DOCSLIB.ORG

Direct Log-Density Gradient Estimation with Gaussian Mixture Models and Its Application to Clustering

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Direct Log-Density Gradient Estimation with Gaussian Mixture Models and Its Application to Clustering IEICE TRANS. INF. & SYST., VOL.E102–D, NO.6 JUNE 2019 1154 PAPER Direct Log-Density Gradient Estimation with Gaussian Mixture Models and Its Application to Clustering Qi ZHANG†a), Hiroaki SASAKI††b), Nonmembers, and Kazushi IKEDA††c), Senior Member SUMMARY Estimation of the gradient of the logarithm of a probabil- detected modes. Therefore, MS has been applied to a va- ity density function is a versatile tool in statistical data analysis. A recent riety of problems such as image segmentation [6], [8], [9] method for model-seeking clustering called the least-squares log-density and object tracking [10], [11] (See also a recent review ar- gradient clustering (LSLDGC) [Sasaki et al., 2014] employs a sophisti- cated gradient estimator, which directly estimates the log-density gradients ticle [12]). Furthermore, MS and its related methods have without going through density estimation. However, the typical implemen- been extended for handling manifold data [13], [14] and to tation of LSLDGC is based on a spherical Gaussian function, which may hierarchical clustering [15]. not work well when the probability density function for data has highly A native approach is first to estimate the probability correlated local structures. To cope with this problem, we propose a new gradient estimator for log-density gradients with Gaussian mixture models density function (e.g., by kernel density estimation), and (GMMs). Covariance matrices in GMMs enable the new estimator to cap- then to compute its log-gradient. However, this approach ture the highly correlated structures. Through the application of the new can be unreliable because a good density estimator does gradient estimator to mode-seeking clustering and hierarchical clustering, not necessarily mean a good gradient estimator. To allevi- we experimentally demonstrate the usefulness of our clustering methods ate this problem, a recent mode-seeking clustering method over existing methods. key words: probability density gradient, mixture model, clustering, hier- called LSLDG clustering employs a sophisticated gradient archical clustering estimator, which directly fits a gradient model to the true log-density gradient without going through density estima- 1. Introduction tion [16]–[18]. LSLDG clustering has been experimentally demonstrated to significantly improve the performance of Estimating the gradient of the logarithm of the probabil- MS particularly for high-dimensional data [17], [18].How- ity density function underlying data is an important prob- ever, the gradient estimator in LSLDG clustering is typically lem. For instance, an unsupervised dimensionality reduction implemented based on the spherical Gaussian kernel. Thus, method requires to estimate the log-density gradient [1].In when the probability density function includes clusters with a supervised learning problem, estimating the log-density highly correlated structures, LSLDG may require a huge gradient enables us to capture multiple functional relation- number of samples to produce a smooth gradient estimate. ship between output and input variables [2]. Other statistical This can be problematic particularly in mode-seeking clus- topics can be seen in [3]. Thus, log-density gradient estima- tering because an unsmooth estimate may create spurious tion offers a certain range of applications in statistical data modes as we demonstrate later. analysis. To improve the performance of LSLDG clustering to Among them, an interesting application is mode- the highly correlated structures, we propose to use a Gaus- seeking clustering. Mean shift clustering (MS) [4]–[6] it- sian mixture model (GMM) in log-density gradient estima- eratively updates data samples toward the modes (i.e., lo- tion. Estimating the covariance matrices in GMM makes the cal maxima) of the estimated probability density function gradient estimator much more adaptive to the local struc- by gradient ascent, and then assigns a cluster label to the tures in the probability density function. A challenge is data samples which converged to the same mode. Com- to satisfy the positive semidefinite constraint of the covari- pared with k-means and mixture-model-based clustering [7], ance matrices. To overcome this challenge, we develop MS has two notable points: MS does not make strong as- an estimation algorithm combined with manifold optimiza- sumptions on the probability density function and the num- tion [19]. ber