Set Of Prototypes Research Articles

Comparison of genetic algorithm based prototype selection schemes

1. IntroductionPrototype selection is the process of nding represen-tative patterns from the data. Representative patternshelp in reducing the data on which further operationssuch as data mining can be carried out. The currentwork discusses computation of prototypes usingmedoids [1], leaders [2] and distance based thres-holds. After nding the initial set of prototypes, theoptimal set is found by means of genetic algorithms(GAs). A comparison of stochastic search algorithms iscarriedout by SusheelaDeviand NarasimhaMurty [3].They conclude that performance of genetic algorithmsis the best among the search algorithms. Chang andLipmann [4] suggest the use of genetic algorithms forpattern classication.In the following sections, we discuss and comparevarious prototype selection methods under considera-tion. Comparison of results are based on nearest neigh-bor classier (NNC). Subsequently, considering thoseprototype sets which provided good classication accu-racy, GAs are used for optimal prototype selection.Based on the natureof the data characteristicsa numberofexperimentsbasedonGAsarecarriedout.Asummaryof results is presented.2. Description of dataHandwritten digit data [5] is used for the comparisonexercises. The training data consists of 667 patterns foreach class of digits 0}9, totalling to 6670 patterns. Thetest data consists of 3333 patterns. While carrying out

Learning with Nearest Neighbour Classifiers

This paper introduces a learning strategy for designing a set of prototypes for a 1-nearest-neighbour (NN) classifier. In learning phase, we transform the 1-NN classifier into a maximum classifier whose discriminant functions use the nearest models of a mixture. Then the computation of the set of prototypes is viewed as a problem of estimating the centres of a mixture model. However, instead of computing these centres using standard procedures like the EM algorithm, we derive to compute a learning algorithm based on minimising the misclassification accuracy of the 1-NN classifier on the training set. One possible implementation of the learning algorithm is presented. It is based on the online gradient descent method and the use of radial gaussian kernels for the models of the mixture. Experimental results using hand-written NIST databases show the superiority of the proposed method over Kohonen's LVQ algorithms.

VideoCafé - exploring mediaspaces in public places within organizations

This paper describes our studies of mediaspaces which are embedded within public places in organizations so as to examine the hypothesis that individuals might benefiteven when working apart- from opportunities for light informal interaction. A set of full-scale prototypes were used and assessed over extensive periods of time. The informal observations and reflection in design of these places have been supplemented by formal studies. We found that great care needs to be taken when designing these places from an architectural point of view. For some of the places, we would like to suggest using architectural features when altering the room rather than technology. In other settings, the artful deployment of communication media might be more effective.

A three-dimensional human agent metaphor for modeling and simulation

The use of metaphor can be a potential aid to the novice modeler in several ways. Metaphor can imbue abstract ideas with concrete properties, thereby making the abstract ideas more accessible. The analogies suggested by metaphor might also aid reasoning about modeling and implementation problems. Another potential benefit of metaphor in modeling is the improvement of mental retention of model architecture and functionality. Traditionally, models and programs have been produced in a two dimensional (2D) or textual medium. However these media may be inferior to a three-dimensional (3D) medium in the development and use of metaphor, as the concrete properties that metaphors often provide are real-world phenomena, which are naturally 3D. We developed an example of the use of metaphors in modeling and 3D simulation. The example consists of a simplified operating system task scheduler along with associated hardware devices, developed in a VRML environment using VRML PROTO nodes. These nodes are designed as modular objects based on real-world metaphors. We were able to construct a set of metaphors and prototypes that may, if extended ease the modeling and design of agent oriented systems for novices. A proposed extension of one metaphor presented in the research is the synthetic human agent.

