It has three meanings: (1) a "substitute formula" used to simplify (speed up) the calculation of a function; (2) an "formula that calculates a value close to the true function value" used when it is not possible to perform an exact function calculation within a finite time; and (3) an "formula that represents the approximate shape" of a curve or curved surface. [Hayato Togawa] principleIf the range of application (the domain of the independent variables) is divided into several regions and the range covered by one equation is narrowed, it is possible to approximate using low-order polynomials, rational expressions, continued fractions, etc. Specifically, there are several methods, such as (1) using an interpolating polynomial, (2) using the least squares method, (3) using a Taylor series, (4) using a Pade expansion (a type of approximation method using rational function approximation, i.e., the ratio of polynomials), and (5) using a continued fraction expansion. In addition, by applying the theory of best approximation, it is possible to obtain an approximate equation with higher accuracy and better computational efficiency. For example, to make an approximation of a logarithmic function, use the infinite series expansion of log(1 + x ) [Hayato Togawa] Best ApproximationIt is desirable for an approximation formula to have as little error as possible, require as little calculation time as possible, and be applicable to a wide range of areas. From this perspective, research is being conducted into methods for determining the best approximation formula by setting a "comprehensive measure of the quality of the approximation" and adjusting the coefficients of the approximation formula. The classical method is the least squares method, which fixes the form and range of application of the approximation equation and determines the coefficients so that the sum of squares of the error (the integral of the square of the error over the entire range of application) is minimized. However, for the purpose of approximating a function, it is more desirable to minimize the maximum error within the range of applicability. This is called "minimax approximation." [Hayato Togawa] Standard Function ApproximationExcellent approximation formulas are known for trigonometric functions, inverse trigonometric functions, exponential functions, logarithmic functions, square roots, cube roots, error functions, elliptic functions, gamma functions, Bessel functions, etc. [Hayato Togawa] Precautions when using approximate expressionsAn approximation formula usually approximates the value of a function, but does not approximate the rate of change of the function. Therefore, an equation obtained by differentiating an approximation formula cannot be used as an approximation formula for a derivative. If you want to approximate not only the value of a function but also the value of its derivative, you need to use a special method (spline approximation). [Hayato Togawa] "Applications of Algol and Fortran" by Fujino Seiichi (1971, Asakura Publishing)" ▽ "Approximate Calculation Methods for Electronic Computers" by Hastings, translated by Takeuchi Hitoshi (1973, Tokyo Tosho)" ▽ "Numerical Calculation of Elementary Functions" by Ichimatsu Makoto (1974, Kyoiku Shuppan)" ▽ "Numerical Calculation and Its Applications" by Takahashi Iwao and Muroya Yoshiaki (1979, Corona Publishing)" ▽ "Programming Approximate Formulas" by Hamada Hozumi (1995, Baifukan Publishing) [References] | | | | | | |Logarithmic | | | | | |ContinuedSource: Shogakukan Encyclopedia Nipponica About Encyclopedia Nipponica Information | Legend |
(1)関数の計算を簡略化(高速化)するために使われる「代用の式」、(2)有限の時間内では厳密な関数計算ができない場合に使われる「真の関数値に近い値を算出する式」、(3)曲線や曲面の「概形を表す式」、の三つの意味がある。 [戸川隼人] 原理適用範囲(独立変数の変域)をいくつもの区域に分け、一つの式でカバーする範囲を狭くすれば、低い次数の多項式、有理式、連分数などで近似できる。具体的には、(1)補間多項式を使う、(2)最小二乗法を使う、(3)テーラー級数を使う、(4)パデ展開(有理関数近似、すなわち多項式の比で近似する方式の一種)を使う、(5)連分数展開を使う、などの方法がある。また、最良近似の理論を応用すれば、より高精度で計算効率の優れた近似式を得ることができる。 たとえば、対数関数の近似式をつくるには、log(1+x)の無限級数展開 [戸川隼人] 最良近似近似式は、できるだけ誤差が少なく、計算の手間が少なく、適用できる範囲が広いことが望ましい。そのような観点から「近似の良さの総合的な尺度」を設け、近似式の係数を調整して最良の近似式を求める方法が研究されている。 古典的な手法としては最小二乗法がある。これは近似式の形と適用範囲を固定し、誤差の二乗和(誤差の二乗を適用範囲全域にわたって積分したもの)が最小になるように係数を決める手法である。 しかし、関数を近似する、という目的のためには、適用範囲内における誤差の最大値を最小にすることのほうが望ましい。これを「ミニマクス近似」という。 [戸川隼人] 標準的な関数の近似式三角関数、逆三角関数、指数関数、対数関数、平方根、立方根、誤差関数、楕円(だえん)関数、ガンマ関数、ベッセル関数などに関しては優れた近似式が知られている。 [戸川隼人] 近似式を利用する際の注意事項近似式は、普通、関数の値を近似するものであって、関数の変化率までは近似していない。したがって、近似式を微分した式を導関数の近似式として利用することはできない。関数の値だけでなく導関数の値も近似したい場合には、それ専用の手法(スプライン近似)を用いる必要がある。 [戸川隼人] 『藤野精一著『アルゴルとフォートランの応用』(1971・朝倉書店)』▽『ヘースティングス著、竹内均訳『電子計算機のための近似計算法』(1973・東京図書)』▽『一松信著『初等関数の数値計算』(1974・教育出版)』▽『高橋磐郎・室谷義昭著『数値計算とその応用』(1979・コロナ社)』▽『浜田穂積著『近似式のプログラミング』(1995・培風館)』 [参照項目] | | | | | | | | | | | | |出典 小学館 日本大百科全書(ニッポニカ)日本大百科全書(ニッポニカ)について 情報 | 凡例 |
The Nuclear Safety Commission was established in O...
〘Noun〙① A title given to the Left Chief Historian,...
…One of the main streets in New York City, USA. T...
… [Hemangioma] Also called hemangioma. Hemangioma...
... Diagnosis is based on weight loss, apparent t...
...To eliminate this drawback, rubber or plastic ...
A general term for "Government and Relief in ...
One of the boroughs of Greater London, the capita...
The capital of Qatar. It is a port city located o...
〘noun〙 A ceremony to bury the dead. A funeral. A f...
...In 271, Rome withdrew and the so-called Migrat...
Zhao Tuo, a military commander from Hebei Zhendin...
The lyric part of Noh. It is also called ji or ho...
…The background to this idea is that caves were r...
A town in Oura County in the southeastern part of ...