Particle Competition and Cooperation in Networks for Semi-Supervised Learning

Semi-supervised learning is one of the important topics in machine learning, concerning with pattern classification where only a small subset of data is labeled. In this paper, a new network-based (or graph-based) semi-supervised classification model is proposed. It employs a combined random-greedy walk of particles, with competition and cooperation mechanisms, to propagate class labels to the whole network. Due to the competition mechanism, the proposed model has a local label spreading fashion, i.e., each particle only visits a portion of nodes potentially belonging to it, while it is not allowed to visit those nodes definitely occupied by particles of other classes. In this way, a "divide-and-conquer" effect is naturally embedded in the model. As a result, the proposed model can achieve a good classification rate while exhibiting low computational complexity order in comparison to other network-based semi-supervised algorithms. Computer simulations carried out for synthetic and real-world data sets provide a numeric quantification of the performance of the method.

Semi-supervised learning guided by the modularity measure in complex networks

Semi-supervised learning techniques have gained increasing attention in the machine learning community, as a result of two main factors: (1) the available data is exponentially increasing; (2) the task of data labeling is cumbersome and expensive, involving human experts in the process. In this paper, we propose a network-based semi-supervised learning method inspired by the modularity greedy algorithm, which was originally applied for unsupervised learning. Changes have been made in the process of modularity maximization in a way to adapt the model to propagate labels throughout the network. Furthermore, a network reduction technique is introduced, as well as an extensive analysis of its impact on the network. Computer simulations are performed for artificial and real-world databases, providing a numerical quantitative basis for the performance of the proposed method.

Understanding bias in geographic range size estimates

Aim Estimates of geographic range size derived from natural history museum specimens are probably biased for many species. We aim to determine how bias in these estimates relates to range size. Location We conducted computer simulations based on herbarium specimen records from localities ranging from the southern United States to northern Argentina. Methods We used theory on the sampling distribution of the mean and variance to develop working hypotheses about how range size, defined as area of occupancy (AOO), was related to the inter-specific distribution of: (1) mean collection effort per area across the range of a species (MC); (2) variance in collection effort per area across the range of a species (VC); and (3) proportional bias in AOO estimates (PBias: the difference between the expected value of the estimate of AOO and true AOO, divided by true AOO). We tested predictions from these hypotheses using computer simulations based on a dataset of more than 29,000 herbarium specimen records documenting occurrences of 377 plant species in the tribe Bignonieae (Bignoniaceae). Results The working hypotheses predicted that the mean of the inter-specific distribution of MC, VC and PBias were independent of AOO, but that the respective variance and skewness decreased with increasing AOO. Computer simulations supported all but one prediction: the variance of the inter-specific distribution of VC did not decrease with increasing AOO. Main conclusions Our results suggest that, despite an invariant mean, the dispersion and symmetry of the inter-specific distribution of PBias decreases as AOO increases. As AOO increased, range size was less severely underestimated for a large proportion of simulated species. However, as AOO increased, range size estimates having extremely low bias were less common.

Morphological Alterations in Newly Born Dentate Gyrus Granule Cells That Emerge after Status Epilepticus Contribute to Make Them Less Excitable

Computer simulations of external current stimulations of dentate gyrus granule cells of rats with Status Epilepticus induced by pilocarpine and control rats were used to evaluate whether morphological differences alone between these cells have an impact on their electrophysiological behavior. The cell models were constructed using morphological information from tridimensional reconstructions with Neurolucida software. To evaluate the effect of morphology differences alone, ion channel conductances, densities and distributions over the dendritic trees of dentate gyrus granule cells were the same for all models. External simulated currents were injected in randomly chosen dendrites belonging to one of three different areas of dentate gyrus granule cell molecular layer: inner molecular layer, medial molecular layer and outer molecular layer. Somatic membrane potentials were recorded to determine firing frequencies and inter-spike intervals. The results show that morphologically altered granule cells from pilocarpine-induced epileptic rats are less excitable than control cells, especially when they are stimulated in the inner molecular layer, which is the target area for mossy fibers that sprout after pilocarpine-induced cell degeneration. This suggests that morphological alterations may act as a protective mechanism to allow dentate gyrus granule cells to cope with the increase of stimulation caused by mossy fiber sprouting.

