Short-term prediction of chaotic time series by using RBF network with regression weights

We propose a framework for constructing and training a radial basis function (RBF) neural network. The structure of the gaussian functions is modified using a pseudo-gaussian function (PG) in which two scaling parameters sigma are introduced, which eliminates the symmetry restriction and provides the neurons in the hidden layer with greater flexibility with respect to function approximation. We propose a modified PG-BF (pseudo-gaussian basis function) network in which the regression weights are used to replace the constant weights in the output layer. For this purpose, a sequential learning algorithm is presented to adapt the structure of the network, in which it is possible to create a new hidden unit and also to detect and remove inactive units. A salient feature of the network systems is that the method used for calculating the overall output is the weighted average of the output associated with each receptive field. The superior performance of the proposed PG-BF system over the standard RBF are illustrated using the problem of short-term prediction of chaotic time series.     

Feedforward neural network with adaptive reference pattern layer.

A hybrid neural network architecture is investigated for modeling purposes. The proposed hybrid is based on the multilayer perceptron (MLP) network. In addition to the usual hidden layers, the first hidden layer is selected to be an adaptive reference pattern layer. Each unit in this new layer incorporates a reference pattern that is located somewhere in the space spanned by the input variables. The outputs of these units are the component wise-squared differences between the elements of a reference pattern and the inputs. The reference pattern layer has some resemblance to the hidden layer of the radial basis function (RBF) networks. Therefore the proposed design can be regarded as a sort of hybrid of MLP and RBF networks. The presented benchmark experiments show that the proposed hybrid can provide significant advantages over standard MLPs and RBFs in terms of fast and efficient learning, and compact network structure.     

Estimation of Spatiotemporal Neural Activity Using Radial Basis Function Networks

We report a method using radial basis function (RBF) networks to estimate the time evolution of population activity in topologically organized neural structures from single-neuron recordings. This is an important problem in neuroscience research, as such estimates may provide insights into systems-level function of these structures. Since single-unit neural data tends to be unevenly sampled and highly variable under similar behavioral conditions, obtaining such estimates is a difficult task. In particular, a class of cells in the superior colliculus called buildup neurons can have very narrow regions of saccade vectors for which they discharge at high rates but very large surround regions over which they discharge at low, but not zero, levels. Estimating the dynamic movement fields for these cells for two spatial dimensions at closely spaced timed intervals is a difficult problem, and no general method has been described that can be applied to all buildup cells. Estimation of individual collicular cells' spatiotemporal movement fields is a prerequisite for obtaining reliable two-dimensional estimates of the population activity on the collicular motor map during saccades. Therefore, we have developed several computational-geometry-based algorithms that regularize the data before computing a surface estimation using RBF networks. The method is then expanded to the problem of estimating simultaneous spatiotemporal activity occurring across the superior colliculus during a single movement (the inverse problem). In principle, this methodology could be applied to any neural structure with a regular, two-dimensional organization, provided a sufficient spatial distribution of sampled neurons is available.     

海洋浮游植物丰度的空间插值优化

在探索性空间数据分析(Exploratory spatial data analysis)和数据转化的基础上,利用反距离加权(Inverse distance weighting,IDW)、径向基函数(Radial basis functions,RBF)、普通克里格(Ordinary Kriging,OK),3种插值方法,对2003年8月获得的珠江口浮游植物丰度数据进行插值运算,并对插值准确度进行交叉验证。结果显示,珠江口浮游植物丰度数据具有离散性大、存在极大和极小值、呈正偏分布等特点。而对数转化能大大减小数据的离散性和不对称性,有效消除插值结果图中各类插值噪音。交叉验证显示,插值精确度OK最高,RBF次之,IDW最低。观察插值结果等值面图,发现3种方法均能较客观地模拟出浮游植物丰度的总体分布趋势,在对局部趋势的模拟上,OK的表现最好。综合评定,OK为最适合珠江口浮游植物丰度数据的插值方法。半变异模型的选择对OK的插值结果影响不明显。在四种半变异模型中,圆形模型的拟合效果最好。     

Route searching based on neural networks and heuristic reinforcement learning

In this paper, an improved and much stronger RNH-QL method based on RBF network and heuristic Q-learning was put forward for route searching in a larger state space. Firstly, it solves the problem of inefficiency of reinforcement learning if a given problem’s state space is increased and there is a lack of prior information on the environment. Secondly, RBF network as weight updating rule, reward shaping can give an additional feedback to the agent in some intermediate states, which will help to guide the agent towards the goal state in a more controlled fashion. Meanwhile, with the process of Q-learning, it is accessible to the underlying dynamic knowledge, instead of the need of background knowledge of an upper level RBF network. Thirdly, it improves the learning efficiency by incorporating the greedy exploitation strategy to train the neural network, which has been testified by the experimental results.     

