Based on the characteristic of internet date transmission,a cellular automaton model of the date transmission with open boundary condition is proposed.The model is used to study the mean queue length,sending rate and sojourn time versus the input probability α,the output probability β and themaximum input rate Jin in the routing node.Simulation results show that α and β can change themean sending rate in a certain range.There are free phase and jam phase in thesystem.The phase diagram of system is obtained and analyzed.These studies provide the reference for the management of the realistic data transmission.

Among Intersection signal control in urban areas,thePhasing-Sequence design has a very close relationship with the traffic flow path through the intersection.Without changing the phase combination of circumstances,a reasonable Phasing-Sequence can greatly enhance the efficiency of intersections.By analyzing the conflict point of traffic tracks in intersection and ensuring the safe passage of vehicles to pursuit the maximum of green signal ratio,this paper establishes a Phasing-Sequence optimization model.Taking the intersection of Guangzhou as an example,the Phasing-Sequence optimization resultedin an upgrade of efficiency of 13.4%.

Urban heat island effect could be felt with in karstcity,Guilin rapid expansion.After the mixed pixel emissivity of karst hills is computed,land surface temperature (LST) of the karst city could be retrieve from remote sensing image with mono-window algorithm in.Results demonstrated that the difference between the retrieved LST and measured LST is within 2 °C.LST values in the old central district are lower than those in the developing district and industrial districts.LST possesses a negative correlation with vegetation fraction as it decreases with vegetation fraction increasing.LST of water is the minimum,about17 °C.As the karst hills are vegetated,LST of the shaded slope ranges from 18to 19 °C and LST of sunlit slope from 22 to 24 °C.However,the LST range of concrete squares and large buildings is from 32 to 33 °C.As factories with the crowded buildings emit much heat,their LST values range from 33 to 36 °C,the maximum.LST values of some pixels even reach 40 °C.The vegetated karst hills show the urban cool island effect and can improve urban microclimate environment in the karst cities.

In the nonparametric regression model Y_i=g(x_i)+ε_i,1≤i≤n,the strong consistency of the nonparametric regressionweighted function estimator g_n(x)=∑ni=1w_ni(x)Y_i is studied,where ｛ε_i,1≤i≤n｝ are uniformly integrabilited and negatively associated samples.

An example of a strictly stationary associated randomsequence which does not satisfy the central limit theorem and whose partial sums' variance grows in a defined regular way is constructed.The well-known example of N.Herrndorf is generalized and the optimality of conditions in the classical Newman's theorem is shown.

In this paper,a class of new conjugate gradient method is presented,and its global convergence is achieved under the Wolfe line search.Some numerical tests are carried out and the results show that the algorithm is effective.

A subgroup H of a group G is said to be weakly s-permutably embedded in G if there are a subnormal subgroup T ofG and an s-permutably embedded subgroup H_se of G contained in H such that G=HT and H∩T≤H_se.By using the property ofweakly s-permutably embedded groups,some criterions of p-nilpotency offinite groups are obtained.

Reducing and radical scavenging abilities of different extracts from leaves of Streblus asper against DPPH,hydroxyl and superoxide radicals were measured by means of photometric method.The results showed that the different extracts of leaves of Streblus asper have obvious antioxidant activity,amongwhichethyl acetate (EA) fraction exhibited the highest activity.The antioxidant activities of EA extract of Streblus asper leaves was better than that of the synthetic antioxidant (BHT).The EA extract of leaves of Streblus asper was a good natural antioxidant.

A selective and rapid high-performance liquid chromatography coupled with resonance Rayleigh scattering detection method was developed for the determination of multicomponent gentamicin (Gen).Pontamine sky blue was chosen as molecular recognition probe,which could forms ion-association complex with Gen inmobile medium and produce strong scattering signal.The separation of Gen components was achieved on a C₁₈ column and the column eluent was monitored usingfluorescence detection at λex=λem=356 nm.The detection limitsof 6.0～7.2 mg/L were reached,and the response linearity between peak area and components concentration were found in the range of 7.2～115,8.2～132,7.5～79 and 8.5～78 mg/L for C₁,C_1a,C₂,C_2a,respectively.This method has been applied to the determination of Gen components in commercial Gen sulfate injection,and the detection results were in good agreement with the specification of Chinese Pharmacopoeia.

