In this paper, we combine particle swarm optimization (PSO) and levy flight to solve the problem of protein folding prediction, which is based on 3D AB off-lattice model. PSO has slow convergence speed and low precision in its late period, so we introduce levy flight into it to improve the precision and enhance the capability of jumping out of the local optima through particle mutation mechanism. Experiments show that the proposed method outperforms other algorithms on the accuracy of calculating the protein sequence energy value, which is turned to be an effective way to analyze protein structure.

Estimating the irregular function with multi-scale structure is a hard problem. The results achieved by the traditional kernel learning are often unsatisfactory, since underfitting and overfitting cannot be simultaneously avoided, and the performance relative to boundary is often unsatisfactory. In this paper, we investigate the data-based local reweighted regression model under kernel trick and propose an iterative method to solve the kernel regression problem, local reweighted multiple kernel regression (LR-MKR). The new framework of kernel learning approach includes two parts. First, an improved Nadaraya-Watson estimator based on blockwised approach is constructed to organize a data-driven localized reweighted criteria; Second, an iterative kernel learning method is introduced in a series decreased active set. Experiments on simulated and real data sets demonstrate the proposed method can avoid under fitting and over fitting simultaneously and improve the performance relative to the boundary effetely.

The key of fast identification algorithm of time-varying modal parameter based on subspace tracking is to find efficient and fast subspace-tracking algorithm. This paper presents a modified version of NIC(Novel Information Criterion) adopted in two-layer linear neural network learning for subspace tracking, which is applied in time-varying modal parameter identification algorithm based on subspace tracking and get a new time-varying modal parameter identification algorithm. Comparing with the original subspace-tracking algorithm, there is no need to set a key control parameter in advance. Simulation experiments show that new time-varying modal parameter identification algorithm has a faster convergence in the initial period and a real experiment under laboratory conditions confirms further its validity of the time-varying modal identification algorithm presented in this paper.

In this paper, we will study a two-stage model on complex networks. The dynamic behaviors of the model on a heterogeneous scale-free (SF) network are considered, where the absence of the threshold on the SF network is demonstrated, and the stability of the disease-free equilibrium is obtained. Four immunization strategies, proportional immunization, targeted immunization, acquaintance immunization and active immunization are applied in this model. We show that both targeted and acquaintance immunization strategies compare favorably to a proportional scheme in terms of effectiveness. For active immunization, the threshold is easier to apply practically.

With the rapid development of information technology, it is inevitable for the distributed mobile computing will evolve to the pervasive computing gradually and whose final goal is fusing the information space composed of computers with the physical space in which the people are working and living in. Furthermore, with the development of SOA, more and more pervasive applications have been composed of different kinds of services. So how to choose a suitable service from all the useable services is the most important step for pervasive computing. Compare to the traditional service selection we must take more care of the context as well as the quality of them in pervasive environment. However, most of existing researches pay no attention to QoC(Quality of Service) which may lead to unreliable selections. Therefore we proposed a middleware based service selection scheme to support QoC-aware service selection efficiently.

The hippocampal CA1 pyramid neuron has plenty of discharge actions. In the thesis, the dynamic characteristics of the hippocampal neuron model are analyzed and discussed by the neurodynamic theory and methods. Under a certain amplitude current’s stimulation, the change of gNa(the maximum conductance of the transient sodium channel) and gKdr (the maximum conductance of the delay rectification potassium channel) can cause different dynamic characteristics of the neuron model. The transient Na+ current(INa ) caused by gNa is indispensable in the discharge’s formation process of the model. The model can generate the discharge process only when gNa reaches a certain threshold. In the discharge process of the neuron model, gNa’s changing affects little and the ISIs approximate to a straight line. The delay rectification K+ current(Ikdr) caused by gKdr isn’t indispensable in the discharge’s formation process of the model. But gKdr’s changing affects much in the discharge process of the neuron model. With gKdr’s changing, the neuron model undergoes different dynamic bifurcation process, and has plenty of discharge patterns such as the chaos, period, and so on. This investigation is helpful to know and investigate the dynamic characteristics and the bifurcation mechanism of the hippocampal neuron; and it provides a certain theory assist to investigate the neural diseases such as the Alzheimer disease by neurodynamics.

In this report, the amino acid sequence of giant panda interferon-α (gpIFN-α) was determined and compared with 15 corresponding IFN-α sequences. Phylogenetic analysis showed that the 15 interferons fell into two large groups. The giant panda and ferret branched and were most closely related to fox and dog and evolved into a distinct phylogenetic lineage from that of eukaryotic mammalians which evolved into another lineage. After analyzing the encoded amino acid sequence of the gpIFN-α using bioinformatics, the results revealed that in the full amino acid sequence, there were no transmembrane domain, one N-glycosylation sites, eight O-glycosylation sites and nine antigenic determinants. Secondary structure analyzed showed that the Alpha helix, Extended strand, Beta turn and Random coil each occupied 60.37%(99aa), 4.88%(8aa), 9.76%, 25%(41aa) respectively. In conclusion, our results will give the opportunity to investigate more in detail function study in giant panda and add to studies on the evolution of the IFN system in vertebrates and avian more generally.

The previous research revealed some functional coupling among nodes in model of motor control in human brain, which described nondirectional synchronous actions among these nodes during movement-readiness state. However, causal relationships among these nodes, which represent some directional interactions in movement-readiness state, are still lack. In the present study, we used functional magnetic resonance imaging (fMRI) and conditional Granger causality (CGC) method to investigate the interactions in model of motor control in left hand’s movement readiness state. Our results showed that upper precuneus (UPCU) and cingulated motor area (CMA) revealed net causal influences with contra lateral supplementary motor areas and contra lateral caudate nucleus during the left hand’s movement-readiness state. The net causal flows among these nodes can construct a closed circuit, which is similar as the circuit found in monkey’s brain and in human’s brain in right hand’s movement readiness state. This confirmed that there was an intrinsic circuit for motor control in either right hand’s or left hand’s movement readiness. Moreover, the results of Out-In degrees indicated that bilateral primary sensorimotor areas revealed competitive relationship during left hand’s movement-readiness.