Fast Exact Univariate Kernel Density Estimation

Size: px
Start display at page:

Download "Fast Exact Univariate Kernel Density Estimation"

Transcription

1 Fast Exact Univariate Kernel Density Estimation David P. Hofmeyr Department of Statistics and Actuarial Science, Stellenbosc University arxiv: v2 [stat.co] 12 Jul 2018 July 13, 2018 Abstract Tis paper presents new metodology for computationally efficient kernel density estimation. It is sown tat a large class of kernels allows for exact evaluation of te density estimates using simple recursions. Te same metodology can be used to compute density derivative estimates exactly. Given an ordered sample te computational complexity is linear in te sample size. Combining te proposed metodology wit existing approximation metods results in extremely fast density estimation. Extensive erimentation documents te effectiveness and efficiency of tis approac compared wit te existing state-of-te-art. Keywords: linear time, density derivative 1

2 1 Introduction Estimation of density functions is a crucial task in loratory data analysis, wit broad application in te fields of statistics, macine learning and data science. Here a sample of observations, x 1,..., x n, is assumed to represent a collection of realisations of a random variable, X, wit unknown density function, f. Te task is to obtain an estimate of f based on te sample values. Kernel based density estimation is arguably te most popular non-parametric approac. In kernel density estimation, te density estimator, ˆf, is given by a mixture model comprising a large number of (usually n) components. In te canonical form, one as ˆf(x) = 1 n x xi K, (1) were K( ) is called te kernel, and is a density function in its own rigt, satisfying K 0, K = 1. Te parameter > 0 is called te bandwidt, and controls te smootness of ˆf, wit larger values resulting in a smooter estimator. A direct evalution of (1) at a collection of m evaluation points, { x 1,..., x m }, as computational complexity O(nm), wic quickly becomes proibitive as te sample size becomes large, especially if te function estimate is required at a large number of evaluation points. Furtermore many popular metods for bandwidt selection necessitate evaluating te density estimate (or its derivatives) at te sample points temselves (Scott and Terrell, 1987; PW, 1976; Seater and Jones, 1991), making te procedure for coosing quadratic in computational complexity. Existing metods wic overcome tis quadratic complexity barrier are limited to kernels wit bounded support and te Laplace kernel (Fan and Marron, 1994), or tey rely on approximations. Popular approximations including binning (Scott and Seater, 1985; Hall and Wand, 1994) and te fast Gauss (Yang et al., 2003, FGT) and Fourier (Silverman, 1982, FFT) tranforms, as well as combinations of tese. A more recent approac (Raykar et al., 2010) relies on truncations of te Taylor series ansion of te kernel function. Generally speaking tese metods reduce te complexity to O(n + m), wit te constant term depending on te desired accuracy level. In tis paper te class of kernels of te form K(x) = poly( x ) ( x ), were poly( ) denotes a polynomial function of finite degree, is considered. It is sown tat tese kernels 2

3 allow for extremely fast and exact evaluation of te corresponding density estimates. Tis is acieved by defining a collection of O((α + 1)n) terms, were α is te degree of te polynomial, of wic te values { ˆf(x 1 ),..., ˆf(x n )} are linear combinations. Tese terms arise from loiting te binomial ansion of polynomial terms and te trivial factorisation of te onential function. Furtermore tese terms can be computed recursively from te order statistics of te sample. Given an ordered sample, te exact computation of te collection of values { ˆf(x 1 ),..., ˆf(x n )} terefore as complexity O((α + 1)n). Hencefort we will use poly α ( ) to denote a polynomial function of degree α. An important benefit of te proposed kernels over tose used in te fast sum updating approac (Fan and Marron, 1994), is tat bounded kernels cannot be reliably used in cross validation pseudo-likeliood computations. Tis is because te likeliood for points wic do not lie witin te support of te density estimate based on te remaining points is zero. Numerous popular bandwidt selection tecniques can terefore not be applied. Remark 1 Te derivative of a poly α ( x ) ( x ) function is equal to x multiplied by a poly α 1 ( x ) ( x ) function, provided tis derivative exists. Te proposed metodology can terefore be used to exactly and efficiently evaluate { ˆf (k) (x 1 ),..., ˆf (k) (x n )}, were ˆf (k) denotes te k-t derivative of ˆf. Altoug a given poly( x ) ( x ) function is not infinitely differentiable at 0, for a given value of k it is straigtforward to construct a poly( x ) ( x ) function wit at least k continuous derivatives. An alternative is to utilise leave-one-out estimates of te derivative, wic can be computed for any poly( x ) ( x ) function provided no repeated values in te sample. Remark 2 Te proposed class of kernels is extremely ric. Te popular Gaussian kernel is a limit case, wic can be seen by considering tat te density of an arbitrary sum of Laplace random variables lies in tis class. Te remainder of te paper is organised as follows. In Section 2 te kernels used in te proposed metod are introduced, and relevant properties for kernel density estimation are discussed. It is sown tat density estimation using tis class of kernels can be performed in linear time from an ordered sample using te recursive formulae mentioned above. An extensive simulation study is documented in Section 3, wic sows te efficiency and effectiveness of te proposed approac. A final discussion is given in Section 4. 3

4 2 Computing Kernel Density Estimates Exactly Tis section is concerned wit efficient evaluation of univariate kernel density estimates. A general approac for evaluating te estimated density based on kernels wic are of te type K(x) = poly( x ) ( x ) is provided. Tese kernels admit a convenient algebraic ansion of teir sums, wic allows for te evaluation of te density estimates using a few simple recursions. Te resulting computational complexity is O((α + 1)n) for an ordered sample of size n, were α is te degree of te polynomial. To illustrate te proposed approac we need only consider te evaluation of a function of te type ( x x i α x x ) i, (2) for an arbitrary α {0, 1, 2,...}. Te extension to a linear combination of finitely many suc functions, of wic ˆf is an example, is trivial. To tat end let x (1) x (2)... x (n) be te order statistics from te sample. Ten define for eac k = 0, 1, 2,..., α and eac j = 0, 1, 2,..., n te terms l(k, j) = r(k, j) = j ( ( x (i) ) k x(i) x (j) ( (x (i) ) k x(j) x (i) i=j+1 ), (3) ), (4) were for convenience l(k, 0) and r(k, n) are set to zero for all k. Next, for a given x R define n(x) to be te number of sample points less tan or equal to x, i.e., n(x) = 4

5 n δ x i ((, x]), were δ xi ( ) is te Dirac measure for x i. Ten, ( x x i α x x ) i ( = x x (i) α x x ) (i) n(x) = (x x (i) ) α x(i) x x + (x (i) x) α x(i) n(x) x(n(x)) x = x x(n(x)) + ( ( α x(n(x)) x = k i=n(x)+1 α x α k ( x (i) ) k k i=n(x)+1 α x k k (i)( x) α k ) x α k l(k, n(x)) + x(i) x (n(x)) x(n(x)) x (i) ) x x(n(x)) ( x) α k r(k, n(x)). Now, if x is itself an element of te sample, say x = x (j), ten we ave ( x (j) x i α x ) (j) x i α = ((x(j)) α k l(k, j) + ( x (j)) α k r(k, j) ). k Te values ˆf(x (1) ),..., ˆf(x (n) ) can terefore be ressed as linear combinations of terms in k,j {l(k, j), r(k, j)}. Next it is sown tat for eac k = 0,..., α, te terms l(k, j), r(k, j) can be obtained recursively. Consider, j+1 l(k, j + 1) = ( x (i) ) k x(i) x (j+1) j = ( x (i) ) k x(i) x (j) + x (j) x (j+1) + ( x (j+1) ) k j x(j) x (j+1) = ( x (i) ) k x(i) x (j) + ( x (j+1) ) k x(j) x (j+1) = l(k, j) + ( x (j+1) ) k. And similarly, r(k, j 1) = (x (i) ) k x(j 1) x (i) i=j x(j 1) x ( (j) ) = (x (i) ) k x(j) x (i) + (x (j) ) k i=j+1 x(j 1) x (j) (r(k, = j) + (x(j)) k). 5