Capturing Design Space From a User Perspective: The Repertory Grid Technique Revisited

The design of an artifact (e.g., software system, household appliance) requires a multitude of decisions. In the course of narrowing down the design process, "good ideas" have to be divided from "bad ideas." To accomplish this, user perceptions and evaluations are of great value. The individual way people perceive and evaluate a set of prototypes designed in parallel may shed light on their general needs and concerns. The Repertory Grid Technique (RGT) is a method of elucidating the so-called personal constructs (e.g., friendly-hostile, bad-good, playful-expert-like) people employ when confronted with other individuals, events, or artifacts. We assume that the personal constructs (and the underlying topics) generated as a reaction to a set of artifacts mark the artifacts' design space from a user's perspective and that this information may be helpful in separating valuable ideas from the not so valuable. This article explores the practical value of the RGT in gathering design-relevant information about the design space of early artifact prototypes designed in parallel. Ways of treating the information gathered, its quality and general advantages, and limitations of the RGT are presented and discussed. In general, the RGT proved to be a valuable tool in exploring a set of artifact's design space from a user's perspective.

A Batch Learning Vector Quantization Algorithm for Nearest Neighbour Classification

We introduce a batch learning algorithm to design the set of prototypes of 1 nearest-neighbour classifiers. Like Kohonen's LVQ algorithms, this procedure tends to perform vector quantization over a probability density function that has zero points at Bayes borders. Although it differs significantly from their online counterparts since: (1) its statistical goal is clearer and better defined; and (2) it converges superlinearly due to its use of the very fast Newton's optimization method. Experiments results using artificial data confirm faster training time and better classification performance than Kohonen's LVQ algorithms.

10.1162/jmlr.2003.3.4-5.951

In this paper we study a paradigm to generalize online classification algorithms for binary classification problems to multiclass problems. The particular hypotheses we investigate maintain one prototype vector per class. Given an input instance, a multiclass hypothesis computes a similarity-score between each prototype and the input instance and sets the predicted label to be the index of the prototype achieving the highest similarity. To design and analyze the learning algorithms in this paper we introduce the notion of ultraconservativeness. Ultraconservative algorithms are algorithms that update only the prototypes attaining similarity-scores which are higher than the score of the correct label's prototype. We start by describing a family of additive ultraconservative algorithms where each algorithm in the family updates its prototypes by finding a feasible solution for a set of linear constraints that depend on the instantaneous similarity-scores. We then discuss a specific online algorithm that seeks a set of prototypes which have a small norm. The resulting algorithm, which we term MIRA (for Margin Infused Relaxed Algorithm) is ultraconservative as well. We derive mistake bounds for all the algorithms and provide further analysis of MIRA using a generalized notion of the margin for multiclass problems. We discuss the form the algorithms take in the binary case and show that all the algorithms from the first family reduce to the Perceptron algorithm while MIRA provides a new Perceptron-like algorithm with a margin-dependent learning rate. We then return to multiclass problems and describe an analogous multiplicative family of algorithms with corresponding mistake bounds. We end the formal part by deriving and analyzing a multiclass version of Li and Long's ROMMA algorithm. We conclude with a discussion of experimental results that demonstrate the merits of our algorithms.

A mass assignment method for prototype induction

A form of prototypes defined as tuples of marginal probability distributions is introduced. Addition and subtraction operations on such prototypes are then described. A brief introduction to the mass assignment theory of the probability of fuzzy events is given, and it is shown how fuzzy sets can serve as conceptual descriptions of probability distributions. Hence fuzzy descriptions of prototypes can be derived and these can be used for inference as well as enabling rule based representations of a set of prototypes to be formed. A prototype induction algorithm, based on these ideas together with the addition and subtraction operations, is described. The potential of this approach is then illustrated by its application to a number of model and real world machine learning problems. ©1999 John Wiley & Sons, Inc.

An “Openable,” High-Strength Gradient Set for Orthopedic MRI

A novel three-axis gradient set and RF resonator for orthopedic MRI has been designed and constructed. The set is openable and may be wrapped around injured joints. The design methodology used was the minimization of magnetic field spherical harmonics by simulated annealing. Splitting of the longitudinal coil presents the major design challenge to a fully openable gradient set and in order to efficiently design such coils, we have developed a new fast algorithm for determining the magnetic field spherical harmonics generated by an arc of multiturn wire. The algorithm allows a realistic impression of the effect of split longitudinal designs. A prototype set was constructed based on the new designs and tested in a 2-T clinical research system. The set generated 12 mT/m/A with a linear region of 12 cm and a switching time of 100 μs, conforming closely with theoretical predictions. Preliminary images from the set are presented.

