http://honyaku.yahoofs.jp/url_result?ctw_=sT,eCR-EJ,bT,hT,uaHR0cDovL3d3dy50Ym5zLm5ldC9sZW8vR0VzdGlUeHQuaHRt,qlang=ja|for=0|sp=-5|fs=100%|fb=0|fi=0|fc=FF0000|db=T|eid=CR-EJ

Yahoo!翻訳
Yahoo! JAPAN - Yahoo!翻訳 - ヘルプ
crosslanguage
Yahoo! JAPANとページ内のコンテンツとの関連はありません。 原文のページは、こちらから確認できます。
※英単語をダブルクリックすると、別ウィンドウで辞書検索を表示します。(リンクの設定されたテキストは除く)
翻訳設定
翻訳方向:
結果表示:
翻訳技術提供:株式会社クロスランゲージ

General Estimation Theory
一般的な評価論

  

Mathematical Monograph
数学的なモノグラフ

  

by
そばに

 

© Lev Gelimson
© レフGelimson

  

The “Collegium” All World Academy of Sciences Publishers
科学出版者の「委員会」全く世界アカデミー

Munich (Germany)
ミュンヘン(ドイツ)

 

Ninth Edition (2010)
第9のエディション(2010)
Eighth Edition (2004)
第8のエディション(2004)
Seventh Edition (2001)
第7のエディション(2001)
Sixth Edition (2000)
第6のエディション(2000)
Fifth Edition (1995)
第5のエディション(1995)
Fourth Edition (1994)
第4のエディション(1994)
Third Edition (1993)
第3のエディション(1993)
Second Edition (1992)
第2のエディション(1992)
First Edition (1988)
最初のエディション(1988)

 

0. Introduction
0. 導入

 

Urgent scientific and life problems demand development of modern estimation methods [1, 2]. The concept of approximate solutions does not cover inexact pseudo-solutions. Absolute errors are not invariant measures of accuracy. Relative errors are used only for equalities of exact values to their approximations and become indefinite and inadequate if the ratio of the sides of the equality is not close to 1.
緊 急の科学的なおよび生命問題は、最新の評価方法[1、2]の開発を要求します。およその解決の概念は、不正確な偽解決をカバーしません。絶対のエラーは、 正確さの不変の計測でありません。平等の側の比率が約1でないならば、相対的なエラーが彼らの近いものに正確な価値のequalitiesだけのために使 われて、明確でなくて不十分になります。

Estimation methods must be based on the intercommunicated concepts of sets and numbers. But the Cantor's concept [3, 4] of sets ignores multiplicities of their elements, identifies essentially different sets, and gives indefinite numbers of elements and other functions of sets especially if they involve closely spaced elements not exactly known, although the multiplicities of solutions of polynomials are used in algebra. So it is necessary to consider generalized sets with taking the multiplicities of their elements into account. And the known numbers are insufficient to construct sensitive estimations and need their replenishment.
評 価方法は、セットと数の相互に通じられた概念に基づかなければなりません。しかし、セットのカントールの概念[3、4]は、多数の彼らの要素を無視して、 基本的に異なるセットを特定して、多項式の解決の多様性が代数学で使われるが、特に彼らが必ずしも知られていない密接に間隔をあけられた要素を含むなら ば、要素の不明確な数とセットの他の機能を伝えます。それで、それは彼らの要素の多様性を考慮に入れることで分化していないセットを考慮するのに必要で す。そして、既知の数は、敏感な評価を造って、彼らの補充を必要とするには不十分です。

 

 

1. Generalized Sets and Numbers
1.Setsを一般化する、そして、 数

 

Let
されます

ν: x νx
ν: x → νx

be a numerical function (functional), whose domain of definition includes all sets of some type, whose values are some generalized numbers of the elements of sets, and which satisfies the following basic axioms:
明確さの領域が若干のタイプの全セットを含む数の機能(機能的な)です、価値はセットの若干の分化していない数の要素で、そして、以下の基本的な原理を満たす:

