Neural Net Model Research Articles

An architectural model of conscious and unconscious brain functions: Global Workspace Theory and IDA

While neural net models have been developed to a high degree of sophistication, they have some drawbacks at a more integrative, "architectural" level of analysis. We describe a "hybrid" cognitive architecture that is implementable in neuronal nets, and which has uniform brainlike features, including activation-passing and highly distributed "codelets," implementable as small-scale neural nets. Empirically, this cognitive architecture accounts qualitatively for the data described by Baars' Global Workspace Theory (GWT), and Franklin's LIDA architecture, including state-of-the-art models of conscious contents in action-planning, Baddeley-style Working Memory, and working models of episodic and semantic longterm memory. These terms are defined both conceptually and empirically for the current theoretical domain. The resulting architecture meets four desirable goals for a unified theory of cognition: practical workability, autonomous agency, a plausible role for conscious cognition, and translatability into plausible neural terms. It also generates testable predictions, both empirical and computational.

철강 생산 공정에서 Soft Computing 기술을 이용한 온도하락 예측 모형의 비교 연구

This paper is to establish an appropriate model for predicting the temperature decreases in the batch transferred from the refining process to the caster in steel-making companies. Mathematical modeling of the temperature decreases between the processes is difficult, since the reaction mechanism by which the temperature changes in a molten steel batch is dynamic, uncertain and complex. Three soft computing techniques are examined using the same data, namely the multiple regression, fuzzy regression, and neural net (NN) models. To compare the accuracy of these three models, a limited number of input variables are selected from those variables significantly affecting the temperature decrease. The results show that the difference in accuracy between the three models is not statistically significant. Nonetheless, the NN model is recommended because of its adaptive ability and robustness. The method presented in this paper allows the temperature decrease to be predicted without requiring any precise metallurgical knowledge.

Neural net versus classical models for the detection and localization of leaks in pipelines

Four models of a pipeline are compared in the paper: a nonlinear distributed-parameter model, a linear distributed-parameter model, a simplified lumped-parameter model and an extended neural-net-based model. The transcendental transfer function of the linearized model is obtained by a Laplace transformation and corresponding initial and boundary conditions. The lumped-parameter model is obtained by a Taylor series extension of the transencdental transfer function. Based on the experience of linear models the structure of the neural net model, as an addendum to the nonlinear distributed-parameter model, is obtained. All four models are tested on a real pipeline data with an artificially generated leak.

Towards a unifying model of neural net activity in the visual cortex

A neural net model describing the non-linear interactions between axonal spikes is presented. It reconciles aspects of pattern recognition (as action of an associative memory) with those of spike synchronization and phase locking. The stability of the synchronized state is studied in detail.

Undecidability over Continuous Time

Since 1996, some models of recursive functions over the real numbers have been analyzed by several researchers. It could be expected that they exhibit computational power much greater than that of Turing machines (as other well known models of computation over the real numbers already considered in the past fifteen years, like neural net models with real weights). The fact is that they do not have such power. Although they decide the classical halting problem of Turing machines, they otherwise have almost the same limitations of Turing machines.

Perceived Body Size Affected by Garment and Body Mass Index

This study investigated the effect of garment size on perceived body size. The perceived body sizes of nine Chinese men, with Body Mass Index between 17.0 and 37.1 kg/m(2), wearing five sizes of white T-shirts were assessed using Thompson and Gray's Nine-figural Scale. Garment sizes on perceived body sizes were different for those of different Body Mass Index. A backpropagation neural net model was used to model the nonlinear relationship between the perceived body size and the body's BMI, body chest girth, and garment ease (difference between garment and body chest girth). When the BMI was less than 20, wearing larger-sized T-shirts tended to increase perceived body size. For large chest sizes and for taller persons (BMI of 20 to 28) large garments made the wearer look thinner. However, for small persons (BMI of 20 to 28) effect of garment size was relatively small. Obese persons (BMI of > 28), wearing garments too tight or too loose were perceived as larger. Minimum perceived body size was found for garment ease of 2 to 3 cm.