Theoretical study of the absorption and nonradiative deactivation of 1-nitronaphthalene in the low-lying singlet and triplet excited states including methanol and ethanol solvent effects

The photophysics of the 1-nitronaphthalene molecular system, after the absorption transition to the first singlet excited state, is theoretically studied for investigating the ultrafast multiplicity change to the triplet manifold. The consecutive transient absorption spectra experimentally observed in this molecular system are also studied. To identify the electronic states involved in the nonradiative decay, the minimum energy path of the first singlet excited state is obtained using the complete active space self-consistent field//configurational second-order perturbation approach. A near degeneracy region was found between the first singlet and the second triplet excited states with large spin-orbit coupling between them. The intersystem crossing rate was also evaluated. To support the proposed deactivation model the transient absorption spectra observed in the experiments were also considered. For this, computer simulations using sequential quantum mechanic-molecular mechanic methodology was used to consider the solvent effect in the ground and excited states for proper comparison with the experimental results. The absorption transitions from the second triplet excited state in the relaxed geometry permit to describe the transient absorption band experimentally observed around 200 fs after the absorption transition. This indicates that the T-2 electronic state is populated through the intersystem crossing presented here. The two transient absorption bands experimentally observed between 2 and 45 ps after the absorption transition are described here as the T-1 -> T-3 and T-1 -> T-5 transitions, supporting that the intermediate triplet state (T-2) decays by internal conversion to T-1. (C) 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4738757]

Stochastic competitive learning in complex networks

Competitive learning is an important machine learning approach which is widely employed in artificial neural networks. In this paper, we present a rigorous definition of a new type of competitive learning scheme realized on large-scale networks. The model consists of several particles walking within the network and competing with each other to occupy as many nodes as possible, while attempting to reject intruder particles. The particle's walking rule is composed of a stochastic combination of random and preferential movements. The model has been applied to solve community detection and data clustering problems. Computer simulations reveal that the proposed technique presents high precision of community and cluster detections, as well as low computational complexity. Moreover, we have developed an efficient method for estimating the most likely number of clusters by using an evaluator index that monitors the information generated by the competition process itself. We hope this paper will provide an alternative way to the study of competitive learning.

Network-based stochastic semisupervised learning

Semisupervised learning is a machine learning approach that is able to employ both labeled and unlabeled samples in the training process. In this paper, we propose a semisupervised data classification model based on a combined random-preferential walk of particles in a network (graph) constructed from the input dataset. The particles of the same class cooperate among themselves, while the particles of different classes compete with each other to propagate class labels to the whole network. A rigorous model definition is provided via a nonlinear stochastic dynamical system and a mathematical analysis of its behavior is carried out. A numerical validation presented in this paper confirms the theoretical predictions. An interesting feature brought by the competitive-cooperative mechanism is that the proposed model can achieve good classification rates while exhibiting low computational complexity order in comparison to other network-based semisupervised algorithms. Computer simulations conducted on synthetic and real-world datasets reveal the effectiveness of the model.