1)   if a set x consists of one element only,
1)    セットされたxが1つの要素から成るならば、 だけ

νx = 1;
νx = 1;

2)   the estimation function n is quite additive:
2)    評価機能n 非常に添加物です:

ναΑ Gα = αΑ νGα
ν ∪α ∈ΑGα = ∑α ∈Α νGα

where
どこで

Gα is any generalized set indexed by α in Cantor set Α having any cardinality.
Gαは、どんな基数でも持っているΑを課されるカントールでαによってインデックスを付けられるどんな分化していないセットでもあります。

It can be rigorously proved that for the empty set ,
それは、空のセットのためのそれということを厳しく証明されることができます ∅

ν = 0,
ν ∅ =0、

for any finite set,
どんな有限セットのためにでも、

ν{a1, a2, ... , an} = n,
ν{a1、 a2、...}= n、

and for any set if one element is added to or is withdrawn from x, nx is increased or decreased by 1, respectively. This is known only for finite sets because it is considered that any cardinality is absorbed by a greater infinite one although it is not true for ordinal numbers [3, 4]. In particular, if N is the Cantor set of all the natural numbers (positive integers) and it is designated that
そ して、1つの要素が増されるか、xから引っ込められるならば、どんなセットのためにでも、それぞれ、nxは1時までに増減されます。それが序数詞[3、 4]にとって真実でないがどんな基数でもより偉大な無限の人に夢中であると考えられるので、これは有限セットだけで知られています。特に、Nがすべての自 然数(正整数)とそれの配置されるカントールであるかどうかは、それと称されます

νN = N,
νN = ∞N、

N + 1 > N.
∞N + 1 > ∞N.

If N’ is some subset of N and there exists such a (fractional) number μN’ and such infinite subseries nk in N that for every k
Nならば』、いくつかがNのサブセットです、そして、そのような(わずかな)数μNが、存在する』とてもあらゆるkのためのNでその他無限のsubseries nk

ν(N’ { 1, 2, ... , nk}) = μN’ nk
ν』(N ∩ {1、2、...nk})=μN』nk

then N’ is called fractionally numerable and it is designated that
そしてN』はわずかに数えられると言われます、そして、それはそれと称されます

νN’ = μN’N.
νN』 = μN』 ∞N.

If N’ is the former and N’’ is such a subset of N that the both differences
Nならば』、前者とNである」Nのそのようなサブセットが、それである両方の違い

N’’ \ N’
N」 \ N』

and
そして、

N’ \ N’’
N』\ N」

[3, 4] are finite, then N’’ is called finitely fractionally numerable and it is designated that
[3、 4]有限、当時のNです」わずかに数えられて制限的に言います、そして、それがそれと称されます

νN’’ = νN’+ ν(N’’ \ N’) - ν(N’ \ N’’).
νN」 = νN'+ ν」(N」\ N』)- ν』(N \ N」)。

(But there also exist other types of subsets of N, for example,
(しかし、また、たとえば、Nの他のタイプのサブセットが、存在します

{11, 12, ... , 100, 1001, 1002, ... , 10000, ... ,
{11、 12、...100、1001、1002、...10000、...

102n-l + l, 102n-l + 2, ... , 102n, ...}).
102n-l + l、102n-l + 2、...102n、...})。

Let us then designate
それから示そう

1/N = 0+,
1/ ∞N = 0+、

0 - 0+ = 0-,
0 -0+ = 0-、

(0+)ρ = 0ρ+,
(0+)ρ =0ρ+、

r + 0ρ+ = rρ+,
r + 0ρ+ = rρ+、

r - 0ρ+ = rρ-
r -0ρ+ = rρ-

where ρ > 0 and r are usual real numbers or infinities.
そこでρ > 0とrは、普通の実数または無限です。

So every real number (or infinity) represents the infinite set of generalized real numbers (R+ is the set of such positive numbers)
あらゆる実数(または無限)が分化していない実数の無限のセットを表すように、(R+は、そのような正数のセットです)