Neuronal avalanches and criticality: A dynamical model for homeostasis

The dynamics of microelectrode local field potentials from cortical slice cultures shows critical behavior. A desirable feature of criticality is that information transmission is optimal in this state. We explore a biologically plausible neural net model that can dynamically converge on criticality and that can return to criticality if perturbed away from it. Our model assumes the presence of a preferred target firing rate, with dynamical adjustments of internodal connection strengths to approach this firing rate. We suggest that mechanisms for maintaining firing rate homeostasis may also maintain a neural system at criticality.

Acoustic speech recognition method and system using stereo vision neural networks with competition and cooperation

A method and system are provided for speech recognition. The speech recognition method includes the steps of preparing training data representing acoustic parameters of each of phonemes at each time frame; receiving an input signal representing a sound to be recognized and converting the input signal to input data; comparing the input data at each frame with the training data of each of the phonemes to derive a similarity measure of the input data with respect to each of the phonemes; and processing the similarity measures obtained in the comparing step using a neural net model governing development of activities of plural cells to conduct speech recognition of the input signal. In the processing step, each cell is associated with one respective phoneme and one frame, a development of the activity of each cell at each frame in the neural net model is suppressed by the activities of other cells on the same frame corresponding to different phonemes, and the development of the activity of each cell at each frame being enhanced by the activities of other cells corresponding to the same phoneme at different frames. In the process, the phoneme of a cell that has developed the highest activity is determined as a winner at the corresponding frame to produce a list of winners at respective frames. A phoneme is outputted as a recognition result for the input signal in accordance with the list of the winners at the respective frames that have been determined in the step of processing.

Advances in Modeling and Simulation of Grinding Processes

In the last decade the relevance of modeling and simulation of grinding processes has significantly risen which is caused by industrial needs and is indicated by the number of publications and research activities in this area. This keynote paper results from a collaborative work within the STC G and gives an overview of the current state of the art in modeling and simulation of grinding processes: Physical process models (analytical and numerical models) and empirical process models (regression analysis, artificial neural net models) as well as heuristic process models (rule based models) are taken into account, and outlined with respect to their achievements in this paper. The models are characterized by the process parameters such as grinding force, grinding temperature, etc. as well as work results including surface topography and surface integrity. Furthermore, the capabilities and the limitations of the presented model types and simulation approaches will be exemplified.

Optimization of hot-wire thermal flow sensor based on a neural net model

In this paper, we apply an optimization technique that inverts the function of a novel thermal flow sensor. The nominal output of the present sensor is the increase in wire temperature and this output is a function of the time constant of the heated wire and thus also of the velocity of flow. At very low flow velocities, the response is determined almost entirely by the time constant of the wire while at high velocities the device acts almost like a constant current hot-wire anemometer. The present thermal flow sensor, which can be used over a large range of velocities, is well suited for measurements of steady or slowly varying unsteady flows in industrial applications. The calibration data of the sensor consists of a set of curves at different fluid temperatures. Each curve represents the output voltage of the sensor as a function of the flow velocity. An artificial neural net model that approximates the calibration data for the sensor has been trained and it has good accuracy. Based on the neural net model, the optimization algorithm determines the flow velocity of the flowing gas in a pipe if the thermal flow sensor voltage measurement and fluid temperature are known. The problem reduces to minimizing a positive cost function that measures the difference between the neural net approximated voltage and its desired value. The simulation results demonstrate the effectiveness of the algorithm.

