convergence rate 中文意思是什麼

convergence rate 解釋
收斂率
  • convergence : n. 1. 聚合，會聚，輻輳，匯合。2. 集合點；【數、物】收斂；【生物學】趨同（現象）。
  • rate : n 1 比率，率；速度，進度；程度；（鐘的快慢）差率。2 價格；行市，行情；估價，評價；費，費用，運費...
  1. We know that the convergence rate is related to lk, uk according to the previous study, t

    變得越來越大時，上面兩個問題為大型問題且其hesse矩陣為稀疏矩陣。
  2. The hybrid algorithm raises the convergence rate of the conjugate gradient method and solves the problem for which the convergence rate of the steepest descent method get slower when the isopleth of goal function is oblong

    這種混合優化演算法結合了共扼梯度法和最速下降法產生搜索方向，既提高了共扼梯度演算法的收斂速度，又解決了目標函數的等值線是扁長橢球時，最速下降法下降緩慢的問題，具有收斂速度快、收斂范圍大、適應面廣等特點。
  3. Using cross - validation to select h as hn *. under proper conditions, this paper gives the bounds of hn *, and the convergence rate and the weak consistency of g ( t )

    ） ，在一定的正則性條件下，給出了（ ? ）的上下界估計和g （ t ）的估計的收斂速度。
  4. Compared with the classical bp algorithm, robust adaptive bp algorithm possesses some advantages as following : ( 1 ) increasing the accuracy of the network training by means of using both the relative and absolute residual to adjust the weight values ; ( 2 ) improve the robustness and the network convergence rate through combining with the robust statistic technique by way of judging the values of the samples " relative residual to establish the energy function so that can suppress the effect on network training because of the samples with high noise disturbances ; ( 3 ) prevent entrapping into the local minima area and obtain the global optimal result owing to setting the learning rate to be the function of the errors and the error gradients when network is trained. the learning rate of the weights update change with the error values of the network adaptively so that can easily get rid of the disadvantage of the classical bp algorithm that is liable to entrap into the local minima areas

    與基本bp演算法相比，本文提出的魯棒自適應bp演算法具有以下優點： （ 1 ）與魯棒統計技術相結合，通過訓練樣本相對偏差的大小，確定不同訓練樣本對能量函數的貢獻，來抑制含高噪聲干擾樣本對網路訓練的不良影響，從而增強訓練的魯棒性，提高網路訓練的收斂速度； （ 2 ）採用相對偏差和絕對偏差兩種偏差形式對權值進行調整，提高了網路的訓練精度； （ 3 ）在採用梯度下降演算法對權值進行調整的基礎上，通過將學習速率設為訓練誤差及誤差梯度的特殊函數，使學習速率依賴于網路訓練時誤差瞬時的變化而自適應的改變，從而可以克服基本bp演算法容易陷入局部極小區域的弊端，使訓練過程能夠很快的「跳出」局部極小區域而達到全局最優。
  5. Among hese method, one of the simplest is the deepest gradient descent algorithm, but it has one disadvantage that the solutions are oscillatory and slowly convergent. a friction term is ntroduced to avoid oscillatory solutions and raise convergence rate. the third major contribution of this dissertation is study on 3 - d diffuse object surface optical automatic measurement

    其次，結合要處理的具體問題，介紹數值優化及幾種常用的無約束最優化計算方法，以最優梯度法為基礎，討論其演算法的優缺點，通過引進阻尼系數，改進了演算法的收斂速度，從而提高迭代精度。
  6. Because the adaptive algorithm of conventional adaptive noise canceller is the least mean squares ( lms ), and the convergence rate of lms is heavily dependent on the eigenvalue distribution of the autocorrelation matrix of the input signal, thus lms converges at unacceptably low rates when the input signal is colored noise or speech

    由於傳統自適應噪聲抵消系統（ anc ）自適應演算法主要採用lms演算法，而lms演算法收斂速度依賴于輸入信號自相關矩陣特徵值的分散程度。因此，當輸入信號是語音或有色噪聲時， lms的收斂速度很慢。
  7. Analysis of the convergence rate on newton iteration method