[ρ∈R+ rρ-] {r} [ρ∈R+ rρ+],
[ρ∈R+ rρ-] ∪ {r} ∪ [ρ∈R+ rρ+]、

which is intuitively used in limits and improper integrals and allows to obtain sensitive estimations having some parameters (for example, weights) 0+ instead of 0. Besides that, only such generalized numbers provide expressing probability densities and distribution functions in uniform distributions over sets of infinite measures. And such distribution functions can be depicted not in Euclidean but Lobachevskian geometry that is therefore connected with probability theory by the theory of generalized numbers.
そ してそれは限度と異常積分で直観的に使われて、0の代わりに若干のパラメータ(たとえば重さ)がある敏感な評価に0+を得させるために許します。それの他 に、そのような分化していない数だけは、無限の処置のセットの上に均一な分布で確率密度と分布関数を表すことを提供します。そして、そのような分布関数 は、したがって、分化していない数の理論によって確率論と関係があるLobachevskianジオメトリー以外はユークリッドのもので表されることがで きません。

 

 

2. Generalized Least Upper and Greatest Lower Bounds
2. 全身性最も少なく上で最も大きな下界

 

The least upper and the greatest lower bounds [5] on any ordered set M are still less sensitive for estimation than the known sets and numbers. So it is necessary to introduce some generalized bounds.
全く高い方の部分と少しの順序集合Mの下限[5]も、評価のために、既知のセットと数よりまだ敏感でありません。それで、それは若干の分化していない境界を持ち出すのに必要です。

The generalized least upper bound sup M is the generalized set of the usual least upper bounds on generalized subsets of M reduced from above and is numerically equal to the usual sup M.
全身性最小上界は、Mをすすります普通の最小上界の分化していないセットが上から減らされるMの分化していないサブセットの上にあります、そして、普通のものと数値的に等しいですMをすすります。

The generalized greatest lower bound inf M is the generalized set of the usual greatest lower bounds on generalized subsets of M reduced from below and is numerically equal to the usual inf M.
分化していない下限inf Mは、下から減らされるMの分化していないサブセットの普通の下限の分化していないセットで、普通のinf Mと数値的に等しいです。

All these generalizations have a deep analogy and allow proposing some effective estimation methods.
すべてのこれらの一般化は深い類似を持って、若干の効果的評価方法を提案するのを許します。

 

 

3. Unierrors
3. Unierrors

 

The method [6] to determine the unierror
unierrorを決定する方法[6]

E = |a - b| / (|a| + |b|)
E = | - b| /(|| + |b|)

of a formal (true or false) numerica1 equality
形式的(本当であるか間違った)numerica1平等の

a =? b
=?b

can be naturally generalized to any functional equality or equation in some linear normed space and to such combined equalities or equations. Let us consider the equation
若干の線形ノルム空間のどんな機能的な平等または方程式にでも、そして、そのような合同のequalitiesまたは方程式に自然に一般化されることができます。方程式を考慮しよう

(1)
(1)

Lλ[φ∈Φ fφ[ω∈Ω zω]] = 0 (λ∈Λ)
Lλ[φ∈Φ fφ[ω∈Ωzω]]= 0(λ∈Λ)

where
どこで

Lλ is an operator with index λ from an index set Λ;
Lλは、インデックスλがΛを課されるインデックスからにあるオペレーターです;

fφ is a function (dependent variable) with index φ from an index set Φ;
fφは、インデックスφがΦを課されるインデックスからにある機能(従属変数)です;

zω is an independent variable with index ω from an index set Ω;
zωは、インデックスωがΩを課されるインデックスからにある独立変数です;

[ω∈Ω zω]
[ω∈Ωzω]

is a set of indexed elements zω.
一組のインデックスを付けられた要素は、zωです。

The local unierror may be defined by the formula
ローカルunierrorは、公式によって定義されるかもしれません

(2)
(2)