Modelling of a novel hot-wire thermal flow sensor with neural nets under different operating conditions

Thermal flow sensors with a wide dynamic range are widely applied in practical fluid flow measurements to yield local velocity information and also for volume flow-rate measurements. The importance of such flow sensors inspired the authors’ investigations into wide-velocity range thermal sensors and the outcome of this work is summarized in this paper. The present novel sensor is mechanically the same as the hot-wire anemometer, but it is excited by discrete, widely separated, square waves of electrical current rather than a continuous current. The nominal output of the new sensor is a function of the time constant of the heated wire and thus also of the velocity of flow. The time constant decreases as the flow velocity increases, while the heat transfer increases. In this paper, the results obtained suggest that our measurements for flow velocity and volume flow rate are in very good agreement with the theoretical results for the present thermal flow sensor. A neural network has been trained with the output data for the flow sensor and tested on our measurements. It was observed that the quality of the results depends on the number of hidden neurons. The predicted values are close to the real ones which indicate the neural net model gives a good approximation for the calibration curve of the single wire thermal flow sensor under different operating temperatures. The sensor described here was developed for slowly changing unidirectional flows, and uses one wire of 12.5 μm diameter. It is excited at 30 Hz frequency and its usable flow velocity range is 0.01–25 m/s. This yields an effective operating range and corresponds to a bandwidth of 1–2500.

French Gender Attribution on the Basis of Similarity: A Comparison Between AM and Connectionist Models

In simplistic terms the psychological literature on concept representation currently tends to favour either prototype or exemplar-based theories. Similar distinctions can be found within the computational literature. This paper examines the predictions of two such models in the domain of French gender attribution. The comparison between a standard feed-forward neural net model (prototype) and Analogical Modeling (exemplar) is of interest, aside from the broad issues of conceptual representation, since arguments have been presented by proponents of the latter theory as to its theoretical superiority over connectionist models on a variety of grounds. These claims are argued here to be either mistaken or less than compelling. It is, therefore, of some value to evaluate the two models on purely empirical grounds. A variety of simulations using a range of different representations and databases are compared with respect to their ability to generalise gender classifications to a set of unknown French nouns. Analysis shows that both models produce results which are not statistically significantly different.

Analysis and nonlinear modeling of the mound-building ant Formica lugubris spatial multi-scale dynamic in a larch-tree stand of the southern French Alps

Determinism in the evolution of a mound-building ant Formica lugubris (Hymenoptera:Formicidae) colony and the impact of environmental perturbations were analyzed using several methods. Variation in dome volume of ant-hills and their activity were followed in a larch forest of the southern French Alps for 8 consecutive years. The dynamic of domes was graphically visualized, and the deterministic component of variations was assessed using linear and nonlinear models (neural networks) in the context of auto-regressive and spatial multi-scale dependences hypothesis. An analysis of residuals was carried out (errors from the best global model) and nonpredictable data were located in perturbed areas (forest clearings and wind-throws). The dynamic of ant colony in the stand was simulated constructing a web of interacting neural net models. Evolution of virtual ant-hills was in accordance with real observed dynamic. The study revealed a very active dynamical system resulting from ants self-organizing in dome construction and confirmed that silvicultural practices can have a negative impact on ant colonies.

“In-House Likeness”: Comparison of Large Compound Collections Using Artificial Neural Networks

Binary classification models able to discriminate between data sets of compounds are useful tools in a range of applications from compound acquisition to library design. In this paper we investigate the ability of artificial neural networks to discriminate between compound collections from various sources aiming at developing an "in-house likeness" scoring scheme (i.e. in-house vs external compounds) for compound acquisition. Our analysis shows atom-type based Ghose-Crippen fingerprints in combination with artificial neural networks to be an efficient way to construct such filters. A simple measure of the chemical overlap between different compound collections can be derived using the output scores from the neural net models.

Calibrating a neural network‐based urban change model for two metropolitan areas of the Upper Midwest of the United States

We parameterized neural net‐based models for the Detroit and Twin Cities metropolitan areas in the US and attempted to test whether they were transferable across both metropolitan areas. Three different types of models were developed. First, we trained and tested the neural nets within each region and compared them against observed change. Second, we used the training weights from one area and applied them to the other. Third, we selected a small subset (∼1%) of the Twin Cities area where a lot of urban change occurred. Four model performance metrics are reported: (1) Kappa; (2) the scale which correct and paired omission/commission errors exceed 50%; (3) landscape pattern metrics; and (4) percentage of cells in agreement between model simulations. We found that the neural net model in most cases performed well on pattern but not location using Kappa. The model performed well only in one case where the neural net weights from one area were used to simulate the other. We suggest that landscape metrics are good to judge model performance of land use change models but that Kappa might not be reliable for situations where a small percentage of urban areas change.

