Pattern matching in a model of dendritic spines.

Pattern matching, the ability to recognize and maximally respond to an input pattern that is similar to a previously learned pattern, is an essential step in any learning process. To investigate the properties of pattern matching in biological neurons, and in particular the role of a calcium-dependent potassium conductance, a circuit model of a small area of dendritic membrane with a number of dendritic spines is developed. Circuit model simulations show that dendritic membrane depolarization is greater in response to a previously learned pattern of synaptic inputs than in response to a novel pattern of synaptic inputs. These simulations, in combination with an analysis of the circuit model equations, reveal that when a synaptic input pattern is similar to the learned pattern of synaptic inputs, the total dendritic depolarization is a linear combination of dendritic depolarization contributed by individual spines. When at least one synaptic input differs markedly from the learned value, dendritic depolarization is a nonlinear combination of individual spine depolarizations. These principles of spine interactions are captured in a computationally simple set of 'similarity measure' equations which are shown to reproduce the response surface of the circuit model output. Thus, these similarity measure equations not only describe a biologically plausible model of pattern matching, they also satisfy computational requirements for use in artificial neural networks.

Orange juice classification with a biologically based neural network.

Dystal, an artificial neural network, was used to classify orange juice products. Nine varieties of oranges collected from six geographical regions were processed into single-strength, reconstituted or frozen concentrated orange juice. The data set represented 240 authentic and 173 adulterated samples of juices; 16 variables [8 flavone and flavanone glycoside concentrations measured by high-performance liquid chromatography (HPLC) and 8 trace element concentrations measured by inductively coupled plasma spectroscopy] were selected to characterize each juice and were used as input to Dystal. Dystal correctly classified 89.8% of the juices as authentic or adulterated. Classification performance increased monotonically as the percentage of pulpwash in the sample increased. Dystal correctly identified 92.5% of the juices by variety (Valencia vs non-Valencia).

Associative learning in a network model of Hermissenda crassicornis. II. Experiments.

A companion paper in a previous issue of this journal presented a resistance-capacitance circuit computer model of the four-neuron visual-vestibular network of the invertebrate marine mollusk Hermissenda crassicornis. In the present paper, we demonstrate that changes in the model's output in response to simulated associative training is quantitatively similar to behavioral and electrophysiological changes in response to associative training of Hermissenda crassicornis. Specifically, the model demonstrates many characteristics of conditioning: sensitivity to stimulus contingency, stimulus specificity, extinction, and savings. The model's learning features also are shown to be devoid of non-associative components. Thus, this computational model is an excellent tool for examining the information flow and dynamics of biological associative learning and for uncovering insights concerning associative learning, memory, and recall that can be applied to the development of artificial neural networks.

Associative learning in a network model of Hermissenda crassicornis. I. Theory.

A time-varying Resistance-Capacitance (RC) circuit computer model was constructed based on known membrane and synaptic properties of the visual-vestibular network of the marine snail Hermissenda crassicornis. Specific biophysical properties and synaptic connections of identified neurons are represented as lumped parameters (circuit elements) in the model; in the computer simulation, differential equations are approximated by difference equations. The model's output, membrane potential, an indirect measure of firing frequency, closely parallels the behavioral and electrophysiologic outputs of Hermissenda in response to the same input stimuli presented during and after associative learning. The parallelism of the computer modeled and the biologic outputs suggests that the model captures the features necessary and sufficient for associative learning.

Performance comparison of a neural network with human observers on a visual target detection task.

An experiment is described which compares the performance of a neural network to human performance on a visual task which consists of detecting a target in a background image of correlated noise. A three-layer, feed-forward, multi-layer perceptron is trained to indicate the presence or absence of a target in images also presented to human observers. The basis for the comparison between the network and the human observers is the receiver operating characteristic (ROC) curve. Network performance is comparable to human performance for this particular task.

Pattern-recognition by an artificial network derived from biologic neuronal systems.

A novel artificial neural network, derived from neurobiological observations, is described and examples of its performance are presented. This DYnamically STable Associative Learning (DYSTAL) network associatively learns both correlations and anticorrelations, and can be configured to classify or restore patterns with only a change in the number of output units. DYSTAL exhibits some particularly desirable properties: computational effort scales linearly with the number of connections, i.e., it is O(N) in complexity; performance of the network is stable with respect to network parameters over wide ranges of their values and over the size of the input field; storage of a very large number of patterns is possible; patterns need not be orthogonal; network connections are not restricted to multi-layer feed-forward or any other specific structure; and, for a known set of deterministic input patterns, the network weights can be computed, a priori, in closed form. The network has been associatively trained to perform the XOR function as well as other classification tasks. The network has also been trained to restore patterns obscured by binary or analog noise. Neither global nor local feedback connections are required during learning; hence the network is particularly suitable for hardware (VLSI) implementation.