A novel method of capillary electrophoresis (CE) withtris(2,2′-bipyridyl)ruthenium(Ⅱ) electrochemiluminescence (ECL) detection has been developed for the determination of haloperidol in human urine.The parameters such as workelectrode potential,concentration and pH of phosphate buffer solution on effectof ECL intensity of haloperidol were investigated.Under optimized conditions,haloperidol concentration was linear with ECL peak area in the range of 0.01～10mg·L^-1.The detection limit (3σ) for haloperidol was 3.5 μg·L^-1,and the RSD was 1.6% for 3.0 mg·L^-1 haloperidol (n=11).The recoveries were between 95.0% and 99.5%.

α/β Si₃N₄ composites were sintered by hot pressing at 1 600～1 750 °C using MgSiN₂ as additives,and the relationship between mechanical properties and phase composition,microstructure was investigated.Microhardness increased with the increase of the content of α-Si₃N₄,achieved the highest valuewhen the content of α-Si₃N₄ is about 60%,and tended to a steady value (about 23 GPa) with further increase of the content of α-Si₃N₄.With the increaseof the content of β-Si₃N₄,fracture toughness increased and then decreased,and strength increased first and then had little change.

The interaction of Fe₃O₄ magnetic nanoparticles with Hyaluronidase wasinvestigated by fluorescence,ultraviolet and circular dichroism spectral method(CD).The experimental results demonstrated that the quenching mechanism of Fe₃O₄magnetic nanoparticles to Hyaluronidase fluorescence was a static quenching dueto existence of non-radiative energy-transfer.The binding constant of Fe₃O₄ magnetic nanoparticles with Hyaluronidase is four factors of ten,which indicate there is strong association between them.The changes of the thermodynamic parameters showed that the main interaction forces are electrostatic reaction and the hydrophobic interaction between Fe₃O₄ magnetic nanoparticles and Hyaluronidase.At the same time,through the ultraviolet absorption and circular dichroism(CD) spectroscopy,it showed that the secondary structureof Hyaluronidase is changed induced by the binding of Fe₃O₄ magnetic nanoparticles.

Primary dark coniferous forests,dominated with firs,were cut on a large scale from 1950s to 1980s in subalpine region of western Sichuan,China.At thesame time,native tree species regenerated naturally.As a result,there formed theforests with different restoration period.Roots biomass of forests with different restoration period were investigated in a way of space-for-time substitution.The changes and their distribution patterns of roots biomass were analyzed following forests restoration in western Sichuan.The results showed that roots biomass increased gradually with the restoration of forests whether at lower altitude (2 900~3 250 m) or at higher altitude (3 250~3 600 m).At lower altitude,thefluctuation of roots biomass is between 21.000 and 51.583 t/hm².While at higheraltitude,the fluctuation of roots biomass is between 20.363 and 35.316 t/hm².As for the same forests age,roots biomass located at higher altitude is less than that at lower altitude.In lower altitude,the most value of roots biomass is occurred in the soil layer of 10~20 cm,while at higher altitude,which is occurred inthe soil layer of 0~10 cm.The roots biomass in soil layer of 0~30 cm exceeded 90% of total roots biomass.The biomass of coarse roots (＞5 mm) exceeded 65.0% among the three diameter classes,while the small roots (2~5 mm) accounted for theleast percentage of total roots biomass.As for the same forests age,there isa trend that more roots biomass located at lower altitude than at higher altitude,whether coarse roots,small roots or fine roots (＜2 mm).