of clusters in MS is automatically determined by the Next, we apply the proposed estimator to mode- seeking clustering. To update data samples during mode- Manuscript received October 20, 2018. Manuscript revised February 14, 2019. seeking, we derive an update formula based on the fixed- Manuscript publicized March 22, 2019. point method. A similar formula has been proposed in †The author is with the Graduate University for Advanced MS [20], but as shown later, our formula includes it as a spe- Studies, Tachikawa-shi, 190–0014 Japan. cial case. Furthermore, we extend the proposed clustering †† The authors are with Nara Institute of Science and Technol- method to hierarchical clustering, which is a novel exten- ogy, Ikoma-shi, 630–0192 Japan. sion of LSLDG clustering. The usefulness of the proposed a) E-mail: [email protected] b) E-mail: [email protected] clustering methods are experimentally demonstrated. c) E-mail: [email protected] This paper is organized as follows: Sect. 2 reviews an DOI: 10.1587/transinf.2018EDP7354 existing gradient estimator and MS. In Sect. 3, we propose a Copyright c 2019 The Institute of Electronics, Information and Communication Engineers ZHANG et al.: LOG-DENSITY GRADIENT ESTIMATION WITH GAUSSIAN MIXTURE MODELS 1155 ⎛ ⎞ new estimator for log-density gradients with Gaussian mix- ⎜ x − c 2 ⎟ ϕ( j)(x):= ∂ exp ⎝⎜− i ⎠⎟ , (3) ture models. Section 4 develops a mode-seeking clustering i j 2 2σ j method as an application of the proposed gradient estimator. Then, the developed clustering method is further extended = (1) (d) where ci : (ci ,...,ci ) is the kernel centers fixed at a to hierarchical clustering. The performance of the cluster- subset of data samples randomly chosen from D, and σ j de- ing methods are experimentally investigated in Sect. 5. Sec- notes the bandwidth parameter. After substituting the model tion 6 concludes this paper. ( j) = ( j) ( j) (2) into J j, the optimal α : (α1 ,...,αb ) can be com- puted analytically by solving the following problem: 2. Background ( j) ( j) ( j) 2 α := argmin J j(α ) + λ jα This section reviews an existing estimator for log-density α( j) −1 gradients and mode-seeking clustering methods. = −(G j + λ j Ib) h j, 2.1 Review of LSLDG where λ j is the regularization parameter, Ib denotes the b by ( j) = ( j) ( j) b identity matrix, and with ϕ (x): (ϕ1 (x),...,ϕb (x)) , Here, we review a direct estimator for log-density gradi- n n ents which we refer to the least-squares log-density gradi- = 1 ( j) ( j) = 1 ( j) ents (LSLDG) [16], [17]. Suppose that n i.i.d. data samples G j : ϕ (xi)ϕ (xi) , h j : ∂ jϕ (xi). n = n = drawn from a probability distribution with density p(x)are i 1 i 1 available as The gradient estimator is finally given by (1) (d) n b D := {xi = (x ,...,x )} = ∼ p(x). i i i 1 = ( j) ( j) g j(x) αi ϕi (x). The goal of LSLDG is to estimate the gradient of the log- i=1 density from D: 2.2 Review of Mode-Seeking Clustering ∂ p(x) ∂ p(x) ∇ log p(x) = 1 ,..., d , x p(x) p(x) Mean shift clustering (MS) [4], [6] is a mode-seeking clus- tering method based on kernel density estimation (KDE). = ∂ ∇ ff where ∂ j : ∂x( j) and x denotes the di erential operator MS employs KDE to estimate the probability density func- with respect to x. tion as follows: LSLDG directly fits a model g j to the true partial n 1 x − x 2 derivative of the log-density under the squared-loss: p (x):= K i , (4) KDE n 2h2 i=1 1 2 J j(g j):= {g j(x) − ∂ j log p(x)} p(x)dx − C j 2 where h denotes the bandwidth parameter and K is a smooth kernel function for KDE (See [6, Eq. (3)] for definition). = 1 { }2 − g j(x) p(x)dx g j(x) ∂ j p(x) dx 2 Taking the gradient of pKDE(x) with respect to x yields = 1 { }2 + ∇ p (x) g j(x) p(x)dx ∂ jg j(x) p(x)dx, x KDE 2 n 1 x − x x − x 2 = i H i where C := 1 {∂ log p(x)}2 p(x)dx, and we applied the n h2 2h2 j 2 j i=1 integration by parts to the second term on the right-hand ⎡ ⎤ ⎢n − 2 n − 2 ⎥ | |→ | |→∞ 1 ⎢ x xi x xi ⎥ side under the assumption that g j(x)p(x) 0as x j . = ⎣⎢ x H − x H ⎦⎥ nh2 i 2h2 2h2 Then, the empirical risk up to the ignorable constant C j is i=1 i=1 obtained as where H(t):= − d K(t). Based on the fixed-point method, n dt 1 1 setting ∇ p (x) = 0 yields a simple update formula as J g = g x 2 + ∂ g x . x KDE j( j): j( i) j j( i) (1) n = 2 − 2 i 1 n x H x xi ← i=1 i 2h2 x 2 . To estimate g j, LSLDG employs the following model: n x−xi i=1 H 2h2 b g (x):= α( j)ϕ( j)(x), (2) This update formula is iteratively applied for data samples xi j i i i=1 until they converge to the modes of pKDE. MS finally assigns the same cluster label to the data samples converging to the ( j) ( j) ffi where αi and ϕi are coe cients and basis functions re- same mode.