Molecular Dynamics Investigations of PRODAN in a DLPC Bilayer

Molecular dynamics computer simulations have been performed to identify preferred positions of the fluorescent probe PRODAN in a fully hydrated DLPC bilayer in the fluid phase. In addition to the intramolecular charge-transfer first vertical excited state, we considered different charge distributions for the electronic ground state of the PRODAN molecule by distinct atomic charge models corresponding to the probe molecule in vacuum as well as polarized in a weak and a strong dielectric solvent (cyclohexane and water). Independent on the charge distribution model of PRODAN, we observed a preferential orientation of this molecule in the bilayer with the dimethylamino group pointing toward the membrane's center and the carbonyl oxygen toward the membrane's interface. However, changing the charge distribution model of PRODAN, independent of its initial position in the equilibrated DLPC membrane, we observed different preferential positions. For the ground state representation without polarization and the in-cyclohexane polarization, the probe maintains its position close to the membrane's center. Considering the in-water polarization model, the probe approaches more of the polar headgroup region of the bilayer, with a strong structural correlation with the choline group, exposing its oxygen atom to water molecules. PRODAN's representation of the first vertical excited state with the in-water polarization also approaches the polar region of the membrane with the oxygen atom exposed to the bilayer's hydration shell. However, this model presents a stronger structural correlation with the phosphate groups than the ground state. Therefore, we conclude that the orientation of the PRODAN molecule inside the DLPC membrane is well-defined, but its position is very sensitive to the effect of the medium polarization included here by different models for the atomic charge distribution of the probe.

Ranking Configurations of Shading Devices by Its Thermal and Luminous Performance

This paper evaluates the thermal and luminous performance of different louver configurations on an office room model located in Maceió-AL (Brazil), ranking the alternatives in a way that leads to choices for alternatives with potential balanced performance. Parametric analyses were done, based on computer simulations on software Troplux 5 and DesignBuilder 2. The variables examined were number of slats, slat slope and slat reflectance, considering the window facing North, South, East and West and a fixed shading mask for each orientation. Results refer to internal average illuminance and solar heat gains through windows. It was observed that configurations of shading devices with the same shading mask may have different luminous and thermal performance. The alternatives were ranked, so the information here produced has the potential to support decisions on designing shading devices in practice.

Computer simulation study of collective dynamics in the glass former Ca(NO3)(2)center dot 4H(2)O

Time correlation functions of current fluctuations were calculated by molecular dynamics (MD) simulations in order to investigate sound waves of high wavevectors in the glass-forming liquid Ca(NO3)(2)center dot 4H(2)O. Dispersion curves, omega(k), were obtained for longitudinal (LA) and transverse acoustic (TA) modes, and also for longitudinal optic (LO) modes. Spectra of LA modes calculated by MD simulations were modeled by a viscoelastic model within the memory function framework. The viscoelastic model is used to rationalize the change of slope taking place at k similar to 0.3 angstrom(-1) in the omega(k) curve of acoustic modes. For still larger wavevectors, mixing of acoustic and optic modes is observed. Partial time correlation functions of longitudinal mass currents were calculated separately for the ions and the water molecules. The wavevector dependence of excitation energies of the corresponding partial LA modes indicates the coexistence of a relatively stiff subsystem made of cations and anions, and a softer subsystem made of water molecules. (C) 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4751548]

Generalized Metropolis dynamics with a generalized master equation: An approach for time-independent and time-dependent Monte Carlo simulations of generalized spin systems

The extension of Boltzmann-Gibbs thermostatistics, proposed by Tsallis, introduces an additional parameter q to the inverse temperature beta. Here, we show that a previously introduced generalized Metropolis dynamics to evolve spin models is not local and does not obey the detailed energy balance. In this dynamics, locality is only retrieved for q = 1, which corresponds to the standard Metropolis algorithm. Nonlocality implies very time-consuming computer calculations, since the energy of the whole system must be reevaluated when a single spin is flipped. To circumvent this costly calculation, we propose a generalized master equation, which gives rise to a local generalized Metropolis dynamics that obeys the detailed energy balance. To compare the different critical values obtained with other generalized dynamics, we perform Monte Carlo simulations in equilibrium for the Ising model. By using short-time nonequilibrium numerical simulations, we also calculate for this model the critical temperature and the static and dynamical critical exponents as functions of q. Even for q not equal 1, we show that suitable time-evolving power laws can be found for each initial condition. Our numerical experiments corroborate the literature results when we use nonlocal dynamics, showing that short-time parameter determination works also in this case. However, the dynamics governed by the new master equation leads to different results for critical temperatures and also the critical exponents affecting universality classes. We further propose a simple algorithm to optimize modeling the time evolution with a power law, considering in a log-log plot two successive refinements.