Diurnal changes of net photosynthetic rate (P_n)of Bauhinia championii on Guilin's Karst hills were measured by Li-6400 portable photosynthesis analysis system (LI-6400,Li-cor USA) in different seasons.Itwas found that diurnal curves of P_n were “single peak” in spring and winter,but “double peak” in summer and autumn and they were both with a significant “midday depression of photosynthesis”.P_n reduction at noon in summer was mainly originated from nonstomatal limitation,while it was mainly due to stomatal limitation in autumn;Bauhinia championii showed higher growth potential in summer and autumn,the seasonal changes of P_n were in the order of summer＞autumn＞spring＞winter;The positive relationship models between P_nand its influence factors were all significant (P＜0.01) in different seasons.Statistical analysis indicated that diurnal changes of P_n were affected by stomatal conductance (G_s) and intercellular CO₂ concentration (C_i) in spring,photosynthetic active radiation (PAR) was the main factor influencing P_n in summer,the dominant factor for P_n in autumn was G_s,and P_n in winter was contributed to the the co-effects of C_i andtranspiration rate (T_r).

16 species and one variety of ferns are reported forthe first time in Guangxi,China.They are Osmunda angustifolia,Osmunda mildei,Crepidomanes pinnatifidum,Adiantum juxtapositum,Antrophyum vittarioides,Allantodia pinnatifido-pinnata,Leptogramma tottoides,Asplenium interjectum,Polystichum del-todon var.cultripinnum,Polystichum minutissimum,Polystichum bifidum,Polystichum kwangtungense,Polystichum subacutidens,Tectaria leptophylla,Phymatopteris quasidivaricata,Lepidomicrosorum hederaceum,Loxogramme acroscopa.The citation of specimenand distribution have been enumerated in this paper.All the specimens examinedare preserved in Herbarium of Guangxi Institute of Botany,Guangxi Zhuang Autonomous Region and the Chinese Academy of Sciences (IBK).

The effects of different hormone processing,remain leaves,substrate,cutting time and light intensities on survival rate of Camellia nitidissimawere investigated.The results showed that there were remarkable effects on survival rate of C.nitidissima with different hormone processing,remain leaves,substrate,cutting time and light intensities,especially light intensities and remain leaves.The most suitable time for cutting C.nitidissima in northern Guangxi is in the middle of April.It can get higher rooting rate of cutting by usingoneyear shoot with two leaves on the top,dipping the cuttings in concentration of100 mg·L^-1 NAA for 18 h,yellow soil as medium for cutting bed and 90% forlight intensities.

Playing behavior of infants were studied in a group of captive Assamesemacaques (Macaca assamensis) in Nanning Zoo,Guangxi,from August 15 to October 25.2 290 playing bouts were recorded in 42 days of observations (210 hours ofobservations) with focal animal sampling and continuous recording.The results indicated that male infants spent significantly more time,exhibited significantly higher frequency in social playing (special fight playing) than females,andwere more active than females as well.They preferred to play with other malesof same age.These results support the motor training hypothesis.The patterns of playing changed with infant's age.Nonsocial play (special solitary locomotor play) declined significantly while social play increased.Moreover,infantswere more active in social play.

Four kinds of soil were digested by electric heatingboard digestion and microwave digestion with two kinds of acid system respectively.Subsequently,concentration of Cu、Pb、Cr、Zn、Mn in the soils was determinated.The results showed that concentration of Cu、Pb、Zn in the standard soil sample digested by the four methods in good agreement with the reference values.The determinated value of Mn concentration in the standard soil sample digested by the two microwave digestionmethods in good agreement with the reference value.Only determinated valueof Cr concentration in the standard soil sample digested by microwave digestion method with HNO₃+HF+HClO₄ agreed with the reference value.There was no significant differencein the concentration of Cu、Pb、Zn in the standard soil sample,soil A (sampled from biological garden of Guangxi Normal University),soil B and soil C (sampled from recovery region and tailing region of Guangxi Lipu Manganese mine).But there was significant difference in the concentration of Cr in the four soils andin the concentration of Mn in the soil B.Conclusively,microwave digestion method with HNO₃+HF+HClO₄ was the most ideal method which could digest effectively all of the five heavy metals in the soils.

