Search

A computer program has been written which calculates DPM for single and dual labeled samples after correction for quenching. The program in ALGOL is executed from a remote terminal and uses punched paper tape input logged directly from the liquid scintillation counter in the proper format. Multiple quench indexes are used to calculate the best estimate of efficiencies for each isotope; channel ratios, AES ratios and AES counts are all used for single isotope samples and the two AES indexes only for dual labeled samples are used in the estimate computed in either case from a single sample counting. A separate counting produces data for the channels ratio computation for dual label samples using either the screening method or the general method where appropriate. Divergence of efficiencies computed from the different indexes indicates non-identity with the standards. The algorithm has been tested with serum, urine and aqueous biological samples.

Publisher Summary This chapter discusses a production system for counting, subitizing, and adding. It presents an explicit model of the process of quantification. The general model states that quantification of n items takes place via subitizing for n ≤ 4 and via subitizing and addition for n > 4. The latter process is what is conventionally called counting. The models form a collection of independent rules, called productions, that form a production system. A production system obeys simple operating rules: (1) the productions are considered in sequence, starting with the first; (2) each condition is compared with the current state of knowledge in the system, as represented by the symbols in STM, and if all of the elements in a condition can be matched with elements in STM, then the condition is satisfied; (3) If a condition is not satisfied, the next production rule in the ordered list of production rules is considered; and (4) if a condition is satisfied, the actions to the right of the arrow are taken, and then the production system is reentered from the first step; (5) when a condition is satisfied, all those STM elements that were matched are moved to the front of STM, providing a form of automatic rehearsal; (6) Actions can change the state of goals, replace elements, apply operators, or add elements to STM; and (7) the STM is a stack in which a new element appears at the top pushing all else in the stack down one position. Since STM is limited in size, elements may be lost.

Publisher Summary This chapter discusses the frequency analysis of stretches of time series resulting from runs made in accordance with an experimental design. The analyses proposed may be viewed as extensions of the familiar ANOVA procedures employed when each run of a design leads to the measurement of a real-valued variate. The chapter presents an approach that consists of the application of a complex version of the computations of ANOVA to the components of the discrete Fourier transforms of the observed stretches of series. It presents a case where the observed series form sections of stationary point processes. It also discusses two different procedures available for effectively increasing the number of observations; one can evaluate the finite Fourier transform at a number of frequencies in the neighborhood of λ, or one can evaluate the finite Fourier transform for a number of disjoint stretches of the series.

This paper is motivated by two papers Selin [1964,1965] on the problem of the sequential detection of signals in normal noise. Selin's problem was to construct a decision procedure with given error probabilities and of minimal expected time for the hypotheses signal present and signal absent. Here, the behaviour of the likelihood ratio test used by Selin is re-examined. It is shown that this test need not be optimal.

Publisher Summary A major part of most large-scale analyses of data on a computer is concerned with: (1) specifying the external arrangement of the data with their associated labels and headings, (2) specifying the internal core storage for the same data, and (3) describing how one arrangement is to be mapped into the other. By contrast, the internal manipulation of data is often much simpler. Some of the difficulties arise because the core store is essentially a one-dimensional string of locations, whereas data on paper are displayed in two dimensions and the effects of this persist even when the data are punched on paper tape, line by line. Other difficulties arise because efficient processing demands internal storage in vectors of homogeneous mode, for example, all of whose components are integer, or all real, whereas for display purposes, it is convenient to intersperse data values with labels, and sub-headings, so that consecutive items have different modes.

This paper discusses research into the further development of the Moire method of experimental stress analysis. Particular emphasis is placed upon achieving complete objectivity in reducing the strain information from the Moire pattern. Objectivity, as applied, implies a method free of measurements made by the operator's sensory perceptors. The objective method is applied to several tests, one of which is reported. The application is to the thermal stress problem of a ring under an axisymmetric temperature distribution. The maximum temperature reported is 1120°F. The results indicate that new levels of sensitivity and accuracy are achieved even though coarse gratings are used. The theory of the process is discussed, and a photo-reading device is described. A computer program is developed to reduce the Moire data. Techniques for using the Moire method in thermal stress problems at moderately high temperatures are devised. The conclusions drawn from this research are: 1. That the Moire method of experimental stress analysis can be made completely objective. 2. That reconsideration should be given the photogrid method applied in an objective fashion as a means of improving accuracy.

This chapter describes the way to draw figures at one stroke of the pencil. It explains that it is possible to trace at one stroke, that is, without lifting the pencil off the paper and without retracing any of the lines, a fairly complex figure, consisting of a number of nodes joined by paths. In tracing the figure in pencil, a distinction is made between the initial, final, and intermediate nodes. A (2m + l)-node can only be an initial one or a final one because by passing through it times, 2m paths are used and it is only possible to start or to complete the figure along the remaining path. It follows from this theorem that it is impossible to draw a figure, possessing more than two odd nodes, at one stroke. The Euler problem about the seven bridges of Konigsberg is discussed. Substituting points for the banks and the islands and lines for the bridges, Euler's problem is reduced to the question of drawing at one stroke.