Load more

Recommended publications
  • Which Moments of a Logarithmic Derivative Imply Quasiinvariance?
    Doc Math J DMV Which Moments of a Logarithmic Derivative Imply Quasi invariance Michael Scheutzow Heinrich v Weizsacker Received June Communicated by Friedrich Gotze Abstract In many sp ecial contexts quasiinvariance of a measure under a oneparameter group of transformations has b een established A remarkable classical general result of AV Skorokhod states that a measure on a Hilb ert space is quasiinvariant in a given direction if it has a logarithmic aj j derivative in this direction such that e is integrable for some a In this note we use the techniques of to extend this result to general oneparameter families of measures and moreover we give a complete char acterization of all functions for which the integrability of j j implies quasiinvariance of If is convex then a necessary and sucient condition is that log xx is not integrable at Mathematics Sub ject Classication A C G Overview The pap er is divided into two parts The rst part do es not mention quasiinvariance at all It treats only onedimensional functions and implicitly onedimensional measures The reason is as follows A measure on R has a logarithmic derivative if and only if has an absolutely continuous Leb esgue density f and is given by 0 f x ae Then the integrability of jj is equivalent to the Leb esgue x f 0 f jf The quasiinvariance of is equivalent to the statement integrability of j f that f x Leb esgueae Therefore in the case of onedimensional measures a function allows a quasiinvariance criterion as indicated in the abstract
    [Show full text]
  • A Logarithmic Derivative Lemma in Several Complex Variables and Its Applications
    TRANSACTIONS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 363, Number 12, December 2011, Pages 6257–6267 S 0002-9947(2011)05226-8 Article electronically published on June 27, 2011 A LOGARITHMIC DERIVATIVE LEMMA IN SEVERAL COMPLEX VARIABLES AND ITS APPLICATIONS BAO QIN LI Abstract. We give a logarithmic derivative lemma in several complex vari- ables and its applications to meromorphic solutions of partial differential equa- tions. 1. Introduction The logarithmic derivative lemma of Nevanlinna is an important tool in the value distribution theory of meromorphic functions and its applications. It has two main forms (see [13, 1.3.3 and 4.2.1], [9, p. 115], [4, p. 36], etc.): Estimate (1) and its consequence, Estimate (2) below. Theorem A. Let f be a non-zero meromorphic function in |z| <R≤ +∞ in the complex plane with f(0) =0 , ∞. Then for 0 <r<ρ<R, 1 2π f (k)(reiθ) log+ | |dθ 2π f(reiθ) (1) 0 1 1 1 ≤ c{log+ T (ρ, f) + log+ log+ +log+ ρ +log+ +log+ +1}, |f(0)| ρ − r r where k is a positive integer and c is a positive constant depending only on k. Estimate (1) with k = 1 was originally due to Nevanlinna, which easily implies the following version of the lemma with exceptional intervals of r for meromorphic functions in the plane: 1 2π f (reiθ) (2) + | | { } log iθ dθ = O log(rT(r, f )) , 2π 0 f(re ) for all r outside a countable union of intervals of finite Lebesgue measure, by using the Borel lemma in a standard way (see e.g.