Set-valued information systems are generalized models of single-valued information systems.It could be divided into two categories:conjunctive and disjunctive set-valued information systems.Through analyzing the limitations of four existing dominance relations in disjunctive set-valued information systems,a novel variable precision dominance relation is proposed anda disjunctive set-valued ordered information system based on variable precisiondominance relation is defined.It is an extension of the classical rough set theory.Then,an approach for sorting objects is introduced to the disjunctive set-valued ordered information systems,which is based on the whole dominance degree of an object in an information system.Finally,a case of a venture investmentproved it to be a valid approach for extracting decision rules from disjunctive set-valued ordered information systems.

Inthis paper,a kind of new condition dependability is proposed for ordered decision tables based on dominance classes,which can be used to give the definition ofattribute significance.Then,a feature selection algorithm is put forward for ordered decision tables,and the validity of this algorithm is exemplified by anillustrative example.

In this paper,the characteristics of lower and upperapproximation basedon Pawlak rough set theory are introduced and analyzed.Combined with approximation-degree,the method of defining upper approximation on covering rough set theory is studied.Then,a new concept named minimal upper approximation is proposed,and its existence and uniqueness are analyzed.In the end,this paper presents and proves the relationship between minimal upper approximation and other upper approximations.

An improved algorithm based on CMAC (cerebella modelarticulation controller) and named DCMAC-AL is proposed.It uses advantage(λ) learning to calculate the state-action function,emphasizes the differences among action values and shuns action oscillation.It creates novel features based on Bellman error to improvethe adaption of CMAC.Besides,it provides a mathematic model for takeaway in RoboCup Soccer Simulation and experiment with DCMAC-AL.The results demonstrate thatDCMAC-AL outperforms advantage(λ) learning in regard to learning effort.

To minimize “0-1” loss,most of conventional classification algorithmsnon-explicitly assume that all results of classification are accepted.However,the assumption is inappliable to knowledge extraction in such fields as medical diagnosis,fault diagnosis and fraud detection.In this paper,the algorithm Cost-sensitive SVM with Class-dependent Misclassification Cost and Class-dependent Reject Cost (CSVM-CMC²RC) is proposed.In CSVM-CMC²RC algorithm,firstly,acost-sensitive SVM is trained to obtain the preliminary classification results.Secondly,the post probability of each sample is computed.Thirdly,the classification reliability of each sample is estimated.Finally,the optimal reject threshold and the final reject decision are determined based on minimizing the average cost.Experimental results demonstrate that the proposed CSVM-CMC²RC algorithm can reduce the misclassification rate and average cost,and the classification reliability is improved.

With the brief introduction of predictive data miningtask,this paper discusses how to reduce the high dimensionality of data samples and classify the dataset correctly,which has very few samples but large amounts of features or variables.Afterwards,this paper proposes a novel classifier based on the above discussion,which is easily implemented.Finally,the application of the proposed classifier is discussed.

Classification is an important sector of Data Mining,and decision tree is one of the efficient classification methods used constantly.Nowadays,thereare several classification algorithms which use decision tree,for instance ID3,C4.5 and CART.But there are some disadvantages to use them,for example,classset must be input manually,attribute must be separated and the best class setis needed.This paper makes the following contributions to avoid these disadvantages:on the one hand,proposing a new concept of using information gain to lineGEP chromosome's head;on the other hand,proposing the algorithms of IG-GEPDTAR (constructing decision tree attribution reduction algorithms with gene expression programming based on information gain) and validate it by using experiment data.The result shows that decision tree constructed by IG-GEPDTAR is absolutely correct and better,it has less redundancies than Candida Ferreira's,has 82.9%less nodes than that of ID3 algorithms,and has 31.2% less nodes than that of Wang Chunnian's.

From the effect of the linguistic hedge for truth values,linguistic hedge can be divided into three types through the qualitative method.Then six truth valuescan be obtained.Based on the six-element lattice implication algebra,the six lattice-valued propositional logic system which can express both comparable and incomparable information is proposed.Its properties and soft resolution methodon filter is discussed.Reasoning and operation are directly acted by linguistictruth values in the resolution process and the result is also preseuted with linguistic truth values.