Work Site Trip Reduction Model and Manual

Today's transportation professionals often use the ITE Trip Generation Manual and the Parking Generation Manual for estimating future traffic volumes to base off-site transportation improvements and identify parking requirements. But these manuals are inadequate for assessing the claims made by specific transportation demand management (TDM) programs in reducing vehicle trips by a certain amount at particular work sites. This paper presents a work site trip reduction model (WTRM) that can help transportation professionals in assessing those claims. WTRM was built on data from three urban areas in the United States: Los Angeles, California; Tucson, Arizona; and nine counties in Washington State. The data consist of work sites’ employee modal characteristics aggregated at the employer level and a listing of incentives and amenities offered by employers. The dependent variable chosen was the change in vehicle trip rate that correlated with the goals of TDM programs. Two different approaches were used in the model-building process: linear statistical regression and nonlinear neural networks. For performance evaluation the data sets were divided into two disjoint sets: a training set, which was used to build the models, and a validation set, which was used as unseen data to evaluate the models. Because the number of data samples varied from the three areas, two training data sets were formed: one consisted of all training data samples from three areas and the other contained equally sampled training data from the three areas. The best model was the neural net model built on equally sampled training data.

A comparison of novices and experts in the identification of sonar signals

This research contrasted trained sonar operators with novices in their abilities to categorize sonar signals as being of biological or man-made origin. Although the sonar operators performed significantly better than the naïve subjects, the performance of the naïve subjects ( d′ = 1.39) was closer to that of the sonar operators ( d′ = 1.81) than to chance. Additionally, there were qualitative similarities in the two groups’ performances. In order to understand this, a new group of people listened to a subset of 76 of the sounds, and chose verbal descriptors from a detailed list of 146 words such as “pht”, “chuck-chuck”, “rainfall”, etc. to describe each sound. Analyses of the descriptor choices demonstrated that biological sound sources were identifiable as such simply because they sounded like animal noises. Man-made signals were identified as more percussive, a result confirmed by an acoustical analysis of the signals. Two scales were developed for the signals, indicating how “animal like” and how percussive they were, respectively. Either of the two scales could predict with about 75% accuracy the category of the signals, a level of accuracy equal to that of the novices, and not bested by neither a CART model nor a neural net model. A simple acoustical analysis of the percussiveness of the waveforms could predict people’s judgments, but not predict the objective classifications of the signals.

Barrier option pricing: modelling with neural nets

We report call option pricing for up-and-out style barrier options through the use of a neural net model. A synthetic data set was constructed from the real LIFFE standard option price data by use of the Rubenstein and Reiner analytic model (Risk September (1991) 28). Unbiased estimates at the 95% confidence level were achieved for realistic barriers (barrier 4% or more above max ( S 0 , X ) ).

Animal cognition

Animal cognition

A Hybrid Neural Network Approach for the Development of Friction Component Dynamic Model

In this research, a new hybrid neural network is developed to model engagement behaviors of automotive transmission wet friction component. Utilizing known first principles on the physics of engagement, special modules are created to estimate viscous torque and asperity contact torque as preprocessors to a two-layer neural network. Inside these modules, all the physical parameters are represented by neurons with various activation functions derived from first principles. These new features contribute to the improved performance and trainability over a conventional two-layer network model. Both the hybrid and conventional neural net models are trained and tested with experimental data collected from an SAE#2 test stand. The results show that the performance of the hybrid model is much superior to that of the conventional model. It successfully captures detailed characteristics of the friction component engagement torque as a function of time over a wide operating range.