6 Te complete set of values ˆf(x (1) ),..., ˆf(x (n) ) can tus be computed wit a single forward and a single backward pass over te order statistics, requiring O((α + 1)n) operations in total. On te oter and evaluation at an arbitrary collection of m evaluation points requires O((α + 1)(n + m)) operations. Relevant properties of te cosen class of kernels can be simply derived. Consider te kernel given by were c is te normalising constant. K(x) = c ( x ) β k x k, Of course c can be incorporated directly into te coefficients β 0,..., β α, but for completeness te un-normalised kernel formulation is also considered. It migt be convenient to a practitioner to only be concerned wit te sape of a kernel, wic is defined by te relative values of te coefficients β 0,..., β α, witout necessarily concerning temselves initially wit normalisation. Many important properties in relation to te field of kernel density estimation can be simply derived using te fact tat Specifically, one as c 1 = σ 2 K := R(K) := x k ( x )dx = 2 β k x k ( x )dx = 2 = c 2 x 2 K(x)dx = c K(x) 2 = c 2 j=0 β k β j 1 2 j=0 0 x k ( x)dx = 2k! β k k! x k+2 ( x )dx = 2c β k β j x k+j ( 2 x )dx x k+j 2 k+j ( x )dx = c 2 j=0 β k (k + 2)! β k β j (k + j)! 2k+j Furtermore it can be sown tat for K(x) to ave at least k continuous derivatives it is sufficient tat β i 1, for all i = 0, 1,..., k. i! 6

7 Coosing te simplest kernel from te proposed class (i.e., tat wit te lowest degree polynomial) wic admits eac smootness level leads us to te sub-class defined by K α (x) := 1 2(α + 1) x k k! ( x ), α = 1, 2,... (5) In te eriments presented in te following section te kernels K 1 and K 4 will be considered. Te kernel K 1 is cosen as te simplest differentiable kernel in te class, wile K 4 is selected as it as efficiency very close to tat of te ubiquitous Gaussian kernel. Te efficiency of a kernel K relates to te asymptotic mean integrated error wic it induces, and may be defined as eff(k) := (σ K R(K)) 1. It is standard to consider te relative efficiency eff rel (K) := eff(k)/eff(k ). Te kernel K is te kernel wic maximises eff(k) as defined ere, and is given by te Epanecnikov kernel. Te efficiency and sape of te cosen kernels can be seen in Figure 1, and in relation to te popular Gaussian kernel. Remark 3 Te efficiency of a kernel is more frequently defined as te inverse of te definition adopted ere. It is considered preferable ere to speak of maximising efficiency, rater tan minimising it, and ence te above formulation is adopted instead. 2.1 Density Derivative Estimation It is frequently te case tat te most important aspects of a density for analysis can be determined using estimates of its derivatives. For example, te roots of te first derivative provide te locations of te stationary points (modes and anti-modes) of te density. In addition pointwise derivatives are useful for determining gradients of numerous projection indices used in projection pursuit (Huber, 1985). Te natural estimate for te k-t derivative of f at x is simply, f (k) (x) = ˆf (k) (x) = 1 n = 1 n k+1 d k x dx K xi k x K (k) xi Only te first derivative will be considered licitly, were iger order derivatives can be simply derived, given an appropriate kernel (i.e., one wit sufficiently many derivatives). 7

8 Relative Efficiency K(x) α x (a) Relative efficiency of K α for α = 0, 1,..., 15. Relative efficiency of Gaussian kernel ( ). (b) Plots of K 1 ( ), K 4 ( ) and Gaussian kernel ( ) Figure 1: Relative efficiency and sape of te kernels used in eriments Considering again te kernels K α as defined previously, consider tat for α = 1, 2,... we ave K α (1) (x) = d 1 x k ( x ) dx 2(α + 1) k! ( 1 k x k 1 sign(x) = ( x ) sign(x) ( x ) 2(α + 1) k! ( = ( x ) α 1 ) x x k 1 x x k 1 2(α + 1) k! k! = ( x ) 2(α + 1) x x α 1 α! = ( x ) 2(α + 1)! x x α 1 ) x k k! 8

9 To compute estimates of ˆf (1) (x) only a very sligt modification to te metodology discussed previously is required. Specifically, consider tat ( (x x i ) x x i α 1 x x ) i n(x) = (x x (i) ) α x(i) x x (x (i) x) α x(i) i=n(x)+1 ( ) α x(n(x)) x x = x α k x(n(x)) l(k, n(x)) ( x) α k r(k, n(x)). k Te only difference between tis and te corresponding terms in te estimated density is te - separating terms in te final ression above. Now, using te above ression for K (1) α R(K (1) α ) = = = 1 (2(α + 1)!) (2(α + 1)!) 2 2 (x), consider tat (2α)! ((α + 1)!) 2 2 2α 2. x 2α+2 ( 2 x )dx x 2α+2 ( x )dx 22α+2 Unlike for te task of density estimation, te relative efficiency of a kernel for estimating te first derivative of a density function is determined in relation to te biweigt kernel. Te relative efficiency of te adopted class for estimation of te derivative of a density is sown in Figure 2. Te relative efficiency of te Gaussian kernel is again included for context. Again te kernel K 4 as similar efficiency to te Gaussian. Here, unlike for density estimation, we can see a clear maximiser wit K 7. Tis kernel will terefore also be considered for te task of density derivative estimation in te eriments to follow. 3 Simulations Tis section presents te results from a toroug simulation study conducted to illustrate te efficiency and effectiveness of te proposed approac for density and density derivative estimation. A collection of eigt univariate densities are considered, many of wic are taken from te popular collection of bencmark densities given in Marron and Wand (1992). Plots of all eigt densities are given in Figure 3 9

10 Relative Efficiency α Figure 2: Relative efficiency of kernels K α, α = 1, 2,..., 15, for estimating te derivative of a density function. Relative efficiency of Gaussian kernel ( ) (a) Gaussian (b) Uniform (c) Scale mixture (d) Simple bimodal (e) Skew (f) Spiked bimodal (g) Claw () Skew bimodal Figure 3: Collection of densities used in eriments. 10

11 For context, comparisons will be made wit te following existing metods. For tese te Gaussian kernel was used, as te most popular kernel used in te literature. 1. Te exact estimator using te Gaussian kernel, for wic te implementation in te R package kedd (Guidoum, 2015) was used. Tis approac was only applied to samples wit fewer tan observations, due to te ig computation time required for large samples. 2. Te binned estimator wit Gaussian kernel using te package KernSmoot (Wand, 2015). 3. Te fast Fourier transform using R s base stats package. 4. Te truncated Taylor ansion approac (Raykar et al., 2010), for wic a wrapper was created to implement te autors c++ code 1 from witin R. Te main computational components of te proposed metod were implemented in c++, wit te master functions in R via te Rcpp (Eddelbuettel and François, 2011) package 2. Te binned estimator using te proposed class of kernels will also be considered. Because of te nature of te kernels used, as discussed in Section 2, te computational complexity of te corresponding binned estimator is O(n + (α + 1)b), were b is te number of bins. Accuracy will be assessed using te integrated squared error between te kernel estimates and te true sampling densities (or teir derivatives), i.e., ˆf (k) f (k) 2 2 = ( ˆf (k) (x) f(x) (k) ) 2 dx. Exact evaluation of tese integrals is only possible for very specific cases, and so tey are numerically integrated. For simplicity in all cases Silverman s rule of tumb is used to select te bandwidt parameter (Silverman, 2018). Tis extremely popular euristic is motivated by te optimal asymptotic mean integrated squared error (AMISE) bandwidt value. Te euristic is most commonly applied to density estimation, were te direct extension to te first two derivatives will also be used erein. For kernel K te 1 te autors code was obtained from ttps:// Software/optimal_bw/optimal_bw_code.tm 2 A simple R package is available from ttps://gitub.com/davidhofmeyr/fkde 11

