科学网

 找回密码
  注册

tag 标签: process

相关帖子

版块 作者 回复/查看 最后发表

没有相关内容

相关日志

[转载]Science Blog 2012年05月04日 20:16 (星期五)
xupeiyang 2012-5-4 21:19
http://scienceblog.com/ Effects of loneliness mimic aging process, boost heart disease risk Study exposes ‘arms race’ between male and female mating traits Why underweight babies become obese: Hypothalamus is to blame Awake mental replay of past experiences critical for learning Email ‘vacations’ decrease stress, increase concentration How much money parents give to college-age kids North Americans lived with extinct giant beasts Genes may explain why some people turn their noses up at meat Too many years in school turns off science Ph.D. interest in faculty jobs A new understanding of Alzheimer’s trigger Coelacanth find rewrites history of the ancient fish Plant study points to dangers of a warming world Black Hole Caught Red-Handed in a Stellar Homicide Prenatal choline may ‘program’ healthier babies Regular jogging shows dramatic increase in life expectancy
个人分类: 科学博客|1788 次阅读|0 个评论
carbon and ZnO
hubin 2012-4-6 10:46
Accurate Control of Multishelled ZnO Hollow Microspheres for Dye-Sensitized Solar Cells with High Efficiency All reagents were analytical grade and purchased from Beijing Chemical Co. Ltd., and used without further purification. Hydrated zinc nitrates Zn(NO3)2·6H2O were used as metal precursors. Taking triple-shelled ZnO hollow microspheres with close double shells in the interior as an example, the typical synthesis process is described as follows. Carbonaceous microspheres were synthesized through the emulsion polymerization reaction of sugar under hydrothermal conditions as described elsewhere. Briefly, newly prepared carbonaceous microspheres (0.6 g) were dispersed in zinc nitrate solution (30 mL, 5 M) with the aid of ultrasonication. After ultrasonic dispersion for 15 min, the resulting suspension was aged for 6 h at room temperature, filtered, washed, and dried at 80 ℃ for 12 h. The resultant composite microspheres were heated to 500 ℃ in air at the rate of 1 ℃ min−1, with holding of the temperature at 400 ℃ for 30 min. Triple-shelled ZnO hollow microspheres were subsequently formed as a white-powder product. Aerographite: Ultra Lightweight, Flexible Nanowall, Carbon Microtube Material with Outstanding Mechanical Performance CVD-Synthesis : Basic Aerographite confi guration (hollow, with closed graphitic shells) can be gained by placing ZnO templates in the maximum temperature zone of a two zone split tube furnace (quartz tube; l = 1300 mm; d = 110 mm): At a constant temperature profile of 200 ℃ in injection zone and 760 ℃ in main zone under Ar gas flow (0.02 L min −1 , atmosphere pressure), injection of toluene (99% Alfa Aesar) at 5.5 mL h −1 is started. At start of the injection by a syringe pump, gas fl ow rates are changed to 0.2 L min −1 Ar/0.02 L min −1 H 2 for 120 min. A subsequent 45 min pure H 2 (0.6 L min −1 ) gas flow without injection is followed by a 120 min mixed atmosphere Ar/ H 2 (0.2 L min −1 / 0.02 L min −1 ) with injection of toluene. A final pure hydrogen treatment with 0.6 L min −1 was conducted for 20 min until cooling down under pure Ar purge flow of 0.6 L min −1 . Synthesis depends on template surface area and time depended variations of gas flow rates or temperatures. For example, the parameters for the ultra lightweight, hollow-framework variant are given by a decreased 2 mL h −1 toluene injection for 4 h with a 0.06 L min −1 H 2 and 0.2 mL min −1 Ar gas flow at 760 ℃ and a 1 h post-treatment with no injection and pure H 2 flow of 0.09 L min −1 at 800 ℃ . The formation of the Aerographite can be revealed by stopping the synthesis in an intermediate state. The growth of carbon nano layers on the outside and the simultaneous removal of the inner ZnO template are shown in figure 2 e-h (Figures S7-S10). Simultaneously with the carbon deposition a controlled hydrogen gas flow enables a continuous reduction of ZnO to Zn and thus the template removal. The metallic Zn is transported downstream through the CVD exhaust system where it precipitates as metallic thin film on cold areas. This hydrogen etching exhibits a clear preference on crystallographic orientations of the ZnO crystals. Interestingly, decomposition leads to an axial and radial vanishing of template sections. Massive blocks of remaining ZnO are held in place only by the partly grown carbon nano layers, Figure 2f-h (Figures S7-S9). Furthermore, the atomic structure of these layers can be controlled by the CVD growth parameters reaching from a substantially graphitic to a predominantly amorphous to glassy state . This is in contrast to other growth processes for carbon materials which form either highly ordered graphene and graphite like materials or pyrolytic low ordered structures , therefore Aerographite might be also described as ‘pyrolytic graphite’. Besides EELS spectra, also energy filtered TEM (EFTEM) elemental mapping ensures us of the absence of oxygen and confirms the sp²-hybridization state for both arrangements. An amorphous carbon state occurs at higher template decomposition rates, e.g., as induced by higher temperatures and higher hydrogen concentration. As our model proposes, initial deposition of small carbon nucleation belts on template surfaces and in-plane biaxial growth of carbon leads to the variety of possible sp 2 -hybridized layers: For fully enclosing, smooth graphitic layers template etching and growth has to be in equilibrium. A slightly faster template removal by changing CVD-parameters leads to a higher ratio of amorphous to graphitic carbon. Further: A quicker removal of surfaces before enclosing shells can develop (high hydrogen concentration and increased temperatures of 900 ℃ creates amorphous carbon ribbons, which assemble in a hierarchical hollow framework of high mechanical strength (Figure 1e-h, S16).