    [Show full text]
  • Understanding Diagnostic Plots for Well-Test Interpretation
    Published in Hydrogeology Journal 17, issue 3, 589-600, 2009 1 which should be used for any reference to this work Understanding diagnostic plots for well-test interpretation Philippe Renard & Damian Glenz & Miguel Mejias Keywords Hydraulic testing . Conceptual models . Hydraulic properties . Analytical solutions Abstract In well-test analysis, a diagnostic plot is a models in order to infer the hydraulic properties of the aquifer. scatter plot of both drawdown and its logarithmic In that framework, a diagnostic plot (Bourdet et al. derivative versus time. It is usually plotted in log–log 1983) is a simultaneous plot of the drawdown and the scale. The main advantages and limitations of the method logarithmic derivative ðÞ@s=@ ln t ¼ t@s=@t of the draw- are reviewed with the help of three hydrogeological field down as a function of time in log–log scale. This plot is examples. Guidelines are provided for the selection of an used to facilitate the identification of an appropriate appropriate conceptual model from a qualitative analysis conceptual model best suited to interpret the data. of the log-derivative. It is shown how the noise on the The idea of using the logarithmic derivative in well-test drawdown measurements is amplified by the calculation interpretation is attributed to Chow (1952). He demon- of the derivative and it is proposed to sample the signal in strated that the transmissivity of an ideal confined aquifer order to minimize this effect. When the discharge rates are is proportional to the ratio of the pumping rate by the varying, or when recovery data have to be interpreted, the logarithmic derivative of the drawdown at late time.
    [Show full text]
  • Symmetric Logarithmic Derivative of Fermionic Gaussian States
    entropy Article Symmetric Logarithmic Derivative of Fermionic Gaussian States Angelo Carollo 1,2,* ID , Bernardo Spagnolo 1,2,3 ID and Davide Valenti 1,4 ID 1 Department of Physics and Chemistry, Group of Interdisciplinary Theoretical Physics, Palermo University and CNISM, Viale delle Scienze, Ed. 18, I-90128 Palermo, Italy; [email protected] (B.S.); [email protected] (D.V.) 2 Department of Radiophysics, Lobachevsky State University of Nizhni Novgorod, 23 Gagarin Avenue, Nizhni Novgorod 603950, Russia 3 Istituto Nazionale di Fisica Nucleare, Sezione di Catania, Via S. Sofia 64, I-90123 Catania, Italy 4 Istituto di Biomedicina ed Immunologia Molecolare (IBIM) “Alberto Monroy”, CNR, Via Ugo La Malfa 153, I-90146 Palermo, Italy * Correspondence: [email protected] Received: 21 May 2018; Accepted: 16 June 2018; Published: 22 June 2018 Abstract: In this article, we derive a closed form expression for the symmetric logarithmic derivative of Fermionic Gaussian states. This provides a direct way of computing the quantum Fisher Information for Fermionic Gaussian states. Applications range from quantum Metrology with thermal states to non-equilibrium steady states with Fermionic many-body systems. Keywords: quantum metrology; Fermionic Gaussian state; quantum geometric information 1. Introduction Quantum metrology or quantum parameter estimation is the theory that studies the accuracy by which a physical parameter of a quantum system can be estimated through measurements and statistical inference. In many physical scenarios, quantities which are to be estimated may not be directly observable, either due to experimental limitations or on account of fundamental principles. When this is the case, one needs to infer the value of the variable after measurements on a given probe.