12 AMISE optimal bandwidt is given by AMISE = (2k + 1)R(K (k) 1/(2k+5) ) σk 4 R(f. (k+2) )n Tis objective is generally preferred over te mean integrated squared error as it reduces te dependency on te underlying unknown density function to only te functional R(f (k+2) ). Silverman s euristic replaces R(f (k+2) ) wit R(φ (k+2) ˆσ ), were φ σ is te normal density wit scale parameter σ. Te scale estimate ˆσ is computed from te observations, usually as teir standard deviation. 3.1 Density Estimation In tis subsection te accuracy and efficiency of te proposed metod for density estimation are investigated Evaluation on a Grid Many of te approximation metods for kernel density estimation necessitate tat te evaluation points, { x 1,..., x m }, are equally spaced (Scott and Seater, 1985; Silverman, 1982). In addition suc an arrangement is most suitable for visualisation purposes. Here te speed and accuracy of te various metods for evaluation/approximation of te density estimates are considered, were evaluation points are restricted to being on a grid. Accuracy: Te accuracy of all metods is reported in Table 1. Sixty samples were drawn from eac density, tirty of size and tirty of size Te number of evaluation points was kept fixed at Te estimated mean integrated squared error is reported in te table. Te lowest average is igligted in eac case. Te error values for all metods utilising te Gaussian kernel (φ) are extremely similar, wic attests to te accuracy of te approximate metods. Te error of kernel K 4 is also very similar to tat of te Gaussian kernel metods. Tis is unsurprising due to its similar efficiency value. Te kernel K 1 obtains te lowest error over all and in te most cases. In addition te estimated pointwise mean squared error for density (d) was computed for te exact estimates using kernels K 1 and K 4 and te truncated Taylor approximation 12

13 Table 1: Estimated mean integrated squared error of density estimates from 30 replications. Sets of and observations are considered. Lowest average error for eac scenario is igligted in bold. Apparent ties were broken by considering more significant figures. Metod Density Exact φ Tr. Taylor φ Binned φ FFT φ Exact K 1 Exact K 4 Binned K 1 Binned K 4 (a) n=1e e e e e e e e e-04 n=1e e e e e e e e-06 (b) n=1e e e e e e e e e-02 n=1e e e e-03 8e e-03 8e e-03 (c) n=1e e e e e e e e e-01 n=1e e e e e e e e-03 (d) n=1e e e e e e e e e-03 n=1e e e-05 1e e e e e-05 (e) n=1e e e e e e e e e-01 n=1e e e e e e e e-03 (f) n=1e e e e e e e e e-03 n=1e e e e e e e e-03 (g) n=1e e e e e e e e e-02 n=1e e e e e e e e-03 () n=1e e e e e e e e e-01 n=1e e e e e e e e-02 13

14 MSE 0e+00 2e 04 4e 04 6e 04 8e x MSE 0e+00 2e 06 4e 06 6e x (a) n = 1000 (b) n = Tr. Taylor φ ( ), Exact K 1 ( ), Exact K 4 ( ) Figure 4: Estimated pointwise mean squared error for density (d). for te Gaussian kernel estimate. Tese can be seen in Figure 4. In addition te sape of density (d) is sown. Tis density was cosen as it illustrates te improved relative performance of more efficient kernels as te sample size increases. For te smaller sample size kernel K 1 as a lower estimated mean integrated squared error, wic is evident in Figure 4(a). Te mean squared error for te oter two metods is almost indistinguisable. On te oter and for te large sample size, sown in Figure 4(b), te error for kernel K 1 is noticeably larger at te extrema of te underlying density tan K 4 and te Gaussian approximation. A brief discussion will be given in te discussion to follow in relation to kernel efficiency and te coice of kernel. Computational efficiency: Te running times for all densities are extremely similar, and more importantly te comparative running times between different metods are almost exactly te same accross te different densities. It is terefore sufficient for comparisons to consider a single density. Note tat in order to evaluate te density estimate at a point not in te sample, te proposed approac requires all computations needed to evaluate te density at te sample points. Evaluation on a grid may terefore be seen as someting 14

15 of a worst case for te proposed approac. However, once te decision to evaluate te density estimate at points oter tan te sample points as been made, te marginal cost of increasing te number of evaluation points is extremely small. Tis fact is well captured by Figure 5. Tis figure sows plots of te average running times from te metods considered wen applied to density (d), plotted on a log-scale. Figure 5(a) sows te effect of increasing te number of observations, wile keeping te number of evaluation points fixed at On te oter and Figure 5(b) sows te case were te number of observations is kept fixed (at ) and te number of evaluation points is increased. In te former te proposed metod, despite obtaining an exact evalution of te estimate density, is reasonably competitive wit te slower of te approximate metods. It is also orders of magnitude faster tan te exact metod using te Gaussian kernel. In te latter it can be seen tat as te number of evaluation points increases te proposed exact approac is even competitive wit te fastest approximate metods. Overall te binned approximations provide te fastest evaluation. Te nature of te proposed kernels and te proposed metod for fast evaluation means tat te corresponding binned estimators (particularly tat pertaining to kernel K 1 ) are extremely computationally efficient Evaluation at te Sample Points Evaluation of te estimated density at te sample points temselves as important applications in, among oter tings, computation of non-parametric pseudo-likelioods and in te estimation of sample entropy. Of te metods considered only te exact metods and te truncated Taylor ansion approximation are applicable to tis problem. Table 2 sows te average integrated squared error of te estimated densities from 30 replications for eac sampling scenario. Unsuprisingly te accuracy values and associated conclusions are similar to tose for te grid evaluations above. An important difference is tat wen te density estimates are required at all of te sample values, te proposed exact metod outperforms te approximate metod in terms of computation time. Tis is seen in Table 3, were te average running times for all densities are reported. Te exact evalution for kernel K 4 is similar to te truncated Taylor approximate metod, wile te exact evaluation using 15

16 running time (seconds) 1e 03 1e 02 1e 01 1e+00 1e+03 1e+04 1e+05 1e+06 sample size running time (seconds) 1e 02 1e 01 1e+00 1e+03 1e+04 1e+05 1e+06 number of grid points (a) Fixed number of evaluation points, increasing sample size (b) Fixed number of observations, increasing number of evaluation points Exact φ ( ), Tr. Taylor φ ( + ), Binned φ ( ), FFT φ ( ), Exact K 1 ( ), Exact K 4 ( ), Binned K 1 ( ), Binned K 4 ( ) Figure 5: Computation times for density (d) evaluated on a grid kernel K 1 is rougly five times faster wit te current implementations. Remark 4 It is important to reiterate te fact tat te proposed approac is exact. Tis exactness becomes increasingly important wen tese density estimates form part of a larger routine, suc as maximum pseudo-likeliood or in projection pursuit. Wen te density estimates are only approximate it becomes more difficult to determine ow canges in te sample points, or in yperparameters, will affect tese estimated values. 3.2 Density Derivative Estimation In tis subsection te estimation of te first derivative of a density is considered. Te same collection of densities used in density estimation is considered, except tat density (b) is omitted since it is not differentiable at its boundaries. Of te available implementations for te metods considered, only te exact estimation and te truncated Taylor ansion for te Gaussian kernel were available. Only estimation at te sample points was considered, since all available metods are capable of tis task. Te average integrated squared error accuracy is reported in Table 4. Once again te kernel K 1 sows te lowest error most often, 16

