Home About us Contact
|
Home About us Contact | ||
|
|
||
Neural Network Simulations (neural + network_simulation)
Selected AbstractsAn Empirical and Computational Investigation of Perceiving and Remembering Event Temporal RelationsCOGNITIVE SCIENCE - A MULTIDISCIPLINARY JOURNAL, Issue 3 2009Shulan Lu Abstract Events have beginnings, ends, and often overlap in time. A major question is how perceivers come to parse a stream of multimodal information into meaningful units and how different event boundaries may vary event processing. This work investigates the roles of these three types of event boundaries in constructing event temporal relations. Predictions were made based on how people would err according to the beginning state, end state, and overlap heuristic hypotheses. Participants viewed animated events that include all the logical possibilities of event temporal relations, and then made temporal relation judgments. The results showed that people make use of the overlap between events and take into account the ends and beginnings, but they weight ends more than beginnings. Neural network simulations showed a self-organized distinction when learning temporal relations between events with overlap versus those without. [source] Prediction of biodegradation from the atom-type electrotopological state indicesENVIRONMENTAL TOXICOLOGY & CHEMISTRY, Issue 10 2001Jarmo Huuskonen Abstract A group contribution method based on atom-type electrotopological state indices for predicting the biodegradation of a diverse set of 241 organic chemicals is presented. Multiple linear regression and artificial neural networks were used to build the models using a training set of 172 compounds, for which the approximate time for ultimate biodegradation was estimated from the results of a survey of an expert panel. Derived models were validated by using a leave-25%-out method and against two test sets of 12 and 57 chemicals not included in the training set. The squared correlation coefficient (r2) for a linear model with 15 structural parameters was 0.76 for the training set and 0.68 for the test set of 12 molecules. The model predicted correctly the biodegradation of 48 chemicals in the test set of 57 molecules, for which biodegradability was presented as rapid or slow. The use of artificial neural networks gave better prediction for both test sets when the same set of parameters was tested as inputs in neural network simulations. The predictions of rapidly biodegradable chemicals were more accurate than the predictions of slowly bio-degradable chemicals for both the regression and neural network models. [source] Energy Group optimization for forward and inverse problems in nuclear engineering: application to downwell-logging problemsGEOPHYSICAL PROSPECTING, Issue 2 2006Elsa Aristodemou ABSTRACT Simulating radiation transport of neutral particles (neutrons and ,-ray photons) within subsurface formations has been an area of research in the nuclear well-logging community since the 1960s, with many researchers exploiting existing computational tools already available within the nuclear reactor community. Deterministic codes became a popular tool, with the radiation transport equation being solved using a discretization of phase-space of the problem (energy, angle, space and time). The energy discretization in such codes is based on the multigroup approximation, or equivalently the discrete finite-difference energy approximation. One of the uncertainties, therefore, of simulating radiation transport problems, has become the multigroup energy structure. The nuclear reactor community has tackled the problem by optimizing existing nuclear cross-sectional libraries using a variety of group-collapsing codes, whilst the nuclear well-logging community has relied, until now, on libraries used in the nuclear reactor community. However, although the utilization of such libraries has been extremely useful in the past, it has also become clear that a larger number of energy groups were available than was necessary for the well-logging problems. It was obvious, therefore, that a multigroup energy structure specific to the needs of the nuclear well-logging community needed to be established. This would have the benefit of reducing computational time (the ultimate aim of this work) for both the stochastic and deterministic calculations since computational time increases with the number of energy groups. We, therefore, present in this study two methodologies that enable the optimization of any multigroup neutron,, energy structure. Although we test our theoretical approaches on nuclear well-logging synthetic data, the methodologies can be applied to other radiation transport problems that use the multigroup energy approximation. The first approach considers the effect of collapsing the neutron groups by solving the forward transport problem directly using the deterministic code EVENT, and obtaining neutron and ,-ray fluxes deterministically for the different group-collapsing options. The best collapsing option is chosen as the one which minimizes the effect on the ,-ray spectrum. During this methodology, parallel processing is implemented to reduce computational times. The second approach uses the uncollapsed output from neural network simulations in order to estimate the new, collapsed fluxes for the different collapsing cases. Subsequently, an inversion technique is used which calculates the properties of the subsurface, based on the collapsed fluxes. The best collapsing option is chosen as the one that predicts the subsurface properties with a minimal error. The fundamental difference between the two methodologies relates to their effect on the generated ,-rays. The first methodology takes the generation of ,-rays fully into account by solving the transport equation directly. The second methodology assumes that the reduction of the neutron groups has no effect on the ,-ray fluxes. It does, however, utilize an inversion scheme to predict the subsurface properties reliably, and it looks at the effect of collapsing the neutron groups on these predictions. Although the second procedure is favoured because of (a) the speed with which a solution can be obtained and (b) the application of an inversion scheme, its results need to be validated against a physically more stringent methodology. A comparison of the two methodologies is therefore given. [source] Computational constraints between retrieving the past and predicting the future, and the CA3-CA1 differentiationHIPPOCAMPUS, Issue 5 2004Alessandro Treves Abstract The differentiation between the CA3 and CA1 fields of the mammalian hippocampus is one of the salient traits that set it apart from the organization of the homologue medial wall in reptiles and birds. CA3 is widely thought to function as an autoassociator, but what do we need CA1 for? Based on evidence for a specific role of CA1 in temporal processing, I have explored the hypothesis that the differentiation between CA3 and CA1 may help solve a computational conflict. The conflict is between pattern completion, or integrating current sensory information on the basis of memory, and prediction, or moving from one pattern to the next in a stored sequence. CA3 would take care of the former, while CA1 would concentrate on the latter. I have found the hypothesis to be only weakly supported by neural network simulations. The conflict indeed exists, but two mechanisms that would relate more directly to a functional CA3-CA1 differentiation were found unable to produce genuine prediction. Instead, a simple mechanism based on firing frequency adaptation in pyramidal cells was found to be sufficient for prediction, with the degree of adaptation as the crucial parameter balancing retrieval with prediction. The differentiation between the architectures of CA3 and CA1 has a minor but significant, and positive, effect on this balance. In particular, for a fixed anticipatory interval in the model, it increases significantly the information content of hippocampal outputs. There may therefore be just a simple quantitative advantage in differentiating the connectivity of the two fields. Moreover, different degrees of adaptation in CA3 and CA1 cells were not found to lead to better performance, further undermining the notion of a functional dissociation. © 2004 Wiley-Liss, Inc. [source] | ||||||||||