    [Show full text]
  • Differential Algebraic Equations from Definability
    Differential algebraic equations from definability Thomas Scanlon 24 October 2014 Thomas Scanlon ADEs from definability Is the logarithm a function? × The exponential function exp : C ! C has a many-valued analytic × inverse log : C ! C where log is well-defined only up to the adding an element of 2πiZ. Thomas Scanlon ADEs from definability The logarithmic derivative Treating exp and log as functions on functions does not help: If U × is some connected Riemann surface and f : U ! C is analytic, then we deduce a “function” log(f ): U ! C. However, because log(f ) is well-defined up to an additive constant, d @ log(f ) := dz (log(f )) is a well defined function. That is, for M = M (U) the differential field of meromorphic functions on U we have a well-defined differential-analytic function @ log : M× ! M. f 0 Of course, one computes that @ log(f ) = f is, in fact, differential algebraic. Thomas Scanlon ADEs from definability Explanation? Why is the logarithmic differential algebraic? What a silly question! The logarithm is the very first transcendental function whose derivative is computed in a standard calculus course. The differential algebraicity of d dz (log(f )) is merely a consequence of an elementary calculation. The usual logarithmic derivative is an instance of Kolchin’s general theory of algebraic logarithmic derivatives on algebraic groups in which the differential algebraicity is explained by the triviality of the tangent bundle of an algebraic group. It can also be seen as an instance of the main theorem to be discussed today: certain kinds of differential analytic functions constructed from covering maps are automatically differential algebraic due to two key ideas from logic: elimination of imaginaries and the Peterzil-Starchenko theory of o-minimal complex analysis.
    [Show full text]
  • Dictionary of Mathematical Terms
    DICTIONARY OF MATHEMATICAL TERMS DR. ASHOT DJRBASHIAN in cooperation with PROF. PETER STATHIS 2 i INTRODUCTION changes in how students work with the books and treat them. More and more students opt for electronic books and "carry" them in their laptops, tablets, and even cellphones. This is This dictionary is specifically designed for a trend that seemingly cannot be reversed. two-year college students. It contains all the This is why we have decided to make this an important mathematical terms the students electronic and not a print book and post it would encounter in any mathematics classes on Mathematics Division site for anybody to they take. From the most basic mathemat- download. ical terms of Arithmetic to Algebra, Calcu- lus, Linear Algebra and Differential Equa- Here is how we envision the use of this dictio- tions. In addition we also included most of nary. As a student studies at home or in the the terms from Statistics, Business Calculus, class he or she may encounter a term which Finite Mathematics and some other less com- exact meaning is forgotten. Then instead of monly taught classes. In a few occasions we trying to find explanation in another book went beyond the standard material to satisfy (which may or may not be saved) or going curiosity of some students. Examples are the to internet sources they just open this dictio- article "Cantor set" and articles on solutions nary for necessary clarification and explana- of cubic and quartic equations. tion. The organization of the material is strictly Why is this a better option in most of the alphabetical, not by the topic.
    [Show full text]
  • Reminder. Vector and Matrix Differentiation, Integral Formulas
    Euler equations for multiple integrals January 22, 2013 Contents 1 Reminder of multivariable calculus 2 1.1 Vector differentiation . 2 1.2 Matrix differentiation . 3 1.3 Multidimensional integration . 7 1 This part deals with multivariable variational problems that describe equi- libria and dynamics of continua, optimization of shapes, etc. We consideration of a multivariable domain x 2 Ω ⊂ Rd and coordinates x = (x1; : : : ; xd). The variational problem requires minimization of an integral over Ω of a Lagrangian L(x; u; Du) where u = u(x) is a multivariable vector function and ru of mini- mizers and matrix Du = ru of gradients of these minimizers. The optimality conditions for these problems include differentiation with respect to vectors and matrices. First, we recall several formulas of vector (multivariable) calculus which will be commonly used in the following chapters. 1 Reminder of multivariable calculus 1.1 Vector differentiation We remind the definition of vector derivative or derivative of a scalar function with respect to a vector argument a 2 Rn. Definition 1.1 If φ(a) is a scalar function of a column vector argument a = T dφ (a1; : : : ; an) , then the derivative da is a row vector 0 a 1 dφ dφ dφ 1 = ;:::; if a = @ ::: A (1) da da1 dan an assuming that all partial derivatives exist. This definition comes from consideration of the differential dφ of φ(a): dφ(a) dφ(a) = φ(a + da) − φ(a) = · da + o(kak) da Indeed, the left-hand side is a scalar and the second multiplier in the right-hand side is a column vector, therefore the first multiplier is a row vector defined in (1) Examples of vector differentiation The next examples show the calcula- tion of derivative for several often used functions.