In microwave transmission to distances well beyond the horizon, the received signal generally exhibits a characteristic fading. Measurements made on the received signal, in order to be considered representative, must be averaged over a time interval long compared with the period of the slowest fading component. Alternatively it is possible to perform certain kinds of measurements in a time interval short relative to the fastest fading components. Such a measurement yields an “instantaneous” picture of the effects produced by atmospheric conditions. A critical type of experiment involves the swinging of a narrow beam in order to resolve those signal components whose phase interrelationship determines the fading. If this rapid beam-swinging is performed by a phase shifting scheme in a broadside array, not only is the beam more readily controllable, but the interval between successive beam passages through the great-circle bearing can be made short. One then obtains a sequence, as in the frames of a slow-motion picture, in which the details of the statistical variations can be studied. This paper discusses the information that can be obtained by using phased-array techniques. It concludes with an example of an existing array at Stanford University being used for propagation studies at 3.12 kMc over a 101-mile path. The research reported in this document was made possible through the support of the U.S. Army Signal Corps, under Contract DA 36(039)-sc-73151.

The paper describes a method for the synthesis of ternary functions using diagonal negation and two threshold operations. Main advantages over previous methods are a reduction in the number of required inputs and a distribution of input functions over input nodes that automatically takes care of fan-in limitations. It is also shown that in the case of self-dual functions a more economical synthesis is possible by a change in the representation basis and finally the electronic circuits implementing the method are described.

SUMMARY In this paper an attempt is made to show that for an electronic measurement of displacements the group consisting of measuring element + reading head + interpolator can quite appropriately be referred to as measuring system. Thus it is possible to draw a distinctive line between the actual electronic measuring and the electronic processing of measured values. Considering the definition of the term “measuring”, all measuring elements serving for the measurement of lengths are linear measuring elements whose divisions—the primary divisions—must usually, depending on the accuracy requirements with respect to the measuring systems, be interpol ated to obtain the necessary digital steps. If one considers the three interpolation methods—mechanical, optical, and electronic—it becomes evident that for the development of incremental singlestep measuring systems for instruments of increased and highest accuracy the electronic interpolation of primary intervals is gaining in importance to an ever increasing extent. Besides the well-known electronic technique of the production of four counts per cycle a circuit with eight counts per cycle can be realized by comparatively simple means by an electronic bisection of the primary intervals, followed by a division into four counts per cycle. The Cathode-Ray Interpolator moreover offers the possibility of a further division of the primary intervals into 40 or 100 steps, a special cathode-ray tube acting as interpolator.

In this paper we are concerned with learning algorithms which are, in a sense, between series and parallel processing. More distinctly, such iterative learning procedures are considered which may also include a fairly broad class of complex (and even partly heuristically defined) processing steps. Constraints are set up under which such procedures do precisely learn discriminant functions. It is shown, how may one, while retaining convergence, confine the iteration to some fixed search domain, and also keep the required storage under control. These topics are of interest e.g., if (because of complexity) neither completely simultaneous nor subsequent, one by one, handling of the observations appears to be feasible. Techniques of this sort offer in compound machine learning further possibilities to combine, what may be called, mid- and end- procedures.

This paper describes the use of various curve-fitting techniques for development of systems performance algorithms. Nonlinear simple and multiple linear and nonlinear regression methods are described. For the nonlinear two-variable case, where the resulting normal equations are also nonlinear, a novel search and modified method of least squares approach is illustrated. These techniques are examined and compared with other existing techniques from the standpoint of CPU time (and hence cost) requirements for algorithmic development. Several algorithms (or models) generated from real systems data will be described and emphasis will be placed on development, practicality of use and criteria for goodness of fit. In addition, the use of terminal-accessible databank programs which greatly minimize the labor of data handling and curve fitting in problems of this type is described.