17 Table 2: Average integrated squared error of density estimates from 30 replications. Sets of and observations are considered. Evaluation is conducted for te entire collection of sample points in eac case. Lowest average error for eac scenario is igligted in bold. Apparent ties were broken by considering more significant figures. Density (a) (b) (c) (d) (e) (f) (g) () Exact Gauss n = 1e e e e e e e e e-01 Trunc. Taylor Gauss n = 1e e e e e e e e e-01 n = 1e e e e e e e e e-02 Exact K 1 n = 1e e e e e e e e e-02 n = 1e e e e e e e e e-02 Exact K 4 n = 1e e e e e e e e e-01 n = 1e e e e e e e e e-02 Table 3: Average running time of density estimation from 30 replications. Sets of and observations are considered. Evaluation is conducted for te entire collection of sample points in eac case. Lowest average computation time for eac scenario is igligted in bold. Density (a) (b) (c) (d) (e) (f) (g) () Exact Gauss n = 1e e e e e e e e e-01 Trunc. Taylor Gauss n = 1e e e e e e e e e-03 n = 1e e e e e e e e e-01 Exact K 1 n = 1e+03 7e-04 9e e-04 7e-04 9e-04 9e-04 9e e-04 n = 1e e e e e e e e e-02 Exact K 4 n = 1e e e e e e e e e-03 n = 1e e e e e e e e e-01 17

18 Table 4: Average integrated squared error of first derivative estimates from 30 replications.. Sets of and observations are considered. Evaluation is conducted for te entire collection of sample points in eac case. Lowest average for eac scenario is igligted in bold. Density (a) (c) (d) (e) (f) (g) () Exact Gauss n = 1e e e e e e e e+00 Trunc. Taylor Gauss n = 1e e e e e e e e+00 n = 1e e e e e e e e+00 Exact K 1 n = 1e e e e e e e e+00 n = 1e e e e e e e e+00 Exact K 4 n = 1e e e e e e e+00 1e+01 n = 1e e e e e e e e+00 Exact K 7 n = 1e e e e e e e e+01 n = 1e e e e e e e e+00 owever in tis case only wen te densities ave very sarp features. Te performance of te lower efficiency kernel is sligtly worse on densities (a) and (d), for wic te euristic used for bandwidt selection is closer to optimal based on te AMISE objective (in te case of density (a) it is exactly optimal). In tese cases te error of kernel K 7 is lowest. Te relative computational efficiency of te proposed approac is even more apparent in te task of density derivative estimation. Table 5 reports te average running times on all densities considered. Here it can be seen tat te evaluation of te pointwise derivative at te sample points wen using kernel K 1 is an order of magnitude faster tan wen using te truncated Taylor ansion. Evaluation wit te kernel K 4 is rougly tree times faster tan te approximate metod wit te current implementations, and te running time wit kernel K 7 is similar to te approximate approac. 4 Discussion and A Brief Comment on Kernel Coice In tis work a ric class of kernels was introduced wose members allow for extremely efficient and exact evaluation of kernel density and density derivative estimates. A muc smaller sub-class was investigated more deeply. Kernels in tis sub-class were selected for 18

19 Table 5: Average running time of estimation of first derivative from 30 replications. Sets of and observations are considered. Evaluation is conducted for te entire collection of sample points in eac case. Lowest average computation time for eac scenario is igligted in bold. Density (a) (c) (d) (e) (f) (g) () Exact Gauss n = 1e e e e e e e e-01 Trunc. Taylor Gauss n = 1e e-03 6e-03 5e e e e e-03 n = 1e e e e e e e e-01 Exact K 1 n = 1e e e e e-04 1e e-04 1e-03 n = 1e e e e e e e e-02 Exact K 4 n = 1e e e e e-03 2e e-03 2e-03 n = 1e e e e e e e e-01 Exact K 7 n = 1e e e e e e e e-03 n = 1e e e e e e e e-01 teir simplicity of ression and te fact tat tey admit a large number of derivatives relative to tis simplicity. Toroug erimentation wit kernels from tis sub-class was conducted sowing extremely promising performance in terms of accuracy and empirical running time. It is important to note tat te efficiency of a kernel for a given task relates to te AMISE error wic it induces, but under te assumption tat te corresponding optimal bandwidt parameter is also selected. Te popular euristic for bandwidt selection wic was used erein tends to over-estimate te AMISE optimal value wen te underlying density as sarp features and ig curvature. Wit tis euristic tere is strong evidence tat kernel K 1 represents an excellent coice for its fast computation and its accurate density estimation. On te oter and, if a more sopisticated metod is employed to select a bandwidt parameter closer to te AMISE optimal, ten K 4 is recommended for its very similar error to te popular Gaussian kernel and its comparatively fast computation. An interesting direction for future researc will be in te design of kernels in te broader class introduced erein wic ave simple ressions (in te sense tat te polynomial component as a low degree), and wic ave ig relative efficiency for estimation of a specific derivative of te density wic is of relevance for a given task. 19

20 References Eddelbuettel, D. and R. François (2011). Rcpp: Seamless R and C++ integration. Journal of Statistical Software 40 (8), Fan, J. and J. S. Marron (1994). Fast implementations of nonparametric curve estimators. Journal of computational and grapical statistics 3 (1), Guidoum, A. (2015). kedd: Kernel estimator and bandwidt selection for density and its derivatives. R package version Hall, P. and M. P. Wand (1994). On te accuracy of binned kernel density estimators. Huber, P. J. (1985). Projection pursuit. Te annals of Statistics, Marron, J. S. and M. P. Wand (1992). Exact mean integrated squared error. Te Annals of Statistics, PW, R. (1976). On te coice of smooting parameters for parzen estimators of probability density functions. IEEE Transactions on Computers. Raykar, V. C., R. Duraiswami, and L. H. Zao (2010). Fast computation of kernel estimators. Journal of Computational and Grapical Statistics 19 (1), Scott, D. W. and S. J. Seater (1985). Kernel density estimation wit binned data. Communications in Statistics-Teory and Metods 14 (6), Scott, D. W. and G. R. Terrell (1987). Biased and unbiased cross-validation in density estimation. Journal of te american Statistical association 82 (400), Seater, S. J. and M. C. Jones (1991). A reliable data-based bandwidt selection metod for kernel density estimation. Journal of te Royal Statistical Society. Series B (Metodological), Silverman, B. (1982). Algoritm as 176: Kernel density estimation using te fast fourier transform. Journal of te Royal Statistical Society. Series C (Applied Statistics) 31 (1),

21 Silverman, B. W. (2018). Density estimation for statistics and data analysis. Routledge. Wand, M. (2015). KernSmoot: Functions for Kernel Smooting Supporting Wand & Jones. R package version Yang, C., R. Duraiswami, N. A. Gumerov, L. S. Davis, et al. (2003). Improved fast gauss transform and efficient kernel density estimation. In ICCV, Volume 1, pp

Chapter 1. Density Estimation

Chapter 1. Density Estimation Capter 1 Density Estimation Let X 1, X,..., X n be observations from a density f X x. Te aim is to use only tis data to obtain an estimate ˆf X x of f X x. Properties of f f X x x, Parametric metods f

More information

Fast optimal bandwidth selection for kernel density estimation

Fast optimal bandwidth selection for kernel density estimation Fast optimal bandwidt selection for kernel density estimation Vikas Candrakant Raykar and Ramani Duraiswami Dept of computer science and UMIACS, University of Maryland, CollegePark {vikas,ramani}@csumdedu

More information

The Priestley-Chao Estimator

The Priestley-Chao Estimator Te Priestley-Cao Estimator In tis section we will consider te Pristley-Cao estimator of te unknown regression function. It is assumed tat we ave a sample of observations (Y i, x i ), i = 1,..., n wic are

More information

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx.

Consider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx. Capter 2 Integrals as sums and derivatives as differences We now switc to te simplest metods for integrating or differentiating a function from its function samples. A careful study of Taylor expansions

More information

Boosting Kernel Density Estimates: a Bias Reduction. Technique?