The Tri-training,a semi-supervised learning method,has few restrictions on the classification algorithms and the training data.Moreover,it runs faster than other methods,so it is applied to compute thefreedom in Chinese phrase translation.The result shows that Tri-training can improve the accuracy of the computation of the freedom and have high learning effectiveness.

This paper studies the acquisition of comparable corpora and its application in cross-language information retrieval (CLIR).First,download news articles from news sites,and align them with Lucene,and acquire comparable corpora.Then translation knowledge is extracted from the aligned articles.At last,apply the translation knowledge on TDT4 to test the performance of CLIR system.Theexperiments show that the translation knowledge could improve the performance of CLIR,achieve the MAP value of 0.272 8,35.44 percentage points higher than the method based on dictionary.

Due to the fact that Chinese sentence analysis can not provide deep semanticinformation,a semantic chunk of question sentence analysis model is proposed based on HNC and description logics.With the knowledge of HNC's concept symbol to get semantic information about the word,and by starting with the connotationof the concept categories,it preliminarily analyzes the logical structure of specific question sentences.Finally,with description logic reasoning mechanism,asemantic view of question sentence is provided and its validity is proved.

Study on Chinese article comparison has important application value and practical significance.This paper applies the principles of granular computing tothe computation of similarity degree between Chinese articles.A tolerance granular computing model for computing similarity degree between Chinese articles isbuilt by introducing some concepts such as article tolerance granule,paragraphtolerance granule and granule space information table.Based on tolerance granular computing model,an algorithm for computing similarity degree between Chinese articles is proposed.And the effectiveness of the proposed algorithm is investigated by examples and test results.

Due to the fact that inadequate features of short text makes unsatisfactory classification,an improved algorithm fortext classification based on Bayesian network is presented.The improvement liesin the steps of Bayesian networkstructure learning phase for finding parent node.The algorithm takes into account the connection strength of the factors,thus,a decrease of interference errors between two nodes that are unrelated to the father and son nodes is achieved.It enables each node to find more accurate father node,and improves the classification accuracy of text.

An extended semantic feature representation method isintroduced,anda machine learning based disambiguation is performed using this feature.First,a named entity recognition system is used to detect gene mentions in the literature.Second,different searching strategies are adopted to construct mapping pairs.Thirdly,extended semantic feature is used for supervised machine learning based disambiguation.Then,retrieved Wikipedia results are used to build post-filter.This method achieves an F-measure of 83.2% on the BioCreative Ⅱ GN test dataset.

Most of the traditional sponsored search algorithms only concentrate on semantic similarity of keyword and ads,but this similarity doesn't mean thatthe user intent to buy the product,namely,it doesn't mean that the search keyword has commercial value.This paper focuses on keywords' commercial value,analyzes the relationship between keyword's commercial value and its features likepart of speech,length,popularity,and performs some qualitative analysis,whichmay be helpful for keyword-related advertising and keyword suggestions.

Currently it is difficult to quantitatively observe,analyze and control Emergent Internet Public Opinion (IPO).On the basis ofanalysis of user network relationship,a series of quantitative metrics are selected to describethe propagation of Internet Public Opinion.Their selection reasonsand computing method are explained in details.The validity of quantitative metrics is proved by experiments.According to the quantitative analysis metrics change apparently when the number of users change and these changes mayreflect the changes of Internet Public Opinion in different degrees and levels.

Based on the characteristis of P2P network,the beliefand human society,a new P2P trust valuation model on the basis of belief recommendation and external trust negotiation is put forward after disscussing and analizing relvant trust models.This model not onlyconsiders the external trust evaluation in every network exchange,but also inducts a subjective trust evaluation mechanism.The experiment result indicates thevalidity of the method.