SUMMARY Part I of this paper deals with the amount of degeneracy in the genetic code or, looked at in reverse, the amount of nonsense. Section A considers the experimental evidence and suggests (a) that technical difficulties have prevented assigning meaning to the still unassigned triplets and (b) that the validity of the earlier evidences for nonsense triplets may be questioned in the light of recent discoveries. Complete degeneracy of the code and total absence of nonsense have not yet been excluded. Section B comes to the conclusion, on the basis of general evolutionary considerations, that natural selection would be expected to establish and preserve a completely degenerate code. Section C points out that different nondegenerate codes differ greatly in the builtin frequency of nonsense mutations by single base substitutions. Part II deals with the nature of degeneracy in the existing code. The high frequency of shared doublets between “synonyms” (i.e., codons for the same amino acid) suggests that they are “connected,” that is, interconvertible by altering one base only. Among the sets of synonyms thus far reported, some must be completely connected, and, in the absence of knowledge of the order of the bases, the synonyms of each set could be ordered so as to be completely connected. Current knowledge leads to the expectation of either complete connectedness or a high degree of connectedness among synonyms. Part III explores the consequences of complete connectedness between synonyms; the consequences would be modified only quantitatively, not qualitatively, if connectedness is less than complete. Section A discusses “silent” mutations, i.e., single-base changes that yield a synonym. These must occur in the code as now known. Their frequency (if equal probability of all base substitutions is assumed) depends on the number of synonyms in a set, on the way in which the synonyms are connected, and on whether the various synonyms are used equally or unequally. By selecting one synonym of a set for almost exclusive usage, the minimal percentage of silent mutations stays constant at 11.1% in sets of two or more synonyms, but the maximum increases steadily with set size up to 100% in sets of ten or more. If the synonyms of a set are equally used, the pattern of connection that yields the highest frequency of silent mutations gives peaks for sets of four, seven, and ten synonyms. It would not be surprising if 20% or more of all single-base mutations were silent unless the frequency of trans versions is greatly restricted. Comparable analyses can be made of a quadruplet code. The frequency of mutations to synonyms is not enough to account fully for modulatory mutations, but it is enough to constitute a potential source of error in certain amino acid replacement studies. Section B points out that the degeneracy of the existing code implies as an evolutionary consequence a molecular orthogenesis, that is, marked built-in differences among the synonyms of a set in their possible amino acid replacements by single-base mutations. Section C shows the surprising fact that crossing over between synonyms must in certain cases yield recombinant codons for a different amino acid. Although likely to be scored as mutations, these recombinations would be much more frequent than true mutations. The advantage of connected synonymy is opposed by the disadvantage of these recombinational pseudomutations. Possible ways out of this paradox are discussed.

This chapter discusses the formation of statistical material. Statistics present phenomenon in the form of tables that show how many times or in what numbers certain phenomena appear in a given population. To design a statistical table, it is necessary to know the particular units of which the population is composed and to discover for each of them separately those characteristics which are to appear in the table. For instance, to design a statistical table in which the leaves of a birch tree are classified according to length and width, the length and width of every leaf of the birch is to be known. This individual information is called as statistical material. It is formed by the gathering of individual data on each unit separately. In the case of the leaves of a birch, the task of those collecting the material consists in measuring each leaf and writing down the results. The difficulties that may arise result from the requirement that the measurements be taken correctly. If this work is done by a single investigator, the number of leaves measured must be very limited. The same is true in other fields when the whole material is collected by one or several persons. If it is possible to determine exactly the characteristics for each unit, then the difficulties may result only from the great amount of work required to do the job, there being no difficulties with regard to technique or organization. The problem consists in deciding whether it is better to write down all the information concerning each particular unit on a separate sheet of paper or to make out a form in which to note, one by one, the data on each unit.

In this paper we consider the problem of locating a faulty module in a one-dimensional array. Kautz has previously considered this problem and derived a set of sufficient conditions for fault location which were also claimed to be necessary. We show that these conditions are not necessary for fault location and that in some cases a faulty module can be located even if it contains no external output. Specifically, we define the concept of masking inputs of degree k, which is useful for locating the faulty module within k units. This concept is then generalized to obtain necessary and sufficient conditions to locate the faulty module within k units. Making use of these concepts, we introduce a simple design procedure whereby the addition of one external binary input to each module enables fault location to within two modules for an arbitrary array. These techniques may be applicable to some systems of interconnected modules other than one dimensional arrays.

Publisher Summary This chapter presents several applications of the computer as a tool that are significant in the analysis of the very complex data that come from brain systems. The chapter discusses a model of brain tissue, which has three compartments. The neural compartment is composed of nerve cells. The second component of the tricompartmental arrangement is a neuroglial compartment. Between the neuroglia and the neuron is a substance, which is a hydrated net of large molecules, mucopolysaccharides, and mucoproteins, specially developed under synapses, and which plays a role in the excitability of the cell. The chapter discusses the types of data that are collected for computer analysis. There are two main varieties of codes. One is widely used in the laboratory and can be generated as a 3-digit 4-bit member in binary coded decimal (BCD). The second group of codes includes so-called IRIG A and B methods that may give a faster coding cycle, but are not usually compatible in their speed with a typical EEG paper record because they are written very fast, and thus cannot be interpreted. The first step in preparation of data for handling by the computer involves analog-to-digital conversion. Secondly, the data sampling rate must bear a relationship with the bandwidth of the data. A typical modern analog-to-digital converter for electrophysiological studies incorporates a multiplexing input to handle 10 to 30 channels of data. Some computational techniques can be used on-line; how much can be done depends upon the size of the central processor and the mode of access to it. One may choose to start simply by averaging, or by correlation analyses.