Boosting Kernel Density Estimates: a Bias Reduction. Technique? Boosting Kernel Density Estimates: a Bias Reduction Tecnique? Marco Di Marzio Dipartimento di Metodi Quantitativi e Teoria Economica, Università di Cieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy

More information

Kernel Density Estimation

Kernel Density Estimation Kernel Density Estimation Univariate Density Estimation Suppose tat we ave a random sample of data X 1,..., X n from an unknown continuous distribution wit probability density function (pdf) f(x) and cumulative

More information

Numerical Differentiation

Numerical Differentiation Numerical Differentiation Finite Difference Formulas for te first derivative (Using Taylor Expansion tecnique) (section 8.3.) Suppose tat f() = g() is a function of te variable, and tat as 0 te function

More information

A = h w (1) Error Analysis Physics 141

A = h w (1) Error Analysis Physics 141 Introduction In all brances of pysical science and engineering one deals constantly wit numbers wic results more or less directly from experimental observations. Experimental observations always ave inaccuracies.

More information

Order of Accuracy. ũ h u Ch p, (1)

Order of Accuracy. ũ h u Ch p, (1) Order of Accuracy 1 Terminology We consider a numerical approximation of an exact value u. Te approximation depends on a small parameter, wic can be for instance te grid size or time step in a numerical

More information

Kernel Density Based Linear Regression Estimate

Kernel Density Based Linear Regression Estimate Kernel Density Based Linear Regression Estimate Weixin Yao and Zibiao Zao Abstract For linear regression models wit non-normally distributed errors, te least squares estimate (LSE will lose some efficiency

More information

LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION

LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION LIMITATIONS OF EULER S METHOD FOR NUMERICAL INTEGRATION LAURA EVANS.. Introduction Not all differential equations can be explicitly solved for y. Tis can be problematic if we need to know te value of y

More information

Copyright c 2008 Kevin Long

Copyright c 2008 Kevin Long Lecture 4 Numerical solution of initial value problems Te metods you ve learned so far ave obtained closed-form solutions to initial value problems. A closedform solution is an explicit algebriac formula

More information

lecture 26: Richardson extrapolation

lecture 26: Richardson extrapolation 43 lecture 26: Ricardson extrapolation 35 Ricardson extrapolation, Romberg integration Trougout numerical analysis, one encounters procedures tat apply some simple approximation (eg, linear interpolation)

More information

Chapter 5 FINITE DIFFERENCE METHOD (FDM)

Chapter 5 FINITE DIFFERENCE METHOD (FDM) MEE7 Computer Modeling Tecniques in Engineering Capter 5 FINITE DIFFERENCE METHOD (FDM) 5. Introduction to FDM Te finite difference tecniques are based upon approximations wic permit replacing differential

More information

IEOR 165 Lecture 10 Distribution Estimation

IEOR 165 Lecture 10 Distribution Estimation IEOR 165 Lecture 10 Distribution Estimation 1 Motivating Problem Consider a situation were we ave iid data x i from some unknown distribution. One problem of interest is estimating te distribution tat

More information

UNIMODAL KERNEL DENSITY ESTIMATION BY DATA SHARPENING

UNIMODAL KERNEL DENSITY ESTIMATION BY DATA SHARPENING Statistica Sinica 15(2005), 73-98 UNIMODAL KERNEL DENSITY ESTIMATION BY DATA SHARPENING Peter Hall 1 and Kee-Hoon Kang 1,2 1 Australian National University and 2 Hankuk University of Foreign Studies Abstract:

More information

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines

Lecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines Lecture 5 Interpolation II Introduction In te previous lecture we focused primarily on polynomial interpolation of a set of n points. A difficulty we observed is tat wen n is large, our polynomial as to

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximatinga function fx, wose values at a set of distinct points x, x, x,, x n are known, by a polynomial P x suc

More information

Very fast optimal bandwidth selection for univariate kernel density estimation

Very fast optimal bandwidth selection for univariate kernel density estimation Very fast optimal bandwidt selection for univariate kernel density estimation VIKAS CHANDAKANT AYKA and AMANI DUAISWAMI Perceptual Interfaces and eality Laboratory Department of Computer Science and Institute

More information

Bootstrap confidence intervals in nonparametric regression without an additive model

Bootstrap confidence intervals in nonparametric regression without an additive model Bootstrap confidence intervals in nonparametric regression witout an additive model Dimitris N. Politis Abstract Te problem of confidence interval construction in nonparametric regression via te bootstrap

More information

Local Orthogonal Polynomial Expansion (LOrPE) for Density Estimation

Local Orthogonal Polynomial Expansion (LOrPE) for Density Estimation Local Ortogonal Polynomial Expansion (LOrPE) for Density Estimation Alex Trindade Dept. of Matematics & Statistics, Texas Tec University Igor Volobouev, Texas Tec University (Pysics Dept.) D.P. Amali Dassanayake,

More information

Regularized Regression

Regularized Regression Regularized Regression David M. Blei Columbia University December 5, 205 Modern regression problems are ig dimensional, wic means tat te number of covariates p is large. In practice statisticians regularize

More information

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY

SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY (Section 3.2: Derivative Functions and Differentiability) 3.2.1 SECTION 3.2: DERIVATIVE FUNCTIONS and DIFFERENTIABILITY LEARNING OBJECTIVES Know, understand, and apply te Limit Definition of te Derivative

More information

A Locally Adaptive Transformation Method of Boundary Correction in Kernel Density Estimation

A Locally Adaptive Transformation Method of Boundary Correction in Kernel Density Estimation A Locally Adaptive Transformation Metod of Boundary Correction in Kernel Density Estimation R.J. Karunamuni a and T. Alberts b a Department of Matematical and Statistical Sciences University of Alberta,

More information

On Local Linear Regression Estimation of Finite Population Totals in Model Based Surveys

On Local Linear Regression Estimation of Finite Population Totals in Model Based Surveys American Journal of Teoretical and Applied Statistics 2018; 7(3): 92-101 ttp://www.sciencepublisinggroup.com/j/ajtas doi: 10.11648/j.ajtas.20180703.11 ISSN: 2326-8999 (Print); ISSN: 2326-9006 (Online)

More information

Polynomial Interpolation

Polynomial Interpolation Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximating a function f(x, wose values at a set of distinct points x, x, x 2,,x n are known, by a polynomial P (x

More information

New Streamfunction Approach for Magnetohydrodynamics

New Streamfunction Approach for Magnetohydrodynamics New Streamfunction Approac for Magnetoydrodynamics Kab Seo Kang Brooaven National Laboratory, Computational Science Center, Building 63, Room, Upton NY 973, USA. sang@bnl.gov Summary. We apply te finite

More information

Artificial Neural Network Model Based Estimation of Finite Population Total

Artificial Neural Network Model Based Estimation of Finite Population Total International Journal of Science and Researc (IJSR), India Online ISSN: 2319-7064 Artificial Neural Network Model Based Estimation of Finite Population Total Robert Kasisi 1, Romanus O. Odiambo 2, Antony

More information

The Laplace equation, cylindrically or spherically symmetric case

The Laplace equation, cylindrically or spherically symmetric case Numerisce Metoden II, 7 4, und Übungen, 7 5 Course Notes, Summer Term 7 Some material and exercises Te Laplace equation, cylindrically or sperically symmetric case Electric and gravitational potential,

More information

Te comparison of dierent models M i is based on teir relative probabilities, wic can be expressed, again using Bayes' teorem, in terms of prior probab

Te comparison of dierent models M i is based on teir relative probabilities, wic can be expressed, again using Bayes' teorem, in terms of prior probab To appear in: Advances in Neural Information Processing Systems 9, eds. M. C. Mozer, M. I. Jordan and T. Petsce. MIT Press, 997 Bayesian Model Comparison by Monte Carlo Caining David Barber D.Barber@aston.ac.uk

More information

Introduction to Derivatives

Introduction to Derivatives Introduction to Derivatives 5-Minute Review: Instantaneous Rates and Tangent Slope Recall te analogy tat we developed earlier First we saw tat te secant slope of te line troug te two points (a, f (a))

More information

A Jump-Preserving Curve Fitting Procedure Based On Local Piecewise-Linear Kernel Estimation

A Jump-Preserving Curve Fitting Procedure Based On Local Piecewise-Linear Kernel Estimation A Jump-Preserving Curve Fitting Procedure Based On Local Piecewise-Linear Kernel Estimation Peiua Qiu Scool of Statistics University of Minnesota 313 Ford Hall 224 Curc St SE Minneapolis, MN 55455 Abstract

More information

Applications of the van Trees inequality to non-parametric estimation.