    [Show full text]
  • Official Worksheet 5: the Logarithmic Derivative and the Argument Principle
    Official Worksheet 5: The Logarithmic derivative and the argument principle. April 8, 2020 1 Definitions and the argument principle Today's worksheet is a short one: we cover the logarithmic derivative and the argument principle Suppose that f is a meromorphic function on a domain Ω (recall: this is a function which is either holomorphic on Ω or is singular with only isolated singularities, all of which are poles.) Define the logarithmic derivative d log f 0 f := : dz f Note that for any nonzero function f; this definition makes sense everywhere except zeroes or poles of f (where it has at worst poles), since f 0 is defined everywhere that f is. The notation d log is indicative of the fact that, where d d log defined, dx log(f(x)) = dz f: The logarithmic derivative d log has a number of nice properties. They key property is sometimes known as the argument principle, and is as follows. f 0 Theorem 1. 1. If f is nonzero and nonsingular at z0; then f is nonsingular at z0: f 0 2. If f has a pole of order n at z0 then f has a simple pole with residue equal to −n at z0: f 0 3. If f has a zero of degree n at z0 then f has a simple pole with residue equal to n at z0: Corollary 1. If f is a meromorphic function on Ω and γ is a simple closed H f 0 curve inside Ω such that f is nonsingular and nowhere zero on γ; then γ f (z)dz is equal to 2πi(Z − P ); where Z is the number of zeroes of f inside γ and P is the number of poles inside of γ, both counted with multiplicity (i.e.
    [Show full text]
  • A Note on the Logarithmic Derivative of the Gamma Function
    THE EDINBURGH MATHEMATICAL NOTES PUBLISHED BY THE EDINBURGH MATHEMATICAL SOCIETY EDITED BY D. MARTIN, M.A., B.Sc, Ph.D. No. 38 1952 A note on the logarithmic derivative of the gamma function By A. P. GUINAND. Introduction. The object of this note is to give simpler proof* of two formulae involving the function <p (z) which I have proved elsewhere by more complicated methods.1 The formulae are 2 ^. (x + 1) - log x = 2J" U (t + 1) - log t) cos 2nxt dt (1> for all real positive x, and for | arg z | < TT, where y is Euler's constant. Proof of (I). Now3 1 Journal London Math. Soc., 22 (1947), 14-18. f{z) denotes lv (.-;) / Y (z). 'J The first formula shows that ^{x + 1) - log x is self-reciprocal with respect to the Fourier cosine kernel 2 cos 2TTX. It is strange that this result should have been overlooked, but I can find no trace of it. Cf. B. M. Mehrotra, Journal Indian Math. Soc. (New Series), 1 (1934), 209-27 for a list of self-reciprocal functions and references. 8 C. A. Stewart, Advanced Calculus, (London, 1940), 495 and 497. Downloaded from https://www.cambridge.org/core. IP address: 170.106.40.139, on 26 Sep 2021 at 06:18:02, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0950184300002949 A. P. GUINAND __ _. d^ Also dt Jo by Frullani's integral,1 and o t2 + x* 2x Combining these results with (3) and (4) we have JO tdt (5) \2T7< etirt — 1 Hence 2 f'C|i/;(<+ 1)—logiUos Jol J u e — f»/l 1 \ fM = 2 - — u , Idtt c~ "'cos 2nxtdt )0\u e ~ll Jo J 0 \u e« - _ 2 -1 * J o \ 2irt • eZnt — ' = ip (x + 1) — log £ by (5).