Eλ[ω∈Ω zω] =
E λ[ ω∈Ω zω]=

αλ ||Lλ’[ω∈Ω zω]||λ /
α λ ||L λ』[ ω∈Ω zω]|| λ /

(||Lλ’[ω∈Ω zω]||λ + aλ ||Lλ“[ω∈Ω zω]||λ) +
(||L λ』[ ω∈Ω zω]|| λ + λ ||L λ」[ ω∈Ω zω]|| λ) +

βλ ||Lλ’[ω∈Ω zω||λ /
β λ ||L λ』[ ω∈Ω zω|| λ /

(|||Lλ’[ω∈Ω zω]|||λ + bl |||Lλ“[ω∈Ω zω]|||λ) +
(|||L λ』[ ω∈Ω zω]||| λ +bl |||L λ」[ ω∈Ω zω]||| λ) +

γλ ||Ll’[ω∈Ω zω]||λ /
γ λ ||Ll』[ ω∈Ω zω]|| λ /

(sup ||Lλ’[ω∈Ω zω]||λ + gλ sup ||Lλ“[ω∈Ω zω]||λ)
(すする ||L λ』[ ω∈Ω zω]|| λ +g λは夕食をとります ||L λ」[ ω∈Ω zω]|| λ)

where
どこで

αλ , βλ , γλ
αλ β λ、 γ λ

are positive uninumbers, their sum be equal to 1;
あります ポジティブなuninumbers、彼らの金額は、1と等しいです;

αλ , βλ , γλ
α λ、 β λ、 γ λ

are generalized positive numbers;
分化していない正数です;

Lλ’[ω∈Ω zω]
L λ』[ ω∈Ω zω]

is the left-hand side of (1) as a direct (not composite) function of the independent variables;
(1)の左側は、独立変数のダイレクト(合成でない)機能としてあります;

the both sup are taken in the domain of definition zl of the equation;     
両方のすすります方程式の定義zlの領域でします;     

|||Lλ’[ω∈Ω zω]|||λ
|||L λ』[ ω∈Ω zω]||| λ

is the usual least upper bound on the norm of
普通の最少の高い方の部分は、載って基準を結びつけられます

||Lλ’[ω∈Ω zω]||λ
||L λ』[ ω∈Ω zω]|| λ

when all possibly different isometric (conserving the norms) transformations even of equal elements in
中で等しい要素のさえすべてのなんとかして異なる等尺性(基準を節約する)変化は、いつですか

Lλ’[ω∈Ω zω]
L λ』[ ω∈Ω zω]

are  considered;
考慮されます;

Lλ“[ω∈Ω zω]
L λ」[ ω∈Ω zω]

is some function that is chosen (along with
いくつかは、選ばれる機能です(とともに

αλ, βλ, γλ ,
α λ、 β λ、 γ λ

αλ, βλ, γλ)
α λ、 β λ、 γ λ)

by the principle of tolerable simplicity [6, 7] so that the estimation (2) is the most sensitive one over the set of the classes of the functions fφ under consideration, i.e., the difference
そう評価(2)が考慮(すなわち違い)の下の機能fφの種類のセットの上に最も敏感なものである許容できる単純さ[6、7]の原則によって

sup E - inf E
Eをすすってください - inf E

has the greatest value.
最も大きな価値はそうします。

Similar estimations can be proposed for other relations, too. If in (1), the equality sign is replaced by inequality sign, let
また、類似した評価は、他の関係のために提案されることができます。(1)でならば、平等徴候は不平等徴候(される)と取り替えられます

Eλ[ω∈Ω zω] = 0
E λ[ ω∈Ω zω]= 0

if the inequation is true, and let us use the formula (2) otherwise. For the generalized comparison
不等式が真実であるならば、そして、さもなければ公式(2)を使おう。分化していない比較のために

a b (mod d)
≡ b(流行の最先端のd)