Applications of the van Trees inequality to non-parametric estimation. Brno-06, Lecture 2, 16.05.06 D/Stat/Brno-06/2.tex www.mast.queensu.ca/ blevit/ Applications of te van Trees inequality to non-parametric estimation. Regular non-parametric problems. As an example of suc

More information

Basic Nonparametric Estimation Spring 2002

Basic Nonparametric Estimation Spring 2002 Basic Nonparametric Estimation Spring 2002 Te following topics are covered today: Basic Nonparametric Regression. Tere are four books tat you can find reference: Silverman986, Wand and Jones995, Hardle990,

More information

Optimal parameters for a hierarchical grid data structure for contact detection in arbitrarily polydisperse particle systems

Optimal parameters for a hierarchical grid data structure for contact detection in arbitrarily polydisperse particle systems Comp. Part. Mec. 04) :357 37 DOI 0.007/s4057-04-000-9 Optimal parameters for a ierarcical grid data structure for contact detection in arbitrarily polydisperse particle systems Dinant Krijgsman Vitaliy

More information

MATH745 Fall MATH745 Fall

MATH745 Fall MATH745 Fall MATH745 Fall 5 MATH745 Fall 5 INTRODUCTION WELCOME TO MATH 745 TOPICS IN NUMERICAL ANALYSIS Instructor: Dr Bartosz Protas Department of Matematics & Statistics Email: bprotas@mcmasterca Office HH 36, Ext

More information

How to Find the Derivative of a Function: Calculus 1

How to Find the Derivative of a Function: Calculus 1 Introduction How to Find te Derivative of a Function: Calculus 1 Calculus is not an easy matematics course Te fact tat you ave enrolled in suc a difficult subject indicates tat you are interested in te

More information

Bootstrap prediction intervals for Markov processes

Bootstrap prediction intervals for Markov processes arxiv: arxiv:0000.0000 Bootstrap prediction intervals for Markov processes Li Pan and Dimitris N. Politis Li Pan Department of Matematics University of California San Diego La Jolla, CA 92093-0112, USA

More information

7 Semiparametric Methods and Partially Linear Regression

7 Semiparametric Methods and Partially Linear Regression 7 Semiparametric Metods and Partially Linear Regression 7. Overview A model is called semiparametric if it is described by and were is nite-dimensional (e.g. parametric) and is in nite-dimensional (nonparametric).

More information

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES

A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES A MONTE CARLO ANALYSIS OF THE EFFECTS OF COVARIANCE ON PROPAGATED UNCERTAINTIES Ronald Ainswort Hart Scientific, American Fork UT, USA ABSTRACT Reports of calibration typically provide total combined uncertainties

More information

INFINITE ORDER CROSS-VALIDATED LOCAL POLYNOMIAL REGRESSION. 1. Introduction

INFINITE ORDER CROSS-VALIDATED LOCAL POLYNOMIAL REGRESSION. 1. Introduction INFINITE ORDER CROSS-VALIDATED LOCAL POLYNOMIAL REGRESSION PETER G. HALL AND JEFFREY S. RACINE Abstract. Many practical problems require nonparametric estimates of regression functions, and local polynomial

More information

Combining functions: algebraic methods

Combining functions: algebraic methods Combining functions: algebraic metods Functions can be added, subtracted, multiplied, divided, and raised to a power, just like numbers or algebra expressions. If f(x) = x 2 and g(x) = x + 2, clearly f(x)

More information

The derivative function

The derivative function Roberto s Notes on Differential Calculus Capter : Definition of derivative Section Te derivative function Wat you need to know already: f is at a point on its grap and ow to compute it. Wat te derivative

More information

Math 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006

Math 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006 Mat 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006 f(x+) f(x) 10 1. For f(x) = x 2 + 2x 5, find ))))))))) and simplify completely. NOTE: **f(x+) is NOT f(x)+! f(x+) f(x) (x+) 2 + 2(x+) 5 ( x 2

More information

ERROR BOUNDS FOR THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BRADLEY J. LUCIER*

ERROR BOUNDS FOR THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BRADLEY J. LUCIER* EO BOUNDS FO THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BADLEY J. LUCIE* Abstract. Te expected error in L ) attimet for Glimm s sceme wen applied to a scalar conservation law is bounded by + 2 ) ) /2 T

More information

ch (for some fixed positive number c) reaching c

ch (for some fixed positive number c) reaching c GSTF Journal of Matematics Statistics and Operations Researc (JMSOR) Vol. No. September 05 DOI 0.60/s4086-05-000-z Nonlinear Piecewise-defined Difference Equations wit Reciprocal and Cubic Terms Ramadan

More information

The Verlet Algorithm for Molecular Dynamics Simulations

The Verlet Algorithm for Molecular Dynamics Simulations Cemistry 380.37 Fall 2015 Dr. Jean M. Standard November 9, 2015 Te Verlet Algoritm for Molecular Dynamics Simulations Equations of motion For a many-body system consisting of N particles, Newton's classical

More information

Average Rate of Change

Average Rate of Change Te Derivative Tis can be tougt of as an attempt to draw a parallel (pysically and metaporically) between a line and a curve, applying te concept of slope to someting tat isn't actually straigt. Te slope

More information

Derivatives. By: OpenStaxCollege

Derivatives. By: OpenStaxCollege By: OpenStaxCollege Te average teen in te United States opens a refrigerator door an estimated 25 times per day. Supposedly, tis average is up from 10 years ago wen te average teenager opened a refrigerator

More information

Parameter Fitted Scheme for Singularly Perturbed Delay Differential Equations

Parameter Fitted Scheme for Singularly Perturbed Delay Differential Equations International Journal of Applied Science and Engineering 2013. 11, 4: 361-373 Parameter Fitted Sceme for Singularly Perturbed Delay Differential Equations Awoke Andargiea* and Y. N. Reddyb a b Department

More information

Lecture 21. Numerical differentiation. f ( x+h) f ( x) h h

Lecture 21. Numerical differentiation. f ( x+h) f ( x) h h Lecture Numerical differentiation Introduction We can analytically calculate te derivative of any elementary function, so tere migt seem to be no motivation for calculating derivatives numerically. However

More information

HOMEWORK HELP 2 FOR MATH 151

HOMEWORK HELP 2 FOR MATH 151 HOMEWORK HELP 2 FOR MATH 151 Here we go; te second round of omework elp. If tere are oters you would like to see, let me know! 2.4, 43 and 44 At wat points are te functions f(x) and g(x) = xf(x)continuous,

More information

Data-Based Optimal Bandwidth for Kernel Density Estimation of Statistical Samples

Data-Based Optimal Bandwidth for Kernel Density Estimation of Statistical Samples Commun. Teor. Pys. 70 (208) 728 734 Vol. 70 No. 6 December 208 Data-Based Optimal Bandwidt for Kernel Density Estimation of Statistical Samples Zen-Wei Li ( 李振伟 ) 2 and Ping He ( 何平 ) 3 Center for Teoretical

More information

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these.

