of feedforward neural networks: algorithms and applications

There are many methods for performing neural network inversion. Multi-element evolutionary inversion procedures are capable of finding numerous inversion points simultaneously. Constrained neural network inversion requires that the inversion solution belong to one or more specified constraint sets. In many cases, iterating between the neural network inversion solution and the constraint set can successfully solve constrained inversion problems. This paper surveys existing methodologies for neural network inversion, which is illustrated by its use as a tool in query-based learning, sonar performance analysis, power system security assessment, control, and generation of codebook vectors.

Inversion of neural network underwater acoustic model for estimation of bottom parameters using modified particle swarm optimizers

Given a complicated and computationally intensive underwater acoustic model in which some acoustic measurement is a function of sonar system and environmental parameters, it is computationally beneficial to train a neural network to emulate the properties of that model. Given this neural network model, we now have a convenient means of performing geoacoustic inversion without the computational intensity required when attempting to do so with the actual model. This paper proposes an efficient and reliable method of performing the inversion of a neural network underwater acoustic model to obtain parameters pertaining to the characteristics of the ocean floor, using two different modified version of particle swarm optimization (PSO): two-step (gradient approximation) PSO and hierarchical cluster-based PSO.

Stochastic resonance of a threshold detector: image visualization and explanation

Stochastic resonance is said to occur when just the right amount of noise enhances the performance of a process. For a simple threshold detector, the first moment of stochastic resonance is obtained by passing the signal through a transfer function equal to a transposed and shifted version of the underlying noises probability distribution function. The process is readily evident in images wherein noise corresponding to a linear transfer function produces a better visual representation than when other noise is used.

A fuzzy-logic autonomous agent applied as a supervisory controller in a simulated environment

An unsupervised learning system, implemented as an autonomous agent is presented. A simulation of a challenging path planning problem is used to illustrate the agent design and demonstrate its problem solving ability. The agent, dubbed the ORG, employs fuzzy logic and clustering techniques to efficiently represent and retrieve knowledge and uses innovative sensor modeling and attention focus to process a large number of stimuli. Simple initial fuzzy rules (instincts) are used to influence behavior and communicate intent to the agent. Self-reflection is utilized so the agent can learn from its environmental constraints and modify its own state. Speculation is utilized in the simulated environment, to produce new rules and fine-tune performance and internal parameters. The ORG is released in a simulated shallow water environment where its mission is to dynamically and continuously plan a path to effectively cover a specified region in minimal time while simultaneously learning from its environment. Several paths of the agent design are shown, and desirable emergent behavior properties of the agent design are discussed.

Neural network training for varying output node dimension

Considers the problem of neural network supervised learning when the number of output nodes can vary for differing training data. The paper proposes irregular weight updates and learning rate adjustment to compensate for this variation. In order to compensate for possible over training, an a posteriori probability that shows how often the weights associated with each output neuron are updated is obtained from the training data set and is used to evenly distribute the opportunity for weight update to each output neuron. The weight space becomes smoother and the generalization performance is significantly improved.

Biospar, A Buoy System For Acoustic Monitoring Of Biological Populations

BIOSPAR is a free floating spar buoy that will use acoustic backscattering in collecting long term data on plankton populations. The system, which is being jointly developed by the Woods Hole Oceanographic Institution and the Applied Physics Laboratory of the University of Washington, will make acoustic measurements at 120 kHz and 420 kHz. The backscatter data will be processed to produce estimates of the plankton density and acoustic size distribution as a function of depth. The processed acoustic data are stored aboard the buoy on an optical disk. Subsets of the data can be transmittcd to shore using a satellite telemetry link.

Team optimization of cooperating systems: application to maximal area coverage

The composite effort of the system team, rather, is significantly more important than a single players individual performance. We consider the case wherein each players performance is tuned to result in maximal team performance for the specific case of maximal area coverage (MAC). The approach is first illustrated through solution of MAC by a fixed number of deformable shapes. An application to sonar is then presented. Here, sonar control parameters determine a range-depth area of coverage. The coverage is also affected by known but uncontrollable environmental parameters. The problem is to determine K sets of sonar ping parameters that result in MAC. The forward problem of determining coverage given control and environmental parameters is computationally intensive. To facilitate real time cooperative optimization among a number of such systems, the sonar input-output is captured in a feedforward layered perceptron neural network.

Self-selective clustering of training data using the maximally-receptive classifier/regression bank

A common approach to pattern classification problems is to train a bank of layered perceptrons or other classifiers by clustering the input training data and training each classifier with just the data from a specific cluster. There is no provision in such an approach, however, to assure the component layered perceptron is well suited to learn the training data cluster it is assigned. An alternate method of training, herein proposed, lets a layered perceptron in a classifier bank choose the cluster of inputs it processes on the basis of the perceptrons ability to successfully classify those inputs. During training, data is therefore processed only by the classifier in the bank that best classifies the data or, equivalently, to which the data is most receptive. This allows each classifier to learn a localized subset of data dictated by the classifiers own classification ability. Once each classifier in the bank is trained, a separate independent cluster pointer is trained to recognize to which cluster an input test pattern belongs. The cluster pointer is used in the test mode to identify which classifier in the bank will best classify the problem. The approach, also applicable to regression type problems, is illustrated through application on a simulated Gaussian data set and an active sonar test estimation problem. In both cases, the maximally receptive classifer/regression bank significantly outperforms a single layered perceptron trained on the same data