In the video processing,shadow is often detected mistakenly for foreground in the segmentation with moving foreground because movingshadow shares the same characters with moving foreground.This may seriously affects post-processing such as tracking and identification.An improved algorithm is presented for shadow detection of moving object.Firstly,a Gaussian mixture model is constructed for each pixel to segment the moving object.For a pixel detected to be moving foreground in the shadow area,it is determined whether it is suspected shadowpixel according to the brightness characteristics of the 8 pixels in its neighboring region.Then all the suspected shadow is clustered by virtue of the color invariance in the Color space vector model for further shadow detection.Experimentalresults show that the detection method has a broad application prospects due to its high precision,efficiency,and rapidness.

As the face image always has a blur boundary and little gradient change,the region segmentations obtained by the original active contour model are generally unsatisfactory.To achieve more accurate facial contour extraction and face segmentation,a new face segmentation scheme based on curve evolution modelis proposed,which is a combination of face detection,active contour model and mathematical morphology operators.Moreover,an improved active contour model isproposed to increase the accuracy of face contour extraction and speed up the convergence process.Experimental results show that the improved active contour model can effectively detect the local blur and breaking boundaries without any fractures in the curve,resulting in a favorable face segmentation.In addition,the improved narrow-band method reduces the computation by 60%.

According to the volume balance principle that volumeof entering headchamber from cutterhead of shield machine is equal to the carried volume by screw conveyor,a discrete state equation is proposed for pressure control of headchamberin shield tunneling based on nonlinear constitutive model of waste soil.BP neural network and PID neural network are used as identifier and controller.The pressure control and model parameter identification of time-varying system of shield tunneling are studied.Numerical simulation result shows that the pressure control method based on neural network in shield tunneling is effective.The proposed algorithms have good stability and robustness.

Because of the complex series of phase and physical change,as well as large amounts of data in hot rolling of steel strip production process,the basic method of data mining can be used to model,extract rules and achieve the performance prediction and function evaluation of hot rolling strip production.The article uses radial basis function (RBF) neural network to model,and implement the performance prediction of hot rolling strip production.The RBF neural network is better than the traditional BP neural network when considering approximation capabilities,learning rate and so on.This paper shows the superiority of RBF neural network according to the network structures of RBF and BP.

For resolvingthe air defense weapon-target assignment of warships,an improved genetic algorithm based on dynamic adaptive is proposed.The new algorithm adopts adaptive encoding,adaptive choosing operator,adaptive crossover and mutation operator.As aresult,the algorithm can get global optimization resolution in a short time.Results demonstrate that the improved algorithm achieves better efficiency than some classical optimization algorithms,can avoid getting into local optimization,and solve air defense weapon-target assignment problems of warships well.

The research of Unmanned System's autonomy is a keyelement of,this field,and the adjustable autonomy is the method for transparently adjusting autonomy and adapting to a variety of environment.In this paper,the given frameworkbases on the adjustable autonomy pinciples,and meets the functional requirements of variable autonomy and the domain requirements of Unmanned Naval System.Then the relationship between the constituent modules is illustrated.Finally,the future of the key issues to be resolved are described.

Conventionally,control methods on rotary wing flyingrobot need to establish its body dynamics model,and then design the controller.However,due tothe strong nonlinearity characteristics of rotary wing flying robot system,itis always difficult to get the exact model which increases the hardness of controller design.This research focuses on the process of people's tele-operation behavior,analyze the manipulating process of skilled and unskilledmanipulator,and establishes controller respectively that can scale the proficiency of the manipulators.Analysis and comparison of the two controllers were carried out,then a single-channel PD control system for rotary wing flying robot with skilled manipulator characteristics was established.Simulation results show that the PD controller can realize the posture stability control in three kinds of state of rotary wing flying robot including hovering,forward flight,and lateral flight.

This paper designs and implements an information value-based mobile subscription system for stock information,which specifics to the problems existing in the current stock information services,such as poor real-timeliness,insufficient accuracy and lack of personalization.The system,which uses vertical search to obtain information based on the open source tools:Lucene and Weblech,builds personalized stock ontology based on economic knowledge,and generates personalized RSS seeds through query expansion and ontology matching.The informationis ranked by value,and the system also utilizes KXML to parse newsfeed to implement information subscription at the mobile terminal.The experimental results show that the stock information service provided by the system is effective.