Target and temporal pattern selection at neocortical synapses

We attempt to summarize the properties of cortical synaptic connections and the precision with which they select their targets in the context of information processing in cortical circuits. High-frequency presynaptic bursts result in rapidly depressing responses at most inputs onto spiny cells and onto some interneurons. These 'phasic' connections detect novelty and changes in the firing rate, but report frequency of maintained activity poorly. By contrast, facilitating inputs to interneurons that target dendrites produce little or no response at low frequencies, but a facilitating-augmenting response to maintained firing. The neurons activated, the cells they in turn target and the properties of those synapses determine which parts of the circuit are recruited and in what temporal pattern. Inhibitory interneurons provide both temporal and spatial tuning. The 'forward' flow from layer-4 excitatory neurons to layer 3 and from 3 to 5 activates predominantly pyramids. 'Back' projections, from 3 to 4 and 5 to 3, do not activate excitatory cells, but target interneurons. Despite, therefore, an increasing complexity in the information integrated as it is processed through these layers, there is little 'contamination' by 'back' projections. That layer 6 acts both as a primary input layer feeding excitation 'forward' to excitatory cells in other layers and as a higher-order layer with more integrated response properties feeding inhibition to layer 4 is discussed.     

Plant physiological ecology of tropical forest canopies

Mechanistic information about tropical canopy function is emerging at the leaf, tree, stand and landscape levels. With improved canopy access, comprehensive data are accumulating about seasonal and spatial variation in light, temperature and humidity, and corresponding variation in leaf carbon gain and water loss. At the whole-plant level, simultaneous measurements at different spatial scales have revealed the role of boundary layer dynamics in regulating transpiration. Emergent properties of canopy function are being explored through models that integrate leaf and landscape-level exchange processes. Integration of exchange processes that include functional diversity at different scales has the potential to validate regional estimates of gas exchange, which are critical to our understanding of the role of tropical forests in global atmospheric carbon balance.     

典型亚热带森林生态系统碳密度及储量空间变异特征

以浙江省森林生态系统为研究对象,基于GIS网格布点,采集了838个森林样地样本(土壤、枯落物等),结合浙江省森林资源监测中心相关数据,利用地统计学和Moran's I相结合的方法系统研究了浙江省森林生态系统碳密度及碳储量空间变异特征。结果表明:浙江省森林生态系统平均碳密度为145.22 t/hm~2,其中森林植被、土壤、枯落物和枯死木层碳密度分别为27.34、108.89、1.79、1.38 t/hm~2。克里格空间插值和局部Moran's I指数结果表明碳密度空间分布规律呈现从西南向东北方向逐渐递减的趋势,与浙江省地形、地势较为一致,受海拔、树龄、森林类型、台风气候等自然因素和人类活动共同影响。浙江省森林生态系统碳储量为877.19 Tg C,森林植被、土壤、枯落物和枯死木层碳储量分别为203.88、656.20、10.84、6.27 Tg C,分别占总碳储量的23%、75%、1.3%、0.7%。在浙江省森林生态系统碳储量空间分布格局中,土壤层是森林生态系统中最大的碳库,约是森林植被层的3.22倍,是整个浙江省森林生态系统碳储量最主要的贡献者。浙江省森林资源丰富,大多数森林仍处于中幼龄林阶段,碳密度水平较低,但是中幼龄林生长速度较快,加强对全省中幼龄林的健康管理,是未来整体提升浙江省森林生态系统固碳潜力的关键。     

基于辅助环境变量的土壤有机碳空间插值——以黄土丘陵区小流域为例

利用普通克里格法(OK)、反距离加权法(IDW)、径向基函数法(RBF)、基于土地利用类型修正的普通克里格法(OK_LU)4种插值方法,对黄土丘陵羊圈沟小流域的土壤有机碳含量进行空间插值。预测结果的准确性通过Pearson相关系数(R),平均绝对误差(MAE),均方根误差(RMSE),准确度(AC)来评价。研究结果表明:(1)在前3种常规空间插值方法中,OK对刻画区域土壤有机碳的空间分布趋势效果最佳,其预测MAE值和RMSE值均为最小,Pearson相关系数(R)和准确度(AC)最大,说明其预测结果的准确性最好、预测的极端误差也最小;其次为RBF;IDW预测的效果最差。(2)OK_LU在空间特征表达方面能够更好地反映复杂地形区的局部变异,其插值结果的精度相比OK有一定程度的提高,其平均绝对误差(MAE)从0.900%降到了0.567%,均方根误差(RMSE)从1.101%降到了0.777%,Pearson相关系数(R)从0.4026提高到0.5589,准确度(AC)从0.9081提高到0.9505。综合比较,在黄土丘陵地区,OK_LU能使插值结果的精度有较大提高,是土壤有机碳空间制图的有效途径。     