(i.e.,
(すなわち、

(a - b)/d
( - b)/d

is an integer),
整数です)、

where
どこで

a, b, d
b(d)

are complex numbers
複素数です

(d 0),
(d ≠ 0)、

the unierror may be given by the formula
unierrorは、公式によって与えられるかもしれません

E = min({|a - b| / |d|}, 1 - {|a - b| / |d|})
E =分({| - b| / |d|}、 1 - {| - b| / |d|}

where {x} is the fractional part of a real number x.
そこで本当のナンバーxのわずかな部分です{x}。

Any pseudosolution
どんなpseudosolutionでも

[φ∈Φ fφ[ω∈Ω zω]]
[φ∈Φfφ[ω∈Ω zω]

to the equation (1) transforms it into equality (1) is also estimated by the formula (2).
(1)が変換する方程式にとって、平等(1)へのそれは、公式(2)によっても推定されます。

Generally, a unierror is some functional
通常、unierrorは少し機能的です

E: U [0, 1]
E: U → [0、1]

(u U = E I)
(=E u∈U ∪ I)

where
どこで

U is a domain for estimation;
U 領域は、評価のためです;

E is the subdomain containing all the exact objects;
E サブドメインは、すべての正確な物を含んでいます;

I is the subdomain containing all the inexact objects,
私 サブドメインは、すべての不正確な物を含んでいます、

which satisfies some basic axioms:
そしてそれは若干の基本的な原理を満たします:

1) for every e E,
1) あらゆるeのために ∈ E、

d(e) = 0;
d(e) = 0;

2) there exists an i I such that
2) iが存在します ∈ 私

d(i) = 1.
d(i) = 1。

The simplest (but insensitive) estimation is locally logical:
最も単純な(しかし、鈍感な)評価は、地元で論理的です:

d(u) = 0
d(u) = 0

if
もしも

u E,
u ∈ E、

 

d(u) = 1
d(u) = 1

if
もしも

u I.
u ∈ 私。

So its sum with the probability, that an object to be estimated is exact, is identically equal to 1.
それで、推定される物が正確である可能性によるその金額は、1と同じく等しいです。

The domain average power weighted unierror in the λth relation can be defined as
λth関係の力加重のunierrorが定義されることができる領域平均

Eλ(m(λ)) = {lim [V(zλ’)]-1∫(||Lλ[φ∈Φ fφ[ω∈Ω zω]]||λ//
Eλ(m(λ)) ={lim[V(zλ』)]-1∫(||Lλ[φ∈Φfφ[ω∈Ω zω]]||λ//

sup||Lλ[φ∈Φ fφ’[ω∈Ω zω]]||λ)m(λ) dV(zλ’)}1/m(λ)
||Lλ[φ∈Φfφ'[ω∈Ωzω]]||λをすする)m(λ)dV(zλ』)}1/m(λ)

(zλ’ → zλ)
(zλ』→zλ)

where
どこで

zλ is the domain of definition of the λth relation;
z λ λthの明確さの領域です 関係;

zλ’ is domain's approximations that have finite measures V(zλ’) and are domains of definition for the both integrals;
z λ』 領域のものは、有限処置Vを持っている近いものです(z λ』)、そして、明確さの領域は、両方の全体のためです;

m(λ) is a positive number, we shall take 1;
m(λ)は正数です、我々は1をとります;

in the denominator, a direct (not composite) function of independent variables is used and by determining the least upper bound, all different isometric transformations (conserving the norms)
分母において、独立変数のダイレクト(合成でない)機能が、使われて、最小上界(すべての異なる等尺性変化)をそばに決定しています(基準を節約すること)

||fφ’[ω∈Ω zω]||φ = ||fφ [ω∈Ω zω]||φ
||fφ'[ω∈Ω zω]||φ= ||fφ[ω∈Ωzω]||φ

of even equal elements are considered.
均一な同等の、要素は考慮されます。

The domain average power weighted unierror of a pseudo-solution to the combined relations may be defined as
複合関係の偽解決の力加重のunierrorが定義されるかもしれない領域平均

nE(m) = {∑λ∈Λ w(λ)[Eλ(m)]n // ∑λ∈Λ w(λ)}1/n
n E(m) ={∑λ∈Λw(λ)[Eλ(m)]n//∑λ∈Λw(λ)}1/n