    牛頓迭代法收斂速度分析
  8. The artificial neural net ( ann ) way is universal regard as one of the most effective ways of stlf. in this paper, some research is developed for stlf using ann ways in several parts : the first part is about the arithmetic of ann based on bp model, namely the advanced of traditional bp arithmetic, one alterable step and scale bp arithmetic based on comparability of model and probability of accepting bp arithmetic is used to enhances a lot the convergence rate of learning process of bp network, but also avoid the stagnation problem to some extent. it indicates that the ann ' s efficiency and precision by the way can be ameliorated by the simulation of real data

    神經網路方法在短期預測中已經被公認為較有效的方法，本文針對神經網路用於電力系統短期負荷預測的幾個方面展開研究工作：第一部分研究一般用於負荷預測的神經網路bp模型的演算法，即對傳統的bp演算法的改進，將一種基於模式逼近度和接受概率的變步長快速bp演算法應用到短期負荷預測，模擬結果表明該方法有效的改善了bp演算法收斂速度慢以及容易陷入局部最小點的缺點，從而提高了神經網路用於負荷預測的效率和精度。
  9. In this paper, we give a kernel shape estimation of m ( x ) using variable bandwidth local linear refression approch, and discuss the asymptotic normality, the convergence rate of mean square and convergence rate with probability

    本文對上述模型，利用變窗寬局部線性回歸方法，給出了m （ x ）的核形估計，並討論了這一估計的漸近正態性、依概率收斂速度、和均方收斂速度。
  10. A class of global convergent memory gradient methods and its linear convergence rate

    一類全局收斂的記憶梯度法及其線性收斂性
  11. In aim solver, the convergence rate depends on spectral properties of the aim matrix, and the near - zone threshold distance has a dramatic impact on the memory and cpu time. a poorly conditioned matrix occurs through aim when a smaller near - zone

    另外，自適應積分方法計算中，在保證計算精度的條件下，較小的耦合距離將極大減小內存需求和計算的復雜度，但同時誤差積累也導致矩陣的性態變壞。
  12. Chapter 5 is contributed to studying convergence rate of standard transition function and relation of all kinds of convergence rates. by revealing the close link among strong ergodicity, stochastically monotone, and the feller transition functions we are able to prove that a monotone ergodic transition function is strongly ergodic if and only if it is not a feller transition function

    第五章討論標準轉移函數的收斂速度以及各種收斂速度之間的關系，通過找出強遍歷、隨機單調性和feller轉移函數之間的緊密關系，我們證明了隨機單調標準轉移函數是強遍歷的充要條件是它不是feller轉移函數。
  13. The results of numerical experiments, using the four - dimensional variational data assimilation system of the nonhydrostatic mesoscale model mm5, show that : the tangent linear model and adjoint model conducted by keeping the " on - off " switches the same as the basic state, can provide a good approximation of the first - order information to the nlm perturbation and a good descent direction for the minimization procedure ; switching on and off at every other time step in the kuo cumulus parameterization scheme do n ' t impact the convergence rate of cost function ; the existence of the switches do n ' t impact improvement to the mm5 model rainfall prediction because that not only specific humidity, but also wind, temperature and pressure are assimilated into the model

    非靜力中尺度數值模式mm5的四維變分資料同化系統進行的數值試驗結果表明： 「開關」變量保持與基態一致，所構造的切向線性模式能夠提供關于非線性模式擾動的一階近似，伴隨模式所計算的梯度值能夠為最小化過程提供較好的下降方向；郭氏降水參數化方案中對流每隔一個積分步的交替發生並不影響目標函數最小化的收斂速度； 「開關」變量的存在也不影響將風、溫度、氣壓和比濕結合起來同化對mm5降水預報準確性的提高。
  14. In chapter 1, we use the p1 nonconforing mortar element to discrete poisson equation, and adopt multigrid method to solve the discrete equations, we have proven that the algorithm has a uniform convergence rate, that is, the convergence rate is independent of the mesh size and level. the multigrid method for the stationary stokes equations based on mortar mixed element method is presented in chapter 2, and the uniform convergence rate of the method independent of mesh size and level for the w - cycle is proven