戴云山黄山松群落生态（火用）空间异质性

热力学指标生态（火用）作为传统群落结构指数的补充,通过衡量物种包含的遗传信息量化群落生物热力学空间结构,评估其群落稳定性和健康程度。基于生态（火用）理论研究了戴云山黄山松（Pinus taiwanensis）群落乔木层、灌木层生态（火用）和结构（火用）的空间异质性及分布规律。主要结果如下：（1）乔木层和灌木层分别是群落生态（火用）和结构（火用）的主要贡献者。（2）黄山松种群在群落中具有更高的竞争能力和生存概率,群落生态（火用）贡献率为98.39%,对维持群落结构稳定性具有关键作用。（3）乔灌层的生态（火用）和结构（火用）在研究尺度上均具有强烈空间自相关,其空间异质性主要由结构性因素引起,所占比例大于75%。（4）乔灌层的生态（火用）和结构（火用）均呈明显条带状及斑块状分布,有多个明显高值区,等值线较弯曲且密集,空间异质性程度较高。研究确定了戴云山黄山松种群的热力学地位,黄山松在维系戴云山群落结构稳定和群落健康中发挥着重要作用,结果可为黄山松群落演替动态和可持续管理提供科学依据。     

宁夏东部荒漠草原灌丛引入下土壤水分空间异质性

结合地理信息系统(GIS),运用经典统计学和地统计学方法对宁夏东部荒漠草原人工灌丛引入下0～200 cm土层土壤水分空间异质性进行研究.结果表明: 0～200 cm土层土壤含水量为0.6%～19.0%,均值为4.4%,土壤水分含量较低,变异系数为49.5%～86.3%,属于中等变异.不同土层半变异函数的最佳理论模型分别为:0～60、80～120 cm各层土壤水分符合球状模型,60～80 cm符合指数模型,120～200 cm符合高斯模型.不同土层土壤含水量均呈不同程度的空间相关,0～40、60～80、120～200 cm各层土壤水分的块金系数C₀/(C₀+C)为26.1%～49.9%,为中等程度空间相关;40～60、80～100、100～120 cm各层土壤含水量的块金系数为15.5%～22.1%,具强烈空间相关.0～200 cm不同土层土壤水分含量变程不同,0～20 和 20～40 cm土层变程较大,为37.10～45.18 km,40～200 cm各层土壤含水量的变程较小,为3.58～8.66 km.荒漠草原人工灌丛引入过程中加速土壤水分利用和深层水分消耗,导致土壤水分的空间异质性和破碎化程度加强,且对深层次土壤水分作用更强.     

Mutating RBF Can Enhance Its Pro-Apoptotic Activity and Uncovers a New Role in Tissue Homeostasis

The tumor suppressor retinoblastoma protein (pRb) is inactivated in a wide variety of cancers. While its role during cell cycle is well characterized, little is known about its properties on apoptosis regulation and apoptosis-induced cell responses. pRb shorter forms that can modulate pRB apoptotic properties, resulting from cleavages at caspase specific sites are observed in several cellular contexts. A bioinformatics analysis showed that a putative caspase cleavage site (TELD) is found in the Drosophila homologue of pRb (RBF) at a position similar to the site generating the p76Rb form in mammals. Thus, we generated a punctual mutant form of RBF in which the aspartate of the TELD site is replaced by an alanine. This mutant form, RBF^D253A, conserved the JNK-dependent pro-apoptotic properties of RBF but gained the ability of inducing overgrowth phenotypes in adult wings. We show that this overgrowth is a consequence of an abnormal proliferation in wing imaginal discs, which depends on the JNK pathway activation but not on wingless (wg) ectopic expression. These results show for the first time that the TELD site of RBF could be important to control the function of RBF in tissue homeostasis in vivo.     

Development of retinal layers

A noticeable characteristic of nervous systems is the arrangement of synapses into distinct layers. Such laminae are fundamental for the spatial organisation of synaptic connections transmitting different kinds of information. A major example of this is the inner plexiform layer (IPL) of the vertebrate retina, which is subdivided into at least ten sublayers. Another noticeable characteristic of these retina layers is that neurons are displayed in the horizontal plane in a non-random array termed as mosaic patterning. Recent studies of vertebrate and invertebrate systems have identified molecules that mediate these interactions. Here, we review the last mechanisms and molecules mediating retinal layering.     

p55, the Drosophila ortholog of RbAp46/RbAp48, is required for the repression of dE2F2/RBF-regulated genes