个人分类: 科研笔记|2697 次阅读|0 个评论
Research on the intelligent process planning technology base
bshen 2012-3-30 10:09
沈斌、欧阳华兵. Research on the intelligent process planning technology based on STEP-NC. 2011 3 rd International Conference on Computer and Automation Engineering, Chongqing, China, January 21-23,2011:v2-221~225, EI : 20120314682977
个人分类: 学术论文|2214 次阅读|0 个评论
tension deformation texture deformation (2007)
zsma81 2012-3-8 21:55
Evolution of deformation-induced texture and surface microstructure of nickel coating under deep cup-drawing process S.G. Long, Z.S. Ma , X.B. Zhang, Y. Pan, Y.C. Zhou The deformation texture and surface microstructure of nickel coating induced by deep cup-drawing were studied by X-ray diffractometry and SEM. The steel sheet with nickel coating was firstly punched to designed degree of cup shapes. Then the texture of the coating was determined by XRD and the surface microstructure was observed by SEM. The results indicate that, the nickel coating material includes three kinds of textures, i.e. Ni (111), Ni (200) and Ni (220). After cup-drawing deformation, there doesn’t appear new texture component for nickel coatings but the change of the intensity of the deformation textures. In the cup-deformation process, the two kinds of main textures Ni (111) and Ni (200) increase in the first and second cup-drawing procedures and then reduce quickly in the third and fourth drawing procedure. However, for the Ni (220) texture, it always increases in whole cup-drawing procedures. With the increase in the drawing, there are some cracks to be found, but not delamination. This shows that nickel coating and the substrate have a good combinable performance. Transactions of Nonferrous Metals Society of China , 17, (2007) s823–s826. Evolution of deformation-induced texture and surface microstructure of nickel co.pdf
个人分类: 论文|2317 次阅读|0 个评论
Author Disambiguation: A Nonparametric Topic and Co-authorsh
chengh3 2011-11-8 17:33
Author Disambiguation: A Nonparametric Topic andCo-authorship Model Andrew M. Dai,University of Edinburgh In Proceedings of the NIPS Workshop on Applications for Topic Models: Text and Beyond , Whistler, Canada, 2009. 将基于层次Dirichlet过程的主题模型与合作关系结合起来,提出一个产生式模型。该模型不依赖于作者的个数。
个人分类: 重名判别|3321 次阅读|0 个评论
[转载]拟合和回归的区别
ChenboBlog 2011-11-1 10:47
Curve fitting is the process of constructing a curve, or mathematical function , that has the best fit to a series of data points, possibly subject to constraints. Curve fitting can involve either interpolation , where an exact fit to the data is required, or smoothing , in which a "smooth" function is constructed that approximately fits the data. A related topic is regression analysis , which focuses more on questions of statistical inference such as how much uncertainty is present in a curve that is fit to data observed with random errors. Fitted curves can be used as an aid for data visualization, to infer values of a function where no data are available, and to summarize the relationships among two or more variables. Extrapolation refers to the use of a fitted curve beyond the range of the observed data, and is subject to a greater degree of uncertainty since it may reflect the method used to construct the curve as much as it reflects the observed data.
2584 次阅读|0 个评论
[转载]Drug design, also sometimes referred to as rational drug des
cai7net 2011-10-31 15:39
Drug design , also sometimes referred to as rational drug design or structure based drug design, is the inventive process of finding new medications based on the knowledge of the biological target . The drug is most commonly an organic small molecule which activates or inhibits the function of a biomolecule such as a protein which in turn results in a therapeutic benefit to the patient . In the most basic sense, drug design involves design of small molecules that are complementary in shape and charge to the biomolecular target to which they interact and therefore will bind to it. Drug design frequently but not necessarily relies on computer modeling techniques. This type of modeling is often referred to as computer-aided drug design . The phrase "drug design" is to some extent a misnomer . What is really meant by drug design is ligand design. Modeling techniques for prediction of binding affinity are reasonably successful. However there are many other properties such as bioavailability , metabolic half-life , lack of side effects , etc. that first must be optimized before a ligand can become a safe and efficacious drug. These other characteristics are often difficult to optimize using rational drug design techniques.