    第二章求解的問題是定常stokes方程，用混合mortar元來離散原方程，每個子區域上用taylor - hood有限元進行離散，離散出的方程組用w -循環多重網格方法進行求解，第四節中證明了多重網格方法的一致收斂性，即收斂率與網格層數和網格尺寸無關。
  15. In this paper, a class of the second order accurate explicit gauss schemes with staggered grids for the computation of solutions of hyperbolic conservation laws are presented, the advantages of these schemes are : riemann solver - free, faster and programming is much simple, no complete set of eigenvectors is needed and hence weakly hyperbolic system can be solved. in one dimensional case, these schemes are and total variation diminishing and convergence under the restriction of cfl condition, the convergence rate is the first order, and a pointwise error bound is presented

    本文在交錯網格的情況下，利用gauss型求積公式構造了一類求解雙曲守恆律的時空一致二階顯式gauss型差分格式，這類gauss型差分格式，具有不需要求解riemann問題、計算簡單、工作量少、編程簡便等優美特點，而且由於這類格式在應用於求解方程組的時候，不需要對方程組進行特徵分解，因此可應用於求解非嚴格的雙曲守恆律方程組。
  16. As the back - propagation neural network ( bpnn ) is adopted to diagnose the dissolved gases of transformer, convergence rate becomes slow, convergence precision becomes bad and even being out of convergence as more sampled data are trained and more complicated relation between input and output becomes

    根據電力變壓器油中溶解氣體分析（ dga ） ，採用反向傳播神經網路（ bpnn ）對電力變壓器進行故障診斷時，隨著學習樣本數目的增多，輸入輸出關系的復雜化，網路收斂的速度將變得緩慢，收斂精度不理想甚至不收斂。
  17. A kind of complete convergence of sums for negatively associated sequences of non - identically distributed random variables, in the second chapter, is obtained and the requirement of known results are weakened to the condition that absoluted moment - larger than zero - is finite. the strong convergence of negatively associated sequences of non - identically distributed random variables is discussed in the third chapter. in the fourth chapter, after extend the laws of the iterated logarithm of strong stationary case to weak stationary case, we obtain the strong convergence rate for negatively associated sequences of non - identically distributed random variables in linear models

    其中第二章討論了一類不同分佈的na列的加權和的完全收斂性，我們把已有的結果對矩的要求放寬到了只要求大於0的絕對矩有限的情形；第三章討論了不同分佈的na列的加權和的強收斂性；第四章首先把文[ 10 ]的關于na的重對數律由強平穩的情形推廣到了弱平穩不同分佈的情形，然後得到了線性模型中不同分佈的na誤差列的收斂速度。
  18. Chapter 6 is devoted to studying convergence rate of the birth and death processes. for a conservative birth and death q - matrix, we prove that the minimal q - function is strongly ergodic if and only if r - and s <. suppose a birth and death q - matrix satisfies r < and s

    第六章對生滅過程中的收斂速度進行了研究，得到了對保守生滅q -矩陣，最小q -函數是強遍歷的充要條件是r =和s ；且當r和s時，存在唯一可配稱誠實q -函數是強遍歷的。
  19. The main contributions of this dissertation are summarized as follow : ( 1 ) an ilc approach combining feedforward with current feedback is developed based on optimal feedback control and the gradient method. a sufficient condition that guarantees the convergences is given for linear system. the procedures of designing the algorithm can employ lqr, h2 or h approaches to improve the convergence rate of learning in iterations

    本文的主要成果有： 1 、在開閉環綜合迭代學習控制結構的基礎上，分析了利用梯度下降法設計前饋迭代學習控制器時，為保證演算法的收斂性，閉環控制系統應該滿足的充分條件，並依據提高演算法收斂速率的優化條件，給出了基於lqr 、 h _ 2和h等優化控制技術的迭代學習控制演算法的設計方法。
  20. Strong convergence rate of recursive probability density estimator for na sequences

    負相協下遞歸密度估計的強收斂速度
分享友人
分享到 Line

分享到臉書