    [Show full text]
  • Timath.Com Calculus
    TImath.com Calculus The Logarithmic Derivative Time required ID: 9092 45 minutes Activity Overview Students will determine the derivative of the function y = ln(x) and work with the derivative of both y = ln(u) and y = loga(u). In the process, the students will show that ln(a+− h) ln(a) 1 lim = . h0→ ha Topic: Formal Differentiation • Derive the Logarithmic Rule and the Generalized Logarithmic Rule for differentiating logarithmic functions. • Prove that ln( x )= ln(a) ⋅ loga ( x ) by graphing f(x) = loga(x) and g(x) = ln(x) for some a and deduce the Generalized Rule for Logarithmic differentiation. • Apply the rules for differentiating exponential and logarithmic functions. Teacher Preparation and Notes • This investigation derives the definition of the logarithmic derivative. The students should be familiar with keystrokes for the limit command, the derivative command, entering the both the natural logarithmic and the general logarithmic functions, drawing a graph, and setting up and displaying a table. • Before starting this activity, students should go to the home screen and select F6:Clean Up > 2:NewProb, and then press . This will clear any stored variables, turn off any functions and plots, and clear the drawing and home screens. • This activity is designed to be student-centered with the teacher acting as a facilitator while students work cooperatively. • To download the student worksheet, go to education.ti.com/exchange and enter “9092” in the keyword search box. Associated Materials • LogarithmicDerivative_Student.doc Suggested Related Activities To download any activity listed, go to education.ti.com/exchange and enter the number in the keyword search box.
    [Show full text]
  • 3 Additional Derivative Topics and Graphs
    3 Additional Derivative Topics and Graphs 3.1 The Constant e and Continuous Compound Interest The Constant e The number e is defined as: 1 n lim 1 + = lim(1 + s)1=s = 2:718281828459 ::: x!1 n s!0 Continuous Compound Interest If a principal P is invested at an annual rate r (expressed as a decimal) compounded continuously, then the amount A in the account at the end of t years is given by the compound interest formula: A = P ert 3.2 Derivatives of Exponential and Logarithmic Functions Derivatives of Exponential Functions For b > 0, b 6= 1: d • ex = ex dx d • bx = bx ln b dx Derivatives of Logarithmic Functions For b > 0, b 6= 1, and x > 0: d 1 • ln x = dx x d 1 1 • log x = dx b ln b x 1 Change-of-Base Formulas The change-of-base formulas allow conversion from base e to any base b, b > 0, b 6= 1: • bx = ex ln b ln x • log x = b ln b 3.3 Derivatives of Products and Quotients Product Rule If y = f(x) = F (x)S(x), then f 0(x) = F (x)S0(x) + S(x)F 0(x) provided that both F 0(x) and S0(x) exist. Quotient Rule T (x) If y = f(x) = , then B(x) B(x)T 0(x) − T (x)B0(x) f 0(x) = [B(x)]2 provided that both T 0(x) and B0(x) exist. 3.4 The Chain Rule Composite Functions A function m is a composite of functions f and g if m = f[g(x)].
    [Show full text]
  • Introduction to Differential Equations
    Derivatives – An Introduction Concepts of primary interest: Definition: first derivative Differential of a function Left-side and right-side limits Higher derivatives Mean value and Rolle’s theorem L’Hospital Rule for indeterminate forms Implicit differentiation Inverse functions Logarithmic Derivative Method for Products Extrema and the Determinant Test for functions of two variables Lagrange Undetermined Multipliers (*** needs serious edit ****) Sample calculations: 2 Derivative of x Chain Rule Product Rule Saddle-points and local extrema Derivative of the inverse function for Squared d(sinθ) /dθ = 1 for θ = 0 Applications Closest point of approach or CPA (nautical term) Tools of the Trade Derivatives of inverse functions Contact: [email protected] 11/1/2012 Plotting inverse functions ***ADD a table of the derivatives of common functions See Tipler; Lagrange Approximating Functions Many functions are complicated, and their evaluations require detailed, tortuous calculations. To avoid the stress, approximate representations are sometimes substituted for the actual functions of interest. For a continuous function f(x), it might be adequate to replace the function in a small interval around x0 by its value at that point f(x0). A better approximation for a function f(x) with a continuous derivative follows from the definition of that derivative. df fx()− fx ( ) = Limit 0 [DI.1] x0 xx→ dx 0 x− x0 df df Clearly, fx()≅+ fx ( ) ( x − x) where is the slope of the function at x0.. 00dx dx x0 x0 11/1/2012 Physics Handout Series.Tank: Derivatives D/DX- 2 Figure DI-1 This handout formalizes the definition of a derivative and illustrates a few basic properties and applications of the derivative.
    [Show full text]