个人分类: 学习心得|0 个评论
[转载]图(graph) 谱(spectrum) 马尔可夫过程(markov process) 聚类结构
热度 3 justinzhao 2011-10-18 04:44
题目中所说到的四个词语,都是Machine Learning以及相关领域中热门的研究课题。表面看属于不同的topic,实际上则是看待同一个问题的不同角度。不少文章论述了它们之间的一些联系,让大家看到了这个世界的奇妙。 从图说起 这里面,最简单的一个概念就是“图”(Graph),它用于表示事物之间的相互联系。每个图有一批节点(Node),每个节点表示一个对象,通过一些边 (Edge)把这些点连在一起,表示它们之间的关系。就这么一个简单的概念,它对学术发展的意义可以说是无可估量的。几乎所有领域研究的东西,都是存在相互联系的,通过图,这些联系都具有了一个统一,灵活,而又强大的数学抽象。因此,很多领域的学者都对图有着深入探讨,而且某个领域关于图的研究成果,可以被其它领域借鉴。 矩阵表示:让代数进入图的世界 在数学上,一种被普遍使用的表达就是邻接矩阵(Adjacency Matrix)。一个有N个节点的图,可以用一个N x N的矩阵G表示,G(i, j)用一个值表示第i个节点和第j个节点的联系,通常来说这个值越大它们关系越密切,这个值为0表示它们不存在直接联系。这个表达,很直接,但是非常重要,因为它把数学上两个非常根本的概念联系在一起:“图”(Graph)和“矩阵”(Matrix)。矩阵是代数学中最重要的概念,给了图一个矩阵表达,就建立了用代数方法研究图的途径。数学家们几十年前开始就看到了这一点,并且开创了数学上一个重要的分支——代数图论(Algebraic Graph Theory)。 代数图论通过图的矩阵表达来研究图。熟悉线性代数的朋友知道,代数中一个很重要的概念叫做“谱”(Spectrum)。一个矩阵的很多特性和它的谱结构 ——就是它的特征值和特征向量是密切相关的。因此,当我们获得一个图的矩阵表达之后,就可以通过研究这个矩阵的谱结构来研究图的特性。通常,我们会分析一个图的邻接矩阵(Adjacency Matrix)或者拉普拉斯矩阵(Laplace Matrix)的谱——这里多说一句,这两种矩阵的谱结构刚好是对称的。 谱:“分而治之”的代数 谱,这个词汇似乎在不少地方出现过,比如我们可能更多听说的频谱,光谱,等等。究竟什么叫“谱”呢?它的概念其实并不神秘,简单地说,谱这个概念来自“分而治之”的策略。一个复杂的东西不好直接研究,就把它分解成简单的分量。如果我们把一个东西看成是一些分量叠加而成,那么这些分量以及它们各自所占的比例,就叫这个东西的谱。所谓频谱,就是把一个信号分解成多个频率单一的分量。 矩阵的谱,就是它的特征值和特征向量,普通的线性代数课本会告诉你定义:如果A v = c v,那么c 就是A的特征值,v就叫特征向量。这仅仅是数学家发明的一种数学游戏么?——也许有些人刚学这个的时候,并一定能深入理解这么个公式代表什么。其实,这里的谱,还是代表了一种分量结构,它为使用“分而治之”策略来研究矩阵的作用打开了一个重要途径。这里我们可以把矩阵理解为一个操作(operator),它的作用就是把一个向量变成另外一个向量:y = A x。对于某些向量,矩阵对它的作用很简单,A v = cv,相当于就把这个向量v 拉长了c倍。我们把这种和矩阵A能如此密切配合的向量v1, v2, ... 叫做特征向量,这个倍数c1, c2, ...叫特征值。那么来了一个新的向量x 的时候,我们就可以把x 分解为这些向量的组合,x = a1 v1 + a2 v2 + ...,那么A对x的作用就可以分解了:A x = A (a1 v1 + a2 v2 + ...) = a1 c1 v1 + a2 c2 v2 ... 所以,矩阵的谱就是用于分解一个矩阵的作用的。 这里再稍微延伸一点。一个向量可以看成一个关于整数的函数,就是输入i,它返回v( i )。它可以延伸为一个连续函数(一个长度无限不可数的向量,呵呵),相应的矩阵 A 变成一个二元连续函数(面积无限大的矩阵)。这时候矩阵乘法中的求和变成了积分。同样的,A的作用可以理解为把一个连续函数映射为另外一个连续函数,这时候A不叫矩阵,通常被称为算子。对于算子,上面的谱分析方法同样适用(从有限到无限,在数学上还需要处理一下,不多说了)——这个就是泛函分析中的一个重要部分——谱论(Spectral Theory)。 马尔可夫过程——从时间的角度理解图 回到“图”这个题目,那么图的谱是干什么的呢?按照上面的理解,似乎是拿来分解一个图的。这里谱的作用还是分治,但是,不是直观的理解为把图的大卸八块,而是把要把在图上运行的过程分解成简单的过程的叠加。如果一个图上每个节点都有一个值,那么在图上运行的过程就是对这些值进行更新的过程。一个简单,大家经常使用的过程,就是马尔可夫过程(Markov Process)。 学过随机过程的朋友都了解马尔可夫过程。概念很简单——“将来只由现在决定,和过去无关”。考虑一个图,图上每个点有一个值,会被不断更新。每个点通过一些边连接到其它一些点上,对于每个点,这些边的值都是正的,和为1。在图上每次更新一个点的值,就是对和它相连接的点的值加权平均。如果图是联通并且非周期(数学上叫各态历经性, ergodicity),那么这个过程最后会收敛到一个唯一稳定的状态(平衡状态)。 图上的马尔可夫更新过程,对于很多学科有着非常重要的意义。这种数学抽象,可以用在什么地方呢?(1) Google对搜索结果的评估(PageRank)原理上依赖于这个核心过程,(2) 统计中一种广泛运用的采样过程MCMC,其核心就是上述的转移过程,(3) 物理上广泛存在的扩散过程(比如热扩散,流体扩散)和上面的过程有很重要的类比,(4) 网络中的信息的某些归纳与交换过程和上述过程相同 (比如Random Gossiping),还有很多。非常多的实际过程通过某种程度的简化和近似,都可以归结为上述过程。因此,对上面这个核心过程的研究,对于很多现象的理解有重要的意义。各个领域的科学家从本领域的角度出发研究这个过程,得出了很多实质上一致的结论,并且很多都落在了图的谱结构的这个关键点上。 图和谱在此联姻 根据上面的定义,我们看到邻接矩阵A其实就是这个马尔可夫过程的转移概率矩阵。我们把各个节点的值放在一起可以得到一个向量v,那么我们就可以获得对这个过程的代数表示, v(t+1) = A v(t)。稳定的时候,v = A v。我们可以看到稳定状态就是A的一个特征向量,特征值就是1。这里谱的概念进来了。我们把A的特征向量都列出来v1, v2, ...,它们有 A vi = ci vi。vi其实就是一种很特殊,但是很简单的状态,对它每进行一轮更新,所有节点的值就变成原来的ci倍。如果0 ci 1,那么,相当于所有节点的值呈现指数衰减,直到大家都趋近于0。 一般情况下,我们开始于一个任意一个状态u,它的更新过程就没那么简单了。我们用谱的方法来分析,把u分解成 u = v1 + c2 v2 + c3 v3 + ... (在数学上可以严格证明,对于上述的转移概率矩阵,最大的特征值就是1,这里对应于平衡状态v1,其它的特征状态v2, v3, ..., 对应于特征值1 c2 c3 ... -1)。那么,我们可以看到,当更新进行了t 步之后,状态变成 u(t) = v1 + c2^t v2 + c3^t v3 + ...,我们看到,除了代表平衡状态的分量保持不变外,其它分量随着t 增长而指数衰减,最后,其它整个趋近于平衡状态。 从上面的分析看到,这个过程的收敛速度,其实是和衰减得最慢的那个非平衡分量是密切相关的,它的衰减速度取决于第二大特征值c2,c2的大小越接近于1,收敛越慢,越接近于0,收敛越快。这里,我们看到了谱的意义。第一,它帮助把一个图上运行的马尔可夫过程分解为多个简单的字过程的叠加,这里面包含一个平衡过程和多个指数衰减的非平衡过程。第二,它指出平衡状态是对应于最大特征值1的分量,而收敛速度主要取决于第二大特征值。 我们这里知道了第二大特征值c2对于描述这个过程是个至关重要的量,究竟是越大越好,还是越小越好呢?这要看具体解决的问题。如果你要设计一个采样过程或者更新过程,那么就要追求一个小的c2,它一方面提高过程的效率,另外一方面,使得图的结构改变的时候,能及时收敛,从而保证过程的稳定。