4. The slope of the line 2x 7y = 8 is (a) 2/7 (b) 7/2 (c) 2 (d) 2/7 (e) None of these. Mat 11. Test Form N Fall 016 Name. Instructions. Te first eleven problems are wort points eac. Te last six problems are wort 5 points eac. For te last six problems, you must use relevant metods of algebra

More information

158 Calculus and Structures

158 Calculus and Structures 58 Calculus and Structures CHAPTER PROPERTIES OF DERIVATIVES AND DIFFERENTIATION BY THE EASY WAY. Calculus and Structures 59 Copyrigt Capter PROPERTIES OF DERIVATIVES. INTRODUCTION In te last capter you

More information

LIMITS AND DERIVATIVES CONDITIONS FOR THE EXISTENCE OF A LIMIT

LIMITS AND DERIVATIVES CONDITIONS FOR THE EXISTENCE OF A LIMIT LIMITS AND DERIVATIVES Te limit of a function is defined as te value of y tat te curve approaces, as x approaces a particular value. Te limit of f (x) as x approaces a is written as f (x) approaces, as

More information

Differential Calculus (The basics) Prepared by Mr. C. Hull

Differential Calculus (The basics) Prepared by Mr. C. Hull Differential Calculus Te basics) A : Limits In tis work on limits, we will deal only wit functions i.e. tose relationsips in wic an input variable ) defines a unique output variable y). Wen we work wit

More information

CDF and Survival Function Estimation with Infinite-Order Kernels

CDF and Survival Function Estimation with Infinite-Order Kernels CDF and Survival Function Estimation wit Infinite-Order Kernels Artur Berg and Dimitris N. Politis Abstract An improved nonparametric estimator of te cumulative distribution function CDF) and te survival

More information

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative

Mathematics 5 Worksheet 11 Geometry, Tangency, and the Derivative Matematics 5 Workseet 11 Geometry, Tangency, and te Derivative Problem 1. Find te equation of a line wit slope m tat intersects te point (3, 9). Solution. Te equation for a line passing troug a point (x

More information

Differentiation in higher dimensions

Differentiation in higher dimensions Capter 2 Differentiation in iger dimensions 2.1 Te Total Derivative Recall tat if f : R R is a 1-variable function, and a R, we say tat f is differentiable at x = a if and only if te ratio f(a+) f(a) tends

More information

Differentiation. Area of study Unit 2 Calculus

Differentiation. Area of study Unit 2 Calculus Differentiation 8VCE VCEco Area of stud Unit Calculus coverage In tis ca 8A 8B 8C 8D 8E 8F capter Introduction to limits Limits of discontinuous, rational and brid functions Differentiation using first

More information

Physically Based Modeling: Principles and Practice Implicit Methods for Differential Equations

Physically Based Modeling: Principles and Practice Implicit Methods for Differential Equations Pysically Based Modeling: Principles and Practice Implicit Metods for Differential Equations David Baraff Robotics Institute Carnegie Mellon University Please note: Tis document is 997 by David Baraff

More information

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES

SECTION 1.10: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES (Section.0: Difference Quotients).0. SECTION.0: DIFFERENCE QUOTIENTS LEARNING OBJECTIVES Define average rate of cange (and average velocity) algebraically and grapically. Be able to identify, construct,

More information

Time (hours) Morphine sulfate (mg)

Time (hours) Morphine sulfate (mg) Mat Xa Fall 2002 Review Notes Limits and Definition of Derivative Important Information: 1 According to te most recent information from te Registrar, te Xa final exam will be eld from 9:15 am to 12:15

More information

EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING

EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING Statistica Sinica 13(2003), 641-653 EFFICIENT REPLICATION VARIANCE ESTIMATION FOR TWO-PHASE SAMPLING J. K. Kim and R. R. Sitter Hankuk University of Foreign Studies and Simon Fraser University Abstract:

More information

MATH1151 Calculus Test S1 v2a

MATH1151 Calculus Test S1 v2a MATH5 Calculus Test 8 S va January 8, 5 Tese solutions were written and typed up by Brendan Trin Please be etical wit tis resource It is for te use of MatSOC members, so do not repost it on oter forums

More information

CS522 - Partial Di erential Equations

CS522 - Partial Di erential Equations CS5 - Partial Di erential Equations Tibor Jánosi April 5, 5 Numerical Di erentiation In principle, di erentiation is a simple operation. Indeed, given a function speci ed as a closed-form formula, its

More information

These errors are made from replacing an infinite process by finite one.

These errors are made from replacing an infinite process by finite one. Introduction :- Tis course examines problems tat can be solved by metods of approximation, tecniques we call numerical metods. We begin by considering some of te matematical and computational topics tat

More information

2.8 The Derivative as a Function

2.8 The Derivative as a Function .8 Te Derivative as a Function Typically, we can find te derivative of a function f at many points of its domain: Definition. Suppose tat f is a function wic is differentiable at every point of an open

More information

Section 3: The Derivative Definition of the Derivative

Section 3: The Derivative Definition of the Derivative Capter 2 Te Derivative Business Calculus 85 Section 3: Te Derivative Definition of te Derivative Returning to te tangent slope problem from te first section, let's look at te problem of finding te slope

More information

NUMERICAL DIFFERENTIATION. James T. Smith San Francisco State University. In calculus classes, you compute derivatives algebraically: for example,

NUMERICAL DIFFERENTIATION. James T. Smith San Francisco State University. In calculus classes, you compute derivatives algebraically: for example, NUMERICAL DIFFERENTIATION James T Smit San Francisco State University In calculus classes, you compute derivatives algebraically: for example, f( x) = x + x f ( x) = x x Tis tecnique requires your knowing

More information

Kernel Smoothing and Tolerance Intervals for Hierarchical Data

Kernel Smoothing and Tolerance Intervals for Hierarchical Data Clemson University TigerPrints All Dissertations Dissertations 12-2016 Kernel Smooting and Tolerance Intervals for Hierarcical Data Cristoper Wilson Clemson University, cwilso6@clemson.edu Follow tis and

More information

Handling Missing Data on Asymmetric Distribution

Handling Missing Data on Asymmetric Distribution International Matematical Forum, Vol. 8, 03, no. 4, 53-65 Handling Missing Data on Asymmetric Distribution Amad M. H. Al-Kazale Department of Matematics, Faculty of Science Al-albayt University, Al-Mafraq-Jordan

More information

Continuity and Differentiability

Continuity and Differentiability Continuity and Dierentiability Tis capter requires a good understanding o its. Te concepts o continuity and dierentiability are more or less obvious etensions o te concept o its. Section - INTRODUCTION

More information

Taylor Series and the Mean Value Theorem of Derivatives

Taylor Series and the Mean Value Theorem of Derivatives 1 - Taylor Series and te Mean Value Teorem o Derivatives Te numerical solution o engineering and scientiic problems described by matematical models oten requires solving dierential equations. Dierential

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL IFFERENTIATION FIRST ERIVATIVES Te simplest difference formulas are based on using a straigt line to interpolate te given data; tey use two data pints to estimate te derivative. We assume tat

More information

1 Introduction to Optimization

1 Introduction to Optimization Unconstrained Convex Optimization 2 1 Introduction to Optimization Given a general optimization problem of te form min x f(x) (1.1) were f : R n R. Sometimes te problem as constraints (we are only interested

More information

Deconvolution problems in density estimation

Deconvolution problems in density estimation Deconvolution problems in density estimation Dissertation zur Erlangung des Doktorgrades Dr. rer. nat. der Fakultät für Matematik und Wirtscaftswissenscaften der Universität Ulm vorgelegt von Cristian