by the law of the nth power
第n番目の力の法律によって

where
どこで

Λ is the set of indices l, which has the cardinality c(Λ);
Λ インデックスのセットがlであるならば、そしてそれは 基数c( Λ);

n = n(Λ)
n = n( Λ)

is a positive number whose value may be
価値があるかもしれない正数です

1, 2, c(Λ),
1(2)c( Λ)、

etc.
その他。

This unierror can be also defined as
このunierrorは、定義されることもできます

EΛ = supλ∈Λ Eλ(m(λ)).
E Λ =夕食をとってください λ∈Λ E λ(m( λ))。

For instance, let two pseudo-solutions to some combined four relations have the partial unierrors
たとえば、なんらかの複合4つの関係の2つの偽解決に部分的なunierrorsを持たせてください

1, 1, 1, 0;
1、1、1、0;

1, 0, 0, 0,
1、0、0、0、

respectively. Their Cantor sets are identical and equal to
それぞれ。 彼らのカントールセットは同一で、等しいです

{1, 0}
{1、 0}

(and are indefinite at all if their elements are inexact, which often happens), but it is intuitively obvious that the second pseudo-solution is much more precise than the first one. So let us consider just the generalized sets
(そして、彼らの要素が不正確であるならば、まったく明確でありません、しばしば起こります)しかし、第2の偽解決が最初のものより非常に正確であることは、直観的に明らかです。それで、まさに分化していないセットを考慮しよう

(3)
(3)

S1 = {1, 1, 1, 0}; 
S1 ={1、1、1、0}; 

S2 = {1, 0, 0, 0}
S2 ={1、0、0、0}

where the multiplicities of elements are included. But
要素の多様性が含まれるところ。しかし、

sup S1 = sup S2 = 1,
S1 =がS2をすすることをすすってください =1、

so it is necessary to determine just the generalized least upper bounds. The generalized subsets reduced from above are
それで、それはまさに全身性最小上界を決定するのに必要です。 上記から減らされる分化していないサブセットは、そうです

{1, 1, 1, 0};
{1、 1、1、0};

{1, 1, 0};
{1、 1、0};

{1, 0};
{1、 0};

{0}
{0}

and
そして、

{1, 0, 0, 0};
{1、 0、0、0};

{0, 0, 0};
{0、 0、0};

{0, 0};
{0、 0};

{0}.
{0}。

Their sets of least upper bounds coincide with the sets (3) themselves in such a case. The minimally reduced from above subsets having different usual least upper bounds 1 and 0 are
最小上界の彼らのセットは、そのような場合セット(3)自体と同時です。最小限に異なる普通の最小上界1と0がある上記のサブセットから減らすものは、そうです

{1, 1, 0}
{1、 1、0}

and
そして、

{0, 0, 0}.
{0、 0、0}。

So
そう

sup S1 > sup S2
S1をすすってください > S2をすすってください

and the second pseudosolution is estimated by EΛ (for
そして、第2のpseudosolutionは、EΛによって推定されます‖(

EΛ(n(Λ))
EΛ(n(Λ))

it is obvious) better than the first one as required.
それは明らかです)、最初のものよりよく必要に応じて。

 

 

4. Summing Methods for
4. Methodsを合計すること

Divergent Series
互いに異なるシリーズ

 

In particular, such methods for estimating unierrors provide many new methods to sum up divergent series [8, 9]
特に、unierrorsを推定するそのような方法は、互いに異なるシリーズを要約するために、多くの新しい方法を提供します[8、9]

iN ai
∑i∈Nミツユビナマケモノ

by obtaining constant A that ensures:
それが確実にする恒常的なAを得ることによって:

infA supn || A - (a1 + a2 + … + an )|| /
infA supn || A - (a1 + a2 +…+)|| /

(1 + || A - (a1 + a2 + … + an )||);
(1 + || A - (a1 + a2 +…+)||);

 

infA limn || A - (a1 + a2 + … + an )|| /
infAに、描いてください→ ∞ || A - (a1 + a2 + …+)|| /