而对于网络而言,小的c2有利于信息的迅速扩散和传播。 聚类结构——从空间的角度理解图 c2的大小往往取决于图上的聚类结构。如果图上的点分成几组,各自聚成一团,缺乏组与组之间的联系,那么这种结构是很不利于扩散的。在某些情况下,甚至需要O(exp(N))的时间才能收敛。这也符合我们的直观想象,好比两个大水缸,它们中间的只有一根很细的水管相连,那么就需要好长时间才能达到平衡。有兴趣的朋友可以就这个水缸问题推导一下,这个水缸系统的第二大特征值和水管流量与水缸的容积的比例直接相关,随比例增大而下降。 对于这个现象进行推广,数学上有一个重要的模型叫导率模型(Conductance)。具体的公式不说了,大体思想是,节点集之间的导通量和节点集大小的平均比例和第二大特征值之间存在一个单调的上下界关系。导率描述的是图上的节点连接的空间结合,这个模型把第二特征值c2和图的空间聚集结构联系在一起了。 图上的聚类结构越明显, c2越大;反过来说,c2越大,聚类的结构越明显,(c2 = 1)时,整个图就断裂成非连通的两块或者多块了。从这个意义上说,c2越大,越容易对这个图上的点进行聚类。机器学习中一个重要课题叫做聚类,近十年来,基于代数图论发展出来的一种新的聚类方法,就是利用了第二大特征值对应的谱结构,这种聚类方法叫做谱聚类(Spectral Clustering)。它在Computer Vision里面对应于一种著名的图像分割方法,叫做Normalized Cut。很多工作在使用这种方法。其实这种方法的成功,取决于c2的大小,也就是说取决于我们如何构造出一个利于聚类的图,另外c2的值本身也可以作为衡量聚类质量,或者可聚类性的标志。遗憾的是,在paper里面,使用此方法者众,深入探讨此方法的内在特点者少。 归纳起来 图是表达事物关系和传递扩散过程的重要数学抽象 图的矩阵表达提供了使用代数方法研究图的途径 谱,作为一种重要的代数方法,其意义在于对复杂对象和过程进行分解 图上的马尔可夫更新过程是很多实际过程的一个重要抽象 图的谱结构的重要意义在于通过它对马尔可夫更新过程进行分解分析 图的第一特征值对应于马尔可夫过程的平衡状态,第二特征值刻画了这个过程的收敛速度(采样的效率,扩散和传播速度,网络的稳定程度)。 图的第二特征分量与节点的聚类结构密切相关。可以通过谱结构来分析图的聚类结构。 马尔可夫过程代表了一种时间结构,聚类结构代表了一种空间结构,“谱”把它们联系在一起了,在数学刻画了这种时与空的深刻关系。
个人分类: 读书日记|5379 次阅读|3 个评论
The first two weeks in Sydney
qiyin546 2011-10-17 20:01
The first two weeks in Sydney
Since I have been Sydney two weeks ago, nearly everything is unknown for me, I need to change and broaden myself to adapte the new word, so that to quickly get a ecological inche for me and make sure the upcoming learning fitness, that is might be the “evolution”process. During those days, several training items have ocurred on me. Firstly, I need to follow the new work and rest timetable. We have three hours’s time difference between Sydney and China, that I will start to work whereas all Chinese friends are still in sleep. Three hours is not so much, but forget a long term time rhythm and set up a new one is not so easy, so much Chinese news and things always remind me that it is not time to work and not time to sleep, I always go to sleep very late althouth I am very tired. The second items is the food, not only the food content, but the food habit. People here do not pay so much on lunch but the dinner, that is opposite to China. For lunch, teachers and students do not eat very much. There is a Kitchen in the research building, professors and students could cook their lunch there for free. That is a good way to save time and money, I like it. boys with a hamburg and few yoghourt, girls with some fruit, somebody even have nothing given not being so hungery. They generally do not have specific rest time at noon. By contrast, their dinner time will be very colorful, and eat too much. I do not adapt this absolutely at the beginning, and I always need to eat too much for lunch so that to delay the afternoon hungery time, and will be very tired aftern the dinner time. The third training item relates to the traffic. This is an very important item to broaden living space, and is also the priority to control time. The straight proximity from my temporay rent house to research building was 2.5 km, but I need to walk nearly 4 km. I need to choose the most convenient bus and train line. Bicycle is impossible here, from one aspect, there is no space for bicycle, few person ride bicycle, from another aspect, there is specific for bicycle, you need to wear helmet and specical bicycle dress, otherwise you will be fired. Train here is convenient, but the management system is so different,they have so many different train lines, and each line is consistent with specific platform, you need to know your traveling plan from traffict websit of New South Wales (131500.com). From this website, you could control your travelling time presise to minutes, I have be good at this now. Up to now, the daily training has been finished, I have adjust myself to the new world now. I could control my daily work and life time, and know where I should go if I what to buy something, I also know how to prepare the lunch food so that to join in the students at noon. Next step for me is science traning, keep going. Lunch time. The girl is laura, a volunteer from germany. The boy is one Ph.D candidate.