Bilirubin dynamics in the Gunn rat. Dose response of intermittent and continuous phototherapy.

Jaundiced Gunn rats underwent continuous and intermittent phototherapy with white and blue lights. Irradiances ranged from 0.088 to 1.4 mW/cm2 in the blue (415-465 nm) wavelengths. Changes in serum bilirubin concentration were analyzed by a single exponential model. Dose-response curves for 10 days of phototherapy indicated similar logarithmic relationships between light dose and serum bilirubin decline for intermittent and continuous blue light and continuous white light phototherapy. The dose-response curve for intermittent white light showed a low slope and correlation coefficient.

Intermediate phototherapy in the treatment of jaundice in the premature infant.

A controlled trial of the use of intermittent phototherapy for the treatment of hyperbilirubinemia in newborn infants is reported. Periods of illumination of (1) 15 minutes light on, 15 minutes light off, (2) 15 minutes on, 30 minutes off, and (3) 15 minutes on, 60 minutes off are as effective as is continuous illumunation. A comparison with previous trials of intermittent phototherapy is made and differences in results are explained using as a model the action of light on bilirubin.

Manual micromethods for bilirubin determination in sera of adults and children and investigation of reasons for observed differences.

This study compared three micromanual methods for determining bilirubin concentration. The two microchemical methods for total bilirubin, a Jendrassik-Grof procedure and a Unopette procedure, using dimethyl sulfoxide as an accelerator and protein solubilizer, gave comparable results in sera of adults and children. A microspectrophotometric method and the microchemical methods for total bilirubin gave similar results in plasmas of newborns with physiologic hyperbilirubinemia and in sera of older children with no hepatic abnormality. However, the microspectrophotometric method gave higher values in normal and hyperbilirubinemic adult sera. The results obtained with the Jendrassik-Grof and Unopette microchemical methods for direct bilirubin in sera of adults and children showed the values determined by the Unopette to be higher. Using the presently accepted normal range, this difference is significant enough to preclude recommendation of the use of the Unopette method for distinguishing normal from elevated levels of direct bilirubin. Direct bilirubin in newborn serum measured by the Unopette method is considerably higher than that measured by the Jendrassik-Grof method. An investigation to determine the reason for the difference in the direct bilirubin results indicated that the Unopette direct method measures diconjugated bilirubin in amounts similar to those measured by the Jendrassik-Grof method but significantly more monoconjugated bilirubin than the Jendrassik-Grof method.

Effect of intermittent phototherapy on bilirubin dynamics in Gunn rats.

To determine whether continuous phototherapy is necessary to control neonatal jaundice, groups of jaundiced (Gunn) rats were exposed to four blue light regimens: continuous light, 30 min light/30 min dark, 6 min light/6 min dark, and 6 min light/18 min dark. An exponential decrease in serum bilirubin concentration (SBC) was found with all regimens. A logarithmic dose response curve was obtained with a 50% light dose exhibiting 75% efficacy, and a 25% light dose exhibiting 59% efficacy compared to continuous phototherapy. The time constants of the decrease in serum bilirubin concentration also varied logarithmically with doses with continuous therapy having a time constant of 1.4 days, and 6 min light/18 min dark therapy having a time constant of 3 days.

Spline functions: an alternative representation of aspheric surfaces.

This paper offers an alternative to current practice in the specification of aspheric optical surfaces. Spline functions, a numerical analog of a flexible draftsman's curve, are used to represent a rotationally symmetric aspheric surface. We develop the interpolation formulas pertinent to ray tracing and apply an optimization procedure to design a simple Schmidt system as an illustration. We conclude by showing that the new formulation seems to be more compatible with the optimization process than the usual polynomial aspheric.

Asymmetric Lens Design Using Bicubic Splines: Application to the Color TV Lighthouse.

Any two-dimensional interpolation scheme which has continuous derivatives may be used to represent an optical surface for ray-tracing purposes. We present bicubic splines in their application to the design of asymmetric surfaces. An as example of a problem requiring an asymmetric system, we analyze the design problems of a color TV lighthouse lens.

Semiautomatic generation of optical prototypes.

This paper presents the theory and application of a computer approach to the problem of finding lens prototypes from given sets of design goals. The procedure is based on the yy method of Delano and has two independent segments: a strictly first-order phase and a combined first- and third-order phase. The former optimizes the first-order layout with respect to geometrical and optical constraints. In this phase the independent variables are the paraxial marginal and principal ray heights. The starting point is a set of yy coordinates which may be either generated by the computer program or supplied by the designer. The second phase combines third-order information with the constraints of the first phase. Here it is more convenient to use ray angles as independent variables. The optimization scheme for both phases is an orthogonal method which has been used successfully on a variety of optical problems.