(1 + || A - (a1 + a2 + … + an )||);
(1 + || A - (a1 + a2 +…+)||);

 

infA limn || A - (a1 + a2 + … + an )|| /
infAに、描いてください→ ∞ || A - (a1 + a2 + …+)|| /

(1 + || A - (a1 + a2 + … + an )||)
(1 + || A - (a1 + a2 +…+)||)

where lim is the upper limit;
limが上限であるところ;

 

infA supn (n-1i=1 n (|| A - i=1 n ai || /
infA supn(n-1 ∑i=1 n (|| A- ∑i=1 n ミツユビナマケモノ || /

(1 + || A - i=1 n ai ||))q(n))1/q(n);
(1 + || A- ∑i=1 n ミツユビナマケモノ ||))q(n))1/q(n);

 

infA limn(n-1i=1 n (|| A - i=1 n ai || /
infAに、描いてください→ ∞(n-1 ∑i=1 n(|| A - ∑i=1 n ミツユビナマケモノ || /

(1 + || A - i=1 n ai ||))q(n))1/q(n);
(1 + || A- ∑i=1 nミツユビナマケモノ ||))q(n))1/q(n);

 

infA limn(n-1i=1 n (|| A - i=1 n ai || /
infAに、描いてください→ ∞(n-1 ∑i=1 n(|| A - ∑i=1 n ミツユビナマケモノ || /

(1 + || A - i=1 n ai ||))q(n))1/q(n)
(1 + || A- ∑i=1 nミツユビナマケモノ ||))q(n))1/q(n)

where q(n) is a positive function of n, which can be, in particular, n itself or a constant q.
q(n)が良い面であるところはnの機能します。そして、それは、特に、n自体または恒常的なqでありえます。

 

 

5. Reserves
5. 予備役

 

But unierrors can be sensitive in principle to inexact pseudo-solutions and not to exact solutions that have different reliabilities, especially if they and relations themselves are inexactly known. For instance, both
しかし、特に彼らと関係が彼ら自身不正確に知られているならば、unierrorsは原則として不正確な偽解決に敏感でありえて、正確な解決に異なる信頼性をそんなに持つことがありえません。たとえば、両方とも

x1 = 1 + 10-10
x1 =1 + 10-10

and 
そして、 

x2 = 1 + 1010
x2 =1 + 1010

are exact solutions to the inequality
正確な解決は、不平等にあります

x > 1,
x > 1、

but x1 is practically dubious and x2 is guaranteed. So it is necessary to introduce some concept of a reserve
しかし、x1はほとんど疑わしいです、そして、x2は保証されます。それが蓄えの若干の概念を持ち出すのに必要であるように、

R: U [-1, 1]
R: U → [-1、1]

whose values satisfy the following basic axioms:
誰の価値が、以下の基本的な原理を満たしますか:

1) for every i I,
1) あらゆるiのために ∈ 私、

R(i) = - E(i) [-1, 0];
R(i) = - E(i)∈[-1、0];

2) for every e E,
2) あらゆるeのために ∈ E、

R(e) [0, 1];
R(e)∈[0、1];

3) there exist such an i I and an e E that
3) そのようなiが存在します ∈ 私 そして、e ∈ Eが、それです

R(i) = -1
R(i) =-1

and
そして、

R(e) = 1.
R(e) = 1。

For instance, the reserve of any pseudosolution x to the combined inequalities
たとえば、複合不平等へのどんなpseudosolution xの蓄えでも

[αΑ aα x bβ βΒ]
[α ∈Α aα ⇐ x ⇐ bβ β ∈Β]

where
どこで

a, b, x
b(x)

are real numbers;
実数です;

is one of the signs
⇐ サインのうちの1つです

<, ,
<、 ≤

can be defined as
定義されることができます

R(x, [αΑ aα x bβ βΒ]) =
R(x‖[α ∈Α aα ⇐ x ⇐ bβ β ∈Β] =

infαΑ, βΒ ((x - aα)/(|x|+2|aα|+1), (bβ - x)/(|x|+2|bβ|+1)).
infα ∈Α、 β ∈Β((x- aα)/(|x|+2|aα|+1)、 (bβ - x)/(|x|+2|bβ|+1))。