个人分类: 感悟生活|1716 次阅读|0 个评论
为行为行
热度 1 liuyangbnu 2011-10-16 12:30
我从来都没有正经研究过行为,确实是个行为学的爱好者,门外汉。 我对动物行为学的粗浅认识还停留在本科和研究生选修《动物行为学课》。我总是认为行为似乎是一个复杂过程,但我从来都是对描述pattern的兴趣大于process。更愿意像观看动物世界那样欣赏行为学家的工作。 后来跑到了瑞典,我记得我们在大冷的冬天跑道恒温恒湿的地下室里观察蟋蟀的行为,还有那个像圣诞老人的老师,Anders Berglund绘声绘色地讲着他如何设计鱼类行为学实验,我充其量只是好奇而已,并没有觉得我有一天可以做这个。 我的硕士导师Jacob Hoglund成名立万的工作是年轻时对于黑琴鸡求偶场的研究,他后来还把类似的工作扩展到了具有同样具有求偶场行为的鴴鹬和娇鶲类,至今那本厚厚的Lek还是研究婚配制度的经典著作。除了有时听他提起黑琴鸡的求偶行为之外,我学到更多的还是种群遗传。 当然我博士导师Gerald Heckel年轻时候也是做婚配制度的,他的导师当年也能坐上欧洲蝙蝠行为研究的头把交椅。不过我从他那里了解行为的也不多。因为他也转行到种群遗传。 我真正对行为学感兴趣的是跟Michael Taborsky对非洲丽鱼研究的合作,虽然我在这个项目里的角色并不是行为学研究,而是种群遗传,当然我们反过来,我的结果为行为的进化提供了证据。Taborsky是足以进入动物行为学名人堂的人物,在行为学数个领域内都颇有建树。从这时起,我学会用进化学的角度来看待形形色色行为的进化。 我在各种会议上见过的行为学者不少。把我真正领入行为学研究大门的是Tamas Szekley。老爷子凭借着在两性冲突和协作的领域上成为一方大牛。他带着我在格罗宁根算是见了世面。和洛伦兹其名的诺奖得主廷伯根成名的地方,何其神圣。Tamas把他写的 Social Behaviour:Genes,Ecology and Evolution送给我,告诉我21世纪的行为学家不仅应该从实验角度,还应该从比较行为,生理,神经和遗传等多个角度,利用先进的技术研究动物行为变异的机理。于是有了我们的第一次合作,尝试用分子遗传和神经生物学来解释两性冲突的自然变异。 不过还是Michael Griesser让我在行为上开了窍。Michael的博士导师Jan Ekman是我硕士一门课的教师,和Michael的博后导师Ben Hatchwell一样是研究合作繁殖行为一方名宿,和Tamas那一代人同属经典的实验学派。我总是说Michael是追随我来到了伯尔尼。我和他在乌普萨拉,伯尔尼还有谢菲尔德,都有很多千丝万缕的联系,这里面有很多很好玩的故事和渊源。Michael的成名作是北噪鸦的亲代和子代冲突,同样属于动物行为中的热门领域。 就是在伯尔尼进入秋天的日子里,我和Michael跑到瑞士和意大利交界的一个谁也找不到的小山村,没有手机信号,没有网络,在他家的祖居里过了整整一周的清静生活。他工作他的论文,我修改我的,有空了就交换看看,切磋一下。就是这样一段集中精力,思考,讨论和写作的日子,使我对行为学又有了新的认识。我所掌握的那些知识,同样可以从宏进化和系统比较的角度来解释行为表型的进化。Tit-for-tat,也许这一周工作后,Michael可以发表他的第一篇种群遗传学文章,我也能发表第一篇行为学论文。这就是一周闭关的结果。 我想我不会是Jacob他们扫描纪录行为的那一代了。我想我要做的是用现代化的生物学知识和技术去像解释形态,分布模式一样去解释行为变异这种复杂的表型。我也可以慢慢从行为学的观察者变成了分享者。
个人分类: 饭后扯淡|5411 次阅读|1 个评论
new
jiangdm 2011-8-19 22:54
new
个人分类: CHI|1 次阅读|0 个评论
review: Automatic Protocol Conformance Checking
jiangdm 2011-8-7 10:49
《Automatic Protocol Conformance Checking of Recursive and Parallel BPEL Systems》 , Andreas Both and Wolf Zimmermann, Sixth European Conference on Web Services ,2008 Abstract Today model checking of Web Services formulated in BPEL is often reduced by transforming BPEL-processes to Petri nets. These can be model checked using traditional approaches. If recursion is present in the BPEL model this approach hides some possible violations of the wished behaviour. We present an approach which allows the Web Service developer to formulate more properties of the required usage of the Web Service and provide a tool that checks whether these requirements are satisfied in a Web Service based system. We use finite state machines to specify permitted sequences of receivable interactions and call them service protocols. In this paper we will show that it is possible to use BPEL representations and service protocols to check if a sequence of receivable interactions that violates a service protocol can occur. We achieve this result by translating BPEL to Process Algebra Nets (introduced by Mayr ) and applying the approach of Mayr for model checking Process Algebra Nets. Our approach computes counterexamples even for recursive and parallel programs including synchronization. 个人点评: 没看懂,印象是: 用Process Algebra Nets形式化BPEL,先分解再综合 看了SCI index: 4, 都是自引用 Automatic Protocol Conformance Checking of Recursive and Parallel BPEL Systems.pdf beamer_Automatic_Protocol_Conformance_Checking_BPEL.tex beamer_Automatic_Protocol_Conformance_Checking_BPEL.pdf
个人分类: CHI|0 个评论
point process, stochastic geometry 在high-level CV中的应用
热度 1 justinzhao 2011-7-17 10:05
前一篇博文写到了MRF在视觉中的应用,它实际是对问题建立graphical model, 然后用bayesian probabilistics的方法做Inference。在建立图模型前,要首先定义vertex/edge/neighborhood,而后续的estimation和inference是基于建立好的图模型的,所以如果一个问题中,vertex(也可以说object)是变化而不是固定的(此时的图模型变了),MRF变得不再实用。 然而,point process(也称为object process,因为它是面向对象的)能对number不固定、shape可变化的对象建模,所以能很好的解决视觉中的某些inverse problems,如object extraction/texture synthesis。Point process率属于随机过程理论,能把prior geometry information用概率来建模,但它与MRF不同在于,它把随机理论用到estimation中,所以Object的数量、参数是随着求解过程变化的。虽然概率模型和MRF中的概率一样是definite的,可是求解中的stochastics使得points in the distribution 更flexible.(MRF中的points,也就是objects,是固定的)。 point process涉及到测度理论和随机过程,当应用到视觉中时,常用gibbs energy 来描述这个point process。接下来就是求最小能量,常用的方法是RJMCMC外加simulated annealing 策略,为了加快收敛速度,在高温使用jump策略,低温使用diffusion. 目前看到的point process的成功案例是其在图像特征提取中的应用,法国的一个研究所INRIA在这方面做了大量的工作。 下面是一些资料: RJMCMC: Reversible jump Markov chain Monte Carlo computation and Bayesian model determination , PJ Green - Biometrika, 1995 stochastic geometric models: Stochastic geometry models in high - level vision, AJ Baddeley, 1993 PAMI 2010: geometric feature extraction by a multi-marked point process, florent lafarge, 2010
个人分类: 读书日记|5180 次阅读|1 个评论
什么是unit process?