Many proteins have been proposed to be involved in retinoblastoma protein (pRB)-mediated repression, but it is largely uncertain which cofactors are essential for pRB to repress endogenous E2F-regulated promoters. Here we have taken advantage of the stream-lined Drosophila dE2F/RBF pathway, which has only two E2Fs (dE2F1 and dE2F2), and two pRB family members (RBF1 and RBF2). With RNA interference (RNAi), we depleted potential corepressors and looked for the elevated expression of groups of E2F target genes that are known to be directly regulated by RBF1 and RBF2. Previous studies have implicated histone deacetylase (HDAC) and SWI/SNF chromatin-modifying complexes in pRB-mediated repression. However, our results fail to support the idea that the SWI/SNF proteins are required for RBF-mediated repression and suggest that a requirement for HDAC activities is likely to be limited to a subset of targets. We found that the chromatin assembly factor p55/dCAF-1 is essential for the repression of dE2F2-regulated targets. The removal of p55 deregulated the expression of E2F targets that are normally repressed by dE2F2/RBF1 and dE2F2/RBF2 complexes in a cell cycle-independent manner but had no effect on the expression of E2F targets that are normally coupled with cell proliferation. The results indicate that the mechanisms of RBF regulation at these two types of E2F targets are different and suggest that p55, and perhaps p55's mammalian orthologs RbAp46 and RbAp48, have a conserved function in repression by pRB-related proteins.     

白鹤中途停歇地主要食物藨草球茎密度的空间插值方法研究

扁杆藨草(Scirpus planiculmis)和三江藨草(S.nipponicus)的球茎是迁徙白鹤(Grus leucogeranus)在莫莫格中途停歇地的主要食物。2010年9月15～25日,对莫莫格白鹤集中停歇地鹅头泡湿地的20条样线上100 m等间隔470个取样点进行了地下0～20 cm的藨草球茎的取样工作。基于此,采用反距离权重、4种径向基函数和普通克里格共6种空间插值方法模拟了该区域藨草球茎密度。交叉验证和插值结果的空间分布图对比,径向基函数中的张力样条插值效果最佳,是最适于该区域挺水植物藨草球茎密度的空间插值方法。张力样条插值结果表明:研究区内藨草球茎密度均值为(46.0±44.3)个/m2。其中,<25个/m2的面积占研究区的41.7%,25～100个/m2的面积占45.9%,而>100个/m2的面积仅占12.4%。研究结果为下一步分析白鹤迁徙种群与环境因子的生态关系、预测该地区白鹤的环境容纳量提供十分重要的基础数据。     

基于Voronoi图的林分空间模型及分布格局研究

以南洞庭湖龙虎山次生林为研究对象,建立林分Voronoi图模型,通过该模型表示林分空间结构特征,提取林木空间量化信息。在此基础上,引入变异系数量化Voronoi多边形面积的变化率,将林木空间格局分析转化为计算Voronoi图变异系数,并将计算结果与角尺度作相关性比较。结果表明:林分Voronoi图模型能准确表达林木邻接关系、相邻木距离及角度、空间密度等重要空间信息;大尺度取样时(76m以上),变异系数趋于稳定,林分整体空间格局呈接近随机分布的轻度聚集;Voronoi图变异系数和角尺度存在显著的正相关关系,进一步说明本研究格局分析方法的可行性和有效性;受多种因素影响,乔木层主要树种在不同发育阶段的空间格局表现出强烈的时间和空间动态特性,且小径木多为聚集分布,大径木呈轻度聚集或随机分布。基于Voronoi图的林分空间模型及格局分析方法,从空间上表征林木的生长繁育特性、种内种间关系、资源及环境分布特征,为计算几何方法应用于林分空间经营提供了新的思路。     

Constructing a query-able radial basis function artificial neural network

Artificial neural networks will be more widely accepted as standard engineering tools if their reasoning process can be made less opaque. This paper describes NetQuery, an explanation mechanism that extracts meaningful explanations from trained Radial Basis Function (RBF) networks. RBF networks are well suited for explanation generation because they contain a set of locally tuned units. Standard RBF networks are modified to identify dependencies between the inputs, to be sparsely connected, and to have an easily interpretable output layer. Given these modifications, the network architecture can be used to extract "Why?" and "Why not?" explanations from the network in terms of excitatory and inhibitory in-puts and their linear relationships, greatly simplified by a run-time pruning algorithm. These query results are validated by creating an expert system based on the explanations. NetQuery is also able to inform a user about a possible change in category for a given pattern by responding to a "How can I...?" query. This kind of query is extremely useful when analyzing the quality of a pattern set.     

一种感知视觉运动信息的简化脑模型

视觉运动信息的感知过程,包括从局域运动检测到对模式整体运动的感知过程.我们以蝇视觉系统的图形-背景相对运动分辨的神经回路网络为基本框架,采用初级运动检测器的六角形阵列作为输入层,构造了一种感知视觉运动信息的简化脑模型,模拟了运动信息应该神经计算模型各个层次上的处理.该模型对差分行为实验结果作出了正确预测.本文并对空间生理整合的神经机制作了讨论.