More information

LECTURE 14 NUMERICAL INTEGRATION. Find

LECTURE 14 NUMERICAL INTEGRATION. Find LECTURE 14 NUMERCAL NTEGRATON Find b a fxdx or b a vx ux fx ydy dx Often integration is required. However te form of fx may be suc tat analytical integration would be very difficult or impossible. Use

More information

THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS. L. Trautmann, R. Rabenstein

THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS. L. Trautmann, R. Rabenstein Worksop on Transforms and Filter Banks (WTFB),Brandenburg, Germany, Marc 999 THE STURM-LIOUVILLE-TRANSFORMATION FOR THE SOLUTION OF VECTOR PARTIAL DIFFERENTIAL EQUATIONS L. Trautmann, R. Rabenstein Lerstul

More information

Dedicated to the 70th birthday of Professor Lin Qun

Dedicated to the 70th birthday of Professor Lin Qun Journal of Computational Matematics, Vol.4, No.3, 6, 4 44. ACCELERATION METHODS OF NONLINEAR ITERATION FOR NONLINEAR PARABOLIC EQUATIONS Guang-wei Yuan Xu-deng Hang Laboratory of Computational Pysics,

More information

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS

HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS HOW TO DEAL WITH FFT SAMPLING INFLUENCES ON ADEV CALCULATIONS Po-Ceng Cang National Standard Time & Frequency Lab., TL, Taiwan 1, Lane 551, Min-Tsu Road, Sec. 5, Yang-Mei, Taoyuan, Taiwan 36 Tel: 886 3

More information

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura

NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL. Georgeta Budura NONLINEAR SYSTEMS IDENTIFICATION USING THE VOLTERRA MODEL Georgeta Budura Politenica University of Timisoara, Faculty of Electronics and Telecommunications, Comm. Dep., georgeta.budura@etc.utt.ro Abstract:

More information

3.1 Extreme Values of a Function

3.1 Extreme Values of a Function .1 Etreme Values of a Function Section.1 Notes Page 1 One application of te derivative is finding minimum and maimum values off a grap. In precalculus we were only able to do tis wit quadratics by find

More information

Integral Calculus, dealing with areas and volumes, and approximate areas under and between curves.

Integral Calculus, dealing with areas and volumes, and approximate areas under and between curves. Calculus can be divided into two ke areas: Differential Calculus dealing wit its, rates of cange, tangents and normals to curves, curve sketcing, and applications to maima and minima problems Integral

More information

= 0 and states ''hence there is a stationary point'' All aspects of the proof dx must be correct (c)

= 0 and states ''hence there is a stationary point'' All aspects of the proof dx must be correct (c) Paper 1: Pure Matematics 1 Mark Sceme 1(a) (i) (ii) d d y 3 1x 4x x M1 A1 d y dx 1.1b 1.1b 36x 48x A1ft 1.1b Substitutes x = into teir dx (3) 3 1 4 Sows d y 0 and states ''ence tere is a stationary point''

More information

Click here to see an animation of the derivative

Click here to see an animation of the derivative Differentiation Massoud Malek Derivative Te concept of derivative is at te core of Calculus; It is a very powerful tool for understanding te beavior of matematical functions. It allows us to optimize functions,

More information

Bounds on the Moments for an Ensemble of Random Decision Trees

Bounds on the Moments for an Ensemble of Random Decision Trees Noname manuscript No. (will be inserted by te editor) Bounds on te Moments for an Ensemble of Random Decision Trees Amit Durandar Received: Sep. 17, 2013 / Revised: Mar. 04, 2014 / Accepted: Jun. 30, 2014

More information

Precalculus Test 2 Practice Questions Page 1. Note: You can expect other types of questions on the test than the ones presented here!

Precalculus Test 2 Practice Questions Page 1. Note: You can expect other types of questions on the test than the ones presented here! Precalculus Test 2 Practice Questions Page Note: You can expect oter types of questions on te test tan te ones presented ere! Questions Example. Find te vertex of te quadratic f(x) = 4x 2 x. Example 2.

More information

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems

5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems 5 Ordinary Differential Equations: Finite Difference Metods for Boundary Problems Read sections 10.1, 10.2, 10.4 Review questions 10.1 10.4, 10.8 10.9, 10.13 5.1 Introduction In te previous capters we

More information

New Distribution Theory for the Estimation of Structural Break Point in Mean

New Distribution Theory for the Estimation of Structural Break Point in Mean New Distribution Teory for te Estimation of Structural Break Point in Mean Liang Jiang Singapore Management University Xiaou Wang Te Cinese University of Hong Kong Jun Yu Singapore Management University

More information

NADARAYA WATSON ESTIMATE JAN 10, 2006: version 2. Y ik ( x i

NADARAYA WATSON ESTIMATE JAN 10, 2006: version 2. Y ik ( x i NADARAYA WATSON ESTIMATE JAN 0, 2006: version 2 DATA: (x i, Y i, i =,..., n. ESTIMATE E(Y x = m(x by n i= ˆm (x = Y ik ( x i x n i= K ( x i x EXAMPLES OF K: K(u = I{ u c} (uniform or box kernel K(u = u

More information

Discriminate Modelling of Peak and Off-Peak Motorway Capacity

Discriminate Modelling of Peak and Off-Peak Motorway Capacity International Journal of Integrated Engineering - Special Issue on ICONCEES Vol. 4 No. 3 (2012) p. 53-58 Discriminate Modelling of Peak and Off-Peak Motorway Capacity Hasim Moammed Alassan 1,*, Sundara

More information

Polynomials 3: Powers of x 0 + h

Polynomials 3: Powers of x 0 + h near small binomial Capter 17 Polynomials 3: Powers of + Wile it is easy to compute wit powers of a counting-numerator, it is a lot more difficult to compute wit powers of a decimal-numerator. EXAMPLE

More information

A h u h = f h. 4.1 The CoarseGrid SystemandtheResidual Equation

A h u h = f h. 4.1 The CoarseGrid SystemandtheResidual Equation Capter Grid Transfer Remark. Contents of tis capter. Consider a grid wit grid size and te corresponding linear system of equations A u = f. Te summary given in Section 3. leads to te idea tat tere migt

More information

4.2 - Richardson Extrapolation

4.2 - Richardson Extrapolation . - Ricardson Extrapolation. Small-O Notation: Recall tat te big-o notation used to define te rate of convergence in Section.: Definition Let x n n converge to a number x. Suppose tat n n is a sequence

More information

2.1 THE DEFINITION OF DERIVATIVE

2.1 THE DEFINITION OF DERIVATIVE 2.1 Te Derivative Contemporary Calculus 2.1 THE DEFINITION OF DERIVATIVE 1 Te grapical idea of a slope of a tangent line is very useful, but for some uses we need a more algebraic definition of te derivative

More information

Quantum Theory of the Atomic Nucleus

Quantum Theory of the Atomic Nucleus G. Gamow, ZP, 51, 204 1928 Quantum Teory of te tomic Nucleus G. Gamow (Received 1928) It as often been suggested tat non Coulomb attractive forces play a very important role inside atomic nuclei. We can

More information

GRID CONVERGENCE ERROR ANALYSIS FOR MIXED-ORDER NUMERICAL SCHEMES

GRID CONVERGENCE ERROR ANALYSIS FOR MIXED-ORDER NUMERICAL SCHEMES GRID CONVERGENCE ERROR ANALYSIS FOR MIXED-ORDER NUMERICAL SCHEMES Cristoper J. Roy Sandia National Laboratories* P. O. Box 5800, MS 085 Albuquerque, NM 8785-085 AIAA Paper 00-606 Abstract New developments

More information

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point

1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point MA00 Capter 6 Calculus and Basic Linear Algebra I Limits, Continuity and Differentiability Te concept of its (p.7 p.9, p.4 p.49, p.55 p.56). Limits Consider te function determined by te formula f Note

More information