热度 3 wanght 2011-2-7 23:53
Unit process(单元过程)是LCA中最常用的基本概念之一。很尴尬的是,这次我在日本开会才发现我的理解一直是错误的,幸好我的理解对这次会议的主题(Global Guidance for LCA database)也很重要,大家也顾不上嘲笑我,算是“因祸得福”吧。 先看这个让人晕死的ISO定义: ISO14040:2006 3.34 unit process: smallest element considered in the life cycle inventory analysis for which input and output data are quantified 大家的标准解释: 1)能有自己的量化的输入输出,这当然是指process(过程)。同一个标准中,Process的定义为3.11 process: set of interrelated or interacting activities that transforms inputs into outputs (我觉得改为model of activities with quantified inputs into outputs更好,即process是activity的模型) 2) 大家都认同(除了我不知道): 一个life cycle inventory(LCI) analysis中所有的process都是unitprocess,它的对立面是LCI results ,即各个unit process之和。这也意味着:当这个LCI results用到另一个LCI个案时,它又变成unit process了。 按照上面的解释,生命周期模型中所有的process都是smallest,并没有bigger process或biggestprocess,所以smallest其实是多余的。事实上,我能错这么多年而未发觉,按照PE的Martin的说法,意味着unit process根本不可能被用错——任何时候你看见生命周期模型中的任何一个框(box)称其为unit process,你都是对的。这其实是个多余的概念,直接称其为process就可以了。 我以前就是被smallest误导了,一直在想,smallest是与谁相比?比的是什么?所以我提了一个不同的解释: 1)当第一次各种原始数据Raw data经过计算被处理为一个清单数据集(dataset)时,这时对应的process是unit process。 2)之后,unit process dataset只能不断与其它datasets汇总(aggregation),而得到的都是aggregated process dataset。 3)随着汇总次数的增加,这个dataset所代表的人类活动的范围(scope of activity)在不断增大。从这个意义说,unit process是smallest! 我这样解释unit process的时候,Ecoinvent的Bo告诫我“I like your idea, but don't challenge ISO!”(他说不定就是当初写ISO定义的家伙?)郁闷中,我重新细看ISO的定义(见上),然后惊奇地发现......ISO只为LCI analysis个案研究定义了unit process,但并没有说数据库中各个清单数据集(dataset)对应的过程(process)应该叫什么! 这一天是2011.2.3,兔年新年,在去横滨中国城吃晚饭的车上我给Bo说:“在数据库里,dataset对应的process既不是unit process也不是LCI results”,他马上说:“Or, can be both.”(这就是我喜欢Bo的原因,他是LCA里最聪明的人之一)。 这意味着我们彻底解放了。既然我们现在讨论的只是数据库,ISO的定义并不适用(事实上,ISO压根就没考虑数据库的事)。所以,我们需要考虑:与LCI和LCA个案研究相比,数据库开发的特点是什么?需要什么样的基本概念和指南? 在我的解释中,区分了三个概念:raw data - unit process dataset - aggregated process dataset 1) raw data并未与process的参照流(reference)成线性比例,必须经过数据处理过程才能得到unit process dataset。这其实就是我们天天在做的数据收集工作。 2)unit process dataset是 第一次 被处理为 一组 相互成线性关系的清单数据(a set of input and output data related to the same reference of a process,所以称为dataset),它来自raw data,但尚未经历过汇总计算;尽管这里也称为unitprocess,但其实与ISO的unit process无关。 3)aggregated process dataset是由existing datasets汇总得到的,汇总的计算方法与普通LCI计算方法是一样的,即找到各过程的比例系数(scaling factor),调整比例后相加即可。 由此,数据库中清单数据集的开发被划分为两步: 1)从raw data - unit process dataset 2)从unit process dataset - aggregated process dataset 两步的数据来源、计算规则都完全不一样,所以需要完全不同的指南。这更清楚地界定了指南的各章之间的边界: 我所在的组负责从raw data - unit process dataset的指南,国际钢铁协会Clare的组负责从unit process dataset - aggregated process dataset的指南。 以Ecoinvent数据库为例,其中的systemdataset是通过建立生命周期模型计算得到的结果(按ISO定义,在那个模型中,这个system dataset是LCI results);然后这个system dataset放到Ecoinvent数据库中(这时ISO的术语失效了),Ecoinvent自己取了一个名字叫System dataset(在Global guidance中称为aggregated process dataset);当这个dataset被用到另一个生命周期模型中时(ISO术语生效了),它对应的process应该被称为unit process。所以我说:“既不是unit process也不是LCI results”,而BO说:“Or, can be both.” 现在不幸的是,ISO是用unit表示“大小”(最小),Global guidance用unit表示dataset汇总的历史(无汇总),放在一起仍然是混淆的,此unit不是彼unit。虽然挑战ISO的前景是很不光明的,但还是忍不住要彻底清算一下:-D -----问题背景----- * LCA中主要有“LCI/LCA模型”、“数据库”两种语境/上下文(context); * 在两种语境中,都会涉及process和dataset两个重要对象,dataset是process的属性。 * process之间(及其dataset之间)有各种差异,可能需要用不同的术语(名称)区分。 * 定义术语的关键问题在于:有哪些差异?哪种差异更重要,值得定义一个术语以示区别? -----具体问题----- 在上面关于unit process术语的讨论中,出现了两种差异:大小的差异,汇总(aggregation)的差异 1)大小的差异:按照ISO定义,一个模型中,最小的是unit process,与其对立的是life cycle/product system,但并不存在中间大小的东西。上述定义有两个问题: A. ISO没有说明是process的“什么”的大小。从unit process与life cycle的对立来看,ISO应该是指process所描述的activity的boundary的大小。但这种差异在LCA工作中几乎没有什么意义,例如我们不会比较一个炼钢的process与一个炼铜的process的大小,也不会比较中国炼钢的process与欧洲炼钢的process的大小。唯一有意义的情况是:用采矿、炼铁、炼钢等process汇总生成一个钢铁生产process,这种“大小差异”其实是由“汇总差异”造成的。如果一定要区分这种差异,记录“汇总历史”即可,完全用不着“大小”的概念! B.而且,life cycle与unit process其实是整体与组成部分的关系,并非最大与最小的关系。ISO不该用smallest,也不该定义术语unitprocess,直接称呼process就可以了。 2)汇总差异:从dataset的汇总历史看,“从raw data转变来的dataset(无汇总)”与“从dataset汇总得到的新dataset”是完全不同的事情,非常需要术语区别。这次的Global guidance中用unit process (dataset)和aggregated process (dataset)加以区分。 -----结论----- 1)“大小”只是浮云,汇总历史才是关键; 2)在各种语境中,区分unit process (dataset)和aggregated process (dataset)即可;无需区分汇总历史的时候,称为process (dataset)即可; 3)aggregated process dataset可以再细分各种差异,例如vertical/horizontal aggregated process, LCI result/partialy terminated process, cradle-to-gate/cradle-to-grave/(gate-to-gate,也可能是unit process)等等 -----经验教训----- 1)LCA是人为定义的统计方法,不严密的地方是存在的。讨论问题的时候,应该小心那些含糊的概念,说不定需要重新定义概念才行。 2)找到具体的问题多交流讨论。不说出来,就不知道自己是错的。 3)不要浪费问题。一个看似简单的问题可能发展为很深入的讨论。 4)我之后又看了一遍ISO,乖乖隆个冬,里面几乎一句都没提数据库的事,全部都是在讲LCI和LCA,说不定还有这种好事等着我们?