Data compression and local metrics for nearest neighbor classification

A local distance measure for the nearest neighbor classification rule is shown to achieve high compression rates and high accuracy on real data sets. In the approach proposed here, first, a set of prototypes is extracted during training and, then, a feedback learning algorithm is used to optimize the metric. Even if the prototypes are randomly selected, the proposed metric outperforms, both in compression rate and accuracy, common editing procedures like ICA, RNN, and PNN. Finally, when accuracy is the major concern, we show how compression can be traded for accuracy by exploiting voting techniques. That indicates how voting can be successfully integrated with instance-based approaches, overcoming previous negative results.

BELT-TYPE SEEDER FOR SOYBEANS

Some soybean growers have expressed a need for a soybean seeder that would combine the narrow rowcapability and relatively low cost of a grain drill with the precision seed spacing of a row-crop seeder. A prototype belt-meteringseeder for soybeans was designed, constructed, and tested as a possible answer to this need. The prototype wascompared with commercially available seeders equipped with fluted wheel, brush, or finger metering systems. At groundspeeds of 3.2 to 9.6 km/h (2-6 mph), the prototype set for a nominal spacing of 24 mm (0.94 in.) was able to metersoybeans with a mean spacing of 23 to 30 mm (0.92-1.17 in.) and a quality of feed index of 38 to 46%. Seeding uniformityof the prototype was not better than the seeding uniformity of the three commercial seeders.

An Integrated Framework for Generalized Nearest Prototype Classifier Design

We propose a Generalized Nearest Prototype Classifier (GNPC) as a common framework for a number of classification techniques. Specifically we consider clustering-and-relabeling; Parzen's classifier; radial basis functions (RBF) networks; learning vector quantization (LVQ) type classifiers; and nearest neighbor rules. To classify an unlabeled point x the GNPC combines the degrees of similarity of x to a set of prototypes. Five questions are addressed for these GNPC families: (1) How many prototypes do we need? (2) How are the prototypes found? (3) How are their class labels obtained? (4) How are the similarities defined? (5) How are the similarities and label information combined? The classification performance of a set of GNPCs is illustrated on two benchmark data sets: IRIS and the 2-spirals data. We study the resubstitution error of the GNPC as a function of the number of prototypes. Our conclusions are that: (a) unsupervised selection (or extraction) of prototypes followed by relabeling is inferior to the techniques that use labels to guide them towards prototypes; (b) the edited nearest neighbor rule is a viable option for GNPC design which has not received the attention it deserves.

The development of PTFE (polytetrafluoroethylene)-faced hydrodynamic thrust bearings for hydrogenerator application in the United Kingdom

Polytetrafluoroethylene (PTFE)-faced hydrodynamic thrust bearings have been used for many years in the former Soviet Union and in the People's Republic of China for large hydrogenerator applications. Although there has been some recent interest in other countries, few installations of this sort have previously been reported in Europe or the USA. This paper describes the state of the art and the development and laboratory testing of a prototype set of PTFE pads. Following these trials, a second larger set of thrust pads was designed for use in the pumped-storage hydroelectric power station at Ffestiniog in North Wales. The pads were installed in September 1996 and inspected in May 1997. After nearly 3000 h of operation including 900 separate occasions when the system was started up under load, the pads were found to be in excellent condition. Further hydrogenerator installations of PTFE-faced pads are planned as a result of this successful development.

A new neural network for cluster-detection-and-labeling

We propose in this paper a new unsupervised neural network which is capable of clustering a set of experimental data according to a given generic interpoint similarity measure, and then assign to each new input its appropriate cluster label. The network is able to do this for clusters of any shape, and without knowing in advance the number of clusters to be created. We call this new two-layer network a cluster-detection-and-labeling (CDL) network. In the CDL network the concept of similarity and closeness with regard to distance are combined. Specifically, clusters are represented by a set of prototypes, and the similarities between an input vector and these prototypes are calculated as the inner products of these vectors compared to some thresholds. These thresholds, which depend on the distance between the input vector and the prototype, are calculated in a separate threshold calculating unit. During clustering, the data are cycled through the network several times. At the end of each cycle the clusters are evaluated, and only those with more than a specified number of samples are retained. The others are fed back to be reclustered by an updated network. This process terminates according to a suitable criterion, such as when a prespecified portion of the data are classified. The performance of the CDL network has been compared with that of the winner-take-all (WTA) network for several different cluster structures, since the latter is widely used in cluster analysis applications. These studies demonstrate that the new network performs well for all the tested cluster shapes, also for those cases where the WTA network completely fails.