个人分类: LCA/LCM|7206 次阅读|5 个评论
[转载]Telomerase reverses ageing process
xupeiyang 2010-12-22 09:51
Telomerase reverses ageing process http://www.nature.com/news/2010/101128/full/news.2010.635.html Published online | Nature | doi:10.1038/news.2010.635 | Nature | doi:10.1038/news.2010.635 http://www.gopubmed.org/web/gopubmed/2?WEB01wfch4uc0t42qI3I1I00h001000j100200010 1,547 documents semantically analyzed top author statistics 1 2 Top Years Publications 2003 139 2008 137 2004 137 2010 128 2002 121 2009 120 2005 118 2007 114 2006 104 2000 91 2001 85 1998 70 1999 69 1997 56 1996 31 1995 15 1994 3 1992 3 1991 2 1990 2 1 2 1 2 3 Top Countries Publications USA 619 United Kingdom 143 Japan 126 Germany 100 China 63 Canada 56 Spain 56 France 39 Italy 37 South Korea 35 Australia 32 Russia 18 Denmark 15 Austria 12 Israel 12 Switzerland 12 Belgium 11 Sweden 10 Netherlands 9 Taiwan 8 1 2 3 1 2 3 ... 16 Top Cities Publications Dallas 50 Madrid 47 Baltimore 46 Boston 39 Los Angeles 33 New York City 32 London 32 Houston 29 Bethesda 27 Seoul, South Korea 27 Tokyo 24 Newcastle upon Tyne 22 Cardiff 22 Berkeley 22 San Francisco 20 Menlo Park 19 Stanford 19 Philadelphia 18 Glasgow 16 Montreal 15 1 2 3 ... 16 1 2 3 ... 28 Top Journals Publications Mech Ageing Dev 46 Exp Gerontol 43 Oncogene 40 Exp Cell Res 36 J Biol Chem 31 Biochem Bioph Res Co 25 P Natl Acad Sci Usa 24 Cancer Res 22 Aging Cell 21 Mol Cell Biol 20 Cell Cycle 19 Ann Ny Acad Sci 18 Embo J 17 Science 16 Circ Res 16 J Immunol 15 Nature 14 Rejuvenation Res 14 Nucleic Acids Res 13 Proc Natl Acad Sci U S A 12 1 2 3 ... 28 1 2 3 ... 311 Top Terms Publications Telomerase 1,519 Humans 1,268 Telomere 1,178 Cell Aging 1,097 chromosome, telomeric region 1,082 Aging 1,053 senescence 808 positive regulation of telomerase activity 768 regulation of telomerase activity 745 telomerase activity 738 negative regulation of telomerase activity 725 Animals 702 DNA 502 Neoplasms 458 Chromosomes 456 cell aging 452 Cell Division 437 Proteins 428 Cells, Cultured 399 Genes 397 1 2 3 ... 311 1 2 3 ... 260 Top Authors Publications Shay J 59 Wright W 47 Blasco M 39 Harley C 22 Mattson M 19 Reddel R 15 Greider C 15 DePinho R 14 Holt S 14 Lansdorp P 14 Hornsby P 14 Kipling D 14 Effros R 13 Rudolph K 13 Weng N 13 Von Zglinicki T 12 Hahn W 12 Ide T 12 Hodes R 11 Oshimura M 11 1 2 3 ... 260 publications over time
个人分类: 科学杂志|2485 次阅读|0 个评论
[Management Science] review process
feicheng 2010-12-20 07:28
(DE ----- AE----- four reviewers ----- AE ----- DE) 1) Within a day or two, it will be transferred to the appropriate Department Editor. You can confirm the status of your paper at any point in the review process at http://mc.manuscriptcentral. com/ms in your Author Center. 2) The DE will review the paper to determine if it should be sent to an Associate Editor (AE). It will be sent to an Associate Editor if the DE feels (i) the paper's research domain falls within the scope of the journal's editorial mission and (ii) the paper has the potential to make a sufficient contribution to the literature to consider it for publication in Management Science. If the DE decides not to send your paper to an AE, then the DE will notify you of this decision and provide an explanation. 3) Assuming the paper is sent to an AE, the AE will also make an initial screening for fit and contribution and decide whether further review is warranted. If not, the AE will upload a report explaining his/her reason for not reviewing the paper. 4) Assuming the AE sends the paper out for review, the AE will collect the reviews and write a recommendation to the DE. 5) The DE will make a final decision on the manuscript and send you the decision along with the set of all reports received. Thegoal at Management Science is to complete reviews in approximately 65 days (conditional that the AE sends the paper out for review) and 90% of our reviews within 90 days. If you have waited longer than 90 days, then you should feel free to contact the Department Editor to inquire regarding the status of your manuscript.
个人分类: 一般认识|6202 次阅读|0 个评论
在读;Process Simulation And Control Using Aspen
wangshilei 2010-4-20 19:00
在线阅读地址: http://books.google.com/books?id=PHaVttn2kgUCprintsec=frontcoverhl=zh-CN#v=onepageqf=false
个人分类: Aspen 软件学习系列|5114 次阅读|0 个评论
PostmasterMain()中的process table的初始化后内存结构
hillpig 2010-4-8 05:51
上次写完了“Postmaster的Memory Context 初始化内存结构” http://www.sciencenet.cn/m/user_content.aspx?id=308964 ,接下来我们看看初始化process table 部分。 Postmaster在PostmasterMain()-reset_shared()-CreateSharedMemoryAndSemaphores()中初始化process table的主要函数为: /* * Set up process table */ InitProcGlobal(); CreateSharedProcArray(); CreateSharedBackendStatus(); 我们可以猜想一下,对于接下来PostmasterMain马上要fork的postgres的进程中肯定是用到这些函数初始化的数据结构的。我们来一一分析一下。 1.InitProcGlobal() void InitProcGlobal(void) { PGPROC *procs; int i; bool found; /* Create the ProcGlobal shared structure */ ProcGlobal = (PROC_HDR *) ShmemInitStruct(Proc Header, sizeof(PROC_HDR), found); Assert(!found); /* * Create the PGPROC structures for auxiliary (bgwriter) processes, too. * These do not get linked into the freeProcs list. */ AuxiliaryProcs = (PGPROC *) ShmemInitStruct(AuxiliaryProcs, NUM_AUXILIARY_PROCS * sizeof(PGPROC), found); Assert(!found); /* * Initialize the data structures. */ ProcGlobal-freeProcs = NULL; ProcGlobal-autovacFreeProcs = NULL; ProcGlobal-spins_per_delay = DEFAULT_SPINS_PER_DELAY; /* * Pre-create the PGPROC structures and create a semaphore for each. */ procs = (PGPROC *) ShmemAlloc((MaxConnections) * sizeof(PGPROC)); if (!procs) ereport(FATAL, (errcode(ERRCODE_OUT_OF_MEMORY), errmsg(out of shared memory))); MemSet(procs, 0, MaxConnections * sizeof(PGPROC)); for (i = 0; i MaxConnections; i++) { PGSemaphoreCreate((procs .sem)); procs .links.next = (SHM_QUEUE *) ProcGlobal-freeProcs; ProcGlobal-freeProcs = procs ; } procs = (PGPROC *) ShmemAlloc((autovacuum_max_workers) * sizeof(PGPROC)); if (!procs) ereport(FATAL, (errcode(ERRCODE_OUT_OF_MEMORY), errmsg(out of shared memory))); MemSet(procs, 0, autovacuum_max_workers * sizeof(PGPROC)); for (i = 0; i autovacuum_max_workers; i++) { PGSemaphoreCreate((procs .sem)); procs .links.next = (SHM_QUEUE *) ProcGlobal-autovacFreeProcs; ProcGlobal-autovacFreeProcs = procs ; } MemSet(AuxiliaryProcs, 0, NUM_AUXILIARY_PROCS * sizeof(PGPROC)); for (i = 0; i NUM_AUXILIARY_PROCS; i++) { AuxiliaryProcs .pid = 0; /* marks auxiliary proc as not in use */ PGSemaphoreCreate((AuxiliaryProcs .sem)); } /* Create ProcStructLock spinlock, too */ ProcStructLock = (slock_t *) ShmemAlloc(sizeof(slock_t)); SpinLockInit(ProcStructLock); } 纵观其中的代码, ProcGlobal-freeProcs = procs ; 是我们比较感兴趣的。因为其他的初始化例如autovacuum类似的辅助进程我们暂时还不怎么关注。初始化的结果就是把这些结构用next指针给链接起来,本身不复杂。 2.CreateSharedProcArray() void CreateSharedProcArray(void){ bool found; /* Create or attach to the ProcArray shared structure */ procArray = (ProcArrayStruct *) ShmemInitStruct(Proc Array, ProcArrayShmemSize(), found); if (!found) { /* * We're the first - initialize. */ procArray-numProcs = 0; procArray-maxProcs = MaxBackends + max_prepared_xacts; } } 比较关键的代码已用红色标出,注意procArray的定义: typedef struct ProcArrayStruct { int numProcs; /* number of valid procs entries */ int maxProcs; /* allocated size of procs array */ /* * We declare procs ; /* VARIABLE LENGTH ARRAY */ } ProcArrayStruct; static ProcArrayStruct *procArray; 很简单,无需解释。 3.CreateSharedBackendStatus() void CreateSharedBackendStatus(void){ Size size; bool found; int i; char *buffer; /* Create or attach to the shared array */ size = mul_size(sizeof(PgBackendStatus), MaxBackends); BackendStatusArray = (PgBackendStatus *) ShmemInitStruct(Backend Status Array, size, found); if (!found) { /* * We're the first - initialize. */ MemSet(BackendStatusArray, 0, size); } /* Create or attach to the shared activity buffer */ size = mul_size(pgstat_track_activity_query_size, MaxBackends); BackendActivityBuffer = (char *) ShmemInitStruct(Backend Activity Buffer, size, found); if (!found) { MemSet(BackendActivityBuffer, 0, size); /* Initialize st_activity pointers. */ buffer = BackendActivityBuffer; for (i = 0; i MaxBackends; i++) { BackendStatusArray .st_activity = buffer; buffer += pgstat_track_activity_query_size; } } } 主要初始化了 BackendStatusArray和 BackendActivityBuffer, 并让 BackendStatusArray 各项指向 BackendActivityBuffer 中的各项。 这样,我们分析结束了,结论就是其实这一部分的初始化,基本上没有设定具体数据结构的值,只是把大体框架给列在那里了。 最后,我给出初始化之后在内存中的表示,图有点大的吓人,不过你可以略去shared buffer和shared memory的部分。 至此,结束。 加我私人微信,交流技术。
个人分类: postgresql|3837 次阅读|0 个评论

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-4-23 17:05

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部