Two efficient connectionist schemes for structure preserving dimensionality reduction

We propose two neural-net-based methods for structure preserving dimensionality reduction. Method 1 selects a small representative sample and applies Sammon's method to project it. This projected data set is then used to train an MLP. Method 2 uses Kohonen's self-organizing feature map (SOFM) to generate a small set of prototypes which is then projected by Sammon's method. This projected data set is then used to train an MLP. Both schemes are quite effective in terms of computation time and quality of output, and both outperform methods of Jain and Mao on the data sets tried.

Ab initio determination of quasi-diabatic states for multiple reaction pathways

Abstract We have extended our method for the determination of quasi-diabatic states, in order to treat systems with several open reaction channels. The method is based on the unitary transformation of the adiabatic states, obtained with any standard ab initio method, into the quasi-diabatic ones. The transformation is such that the quasi-diabatic wavefunctions exhibit a maximum overlap with a set of diabatic prototypes. In the present work we show how the prototypes can be smoothly adapted to represent diabatic states for different reaction pathways. The new method is applied to a preliminary study of the photochemistry of azomethane, where several reaction channels can be envisaged: dissociation of either one or the other CN bond, isomerization by rotation, and isomerization by inversion.

An integrated approach to fuzzy learning vector quantization and fuzzy c-means clustering

Derives an interpretation for a family of competitive learning algorithms and investigates their relationship to fuzzy c-means and fuzzy learning vector quantization. These algorithms map a set of feature vectors into a set of prototypes associated with a competitive network that performs unsupervised learning. Derivation of the new algorithms is accomplished by minimizing an average generalized distance between the feature vectors and prototypes using gradient descent. A close relationship between the resulting algorithms and fuzzy c-means is revealed by investigating the functionals involved. It is also shown that the fuzzy c-means and fuzzy learning vector quantization algorithms are related to the proposed algorithms if the learning rate at each iteration is selected to satisfy a certain condition.

Task priorities and design for an arm orthosis

For people with muscular dystrophy or spinal muscular atrophy - conditions characterized by degenerating muscle strength - the option of using a mechanical aid exists. This would allow free arm movement. This paper presents a consumer-based design approach to construction of such an assistive manipulation device and describes the consumer-driven stages in the orthosis design project within the Applied Science and Engineering Laboratories (ASEL). Results are presented from two consumer meetings held during the initial stages of the design. A design principle has been establish.:d and an initial set of prototypes evaluated. These will be used to modify subsequent design iterations.

Double-sided “radiation resistant” microstrip detectors: Technology and results

The paper describes the technology used for the fabrication of a set of prototypes of microstrip silicon detectors matching the requirements of the CMS inner tracking system. The detectors are single and double-sided devices featuring integrated AC coupling on both sides. Polysilicon resistors are used as bias elements and individual p-stop patterns provide isolation in between the strips on the ohmic side. The design is such that the performance of these detectors will be marginally affected by the heavy irradiation correlated to several years of running at LHC. Different dielectric layers have been used to study rigidity and yield in the integrated capacitors. Different polysilicon implant doses have been used to optimise the uniformity and to increase the reproducibility of the poly resistors. The paper describes the design choices and the results of laboratory tests on a series of devices.

Learning with Preknowledge: Clustering with Point and Graph Matching Distance Measures

Prior knowledge constraints are imposed upon a learning problem in the form of distance measures. Prototypical 2D point sets and graphs are learned by clustering with point-matching and graph-matching distance measures. The point-matching distance measure is approximately invariant under affine transformations—translation, rotation, scale, and shear—and permutations. It operates between noisy images with missing and spurious points. The graph-matching distance measure operates on weighted graphs and is invariant under permutations. Learning is formulated as an optimization problem. Large objectives so formulated (∼ million variables) are efficiently minimized using a combination of optimization techniques—softassign, algebraic transformations, clocked objectives, and deterministic annealing.