 

 

6. Pseudo-Solutions
6.偽Solutions

 

Thus reserves provide arranging all pseudo-solutions. If there exists a pseudo-solution that has the maximal reserve, that may be called a super-pseudo-solution. An exact super-pseudo-solution may be called a super-solution, an inexact one is called a quasi-solution. The super-solution to the inequalities
こ のように、蓄えはすべての偽解決を手配することを提供します。最大限の蓄えを持つ偽解決が存在するならば、それはスーパー疑似解決と呼ばれているかもしれ ません。正確なスーパー疑似解決はスーパー解決と呼ばれているかもしれません、不正確なものは準解決と呼ばれています。不平等のスーパー解決

(6)
(6)

a0 x a1
a0 ⇐ x ⇐ a1

is
あります

xs = 0.25 (((2 + 2|a0| + 2|a1| - |a0 + a1|)2 + 8|a0 + a1|)1/2
xs = 0.25(((2 + 2つの|a0| + 2つの|a1| - |a0 + a1|)2 + 8|a0 + a1|)1/2

- 2 - 2|a0| - 2|a1| + |a0 + a1|) sign (a0 + a1).
- 2 - 2つの|a0| -2つの|a1| + |a0 + a1|)、署名してください(a0 + a1)。

If there exists a pseudo-solution that has the minimal reserve and is inexact, it may be called an anti-solution. The anti-solutions to the same inequalities (6) are
最小の蓄えを持って、不正確である偽解決が存在するならば、それは反解決と呼ばれているかもしれません。同じ不平等(6)の反解決は、そうです

xA = -
xA =- ∞

if
もしも

a1 > |a0|
a1 > |a0|

or
あるいは、

(a1 = |a0|
(a1 = |a0|

and (6) has a form
そして、(6)には形があります

-|a0| < x |a0|);
-|a0| < x ≤ |a0|);

xA = +
xA = + ∞

if
もしも

a0 < -|a1|
a0 <-|a1|

or
あるいは、

(a0 = -|a1|
(a0 = ― |a1| ―

and (6) has a form
そして、(6)には形があります

-|a0| x < |a0|);
-|a0| ≤ x< |a0|);

xA = ±
xA = ± ∞

if
もしも

a0 = -a1
a0 =-a1

and (6) has a form
そして、(6)には形があります

-|a0| < x < |a0|
-|a0| < x< |a0|

or
あるいは、

-|a0| x |a0|.
-|a0| ≤ x ≤ |a0|。

If the set of pseudo-solutions is compact, super-quasi-solutions exist. They can be determined by the method of equalizing the partial reserves of a pseudo-solution to separate relations.
偽解決のセットがコンパクトであるならば、スーパー-準-解決が存在します。 彼らは、関係を切り離すために偽解決の部分的な蓄えを等しくする方法で測定されることができます。

 

 

7. Probabilities
7.可能性

 

Estimations are often given by corresponding probabilities. It is possible to generalize the concept of a probability [10] that is locally logical because every elementary outcome gives either 0 (if it is not favorable) or 1 (if it is favorable). So a usual probability may be considered as a logical reserve that is equal to 1 for exact solutions and to 0 for inexact pseudo-solutions.
評 価は、対応する可能性によってしばしばされます。あらゆる基本の結果が0(それが有利でないならば)か1(それが有利であるならば)を与えるので、地元で 論理的である可能性[10]の概念を一般化することは、可能です。それで、普通の可能性は、正確な解決のための1まで、そして、不正確な偽解決のための0 まで等しい論理的蓄えと思われるかもしれません。

Let us consider the conditional probability
条件付き確率を考慮しよう

P(s|p)
P(s|p)

that some pseudo-solution is an exact solution. When using uniform distributions on
とても若干の偽解決は、正確な解決です。やっている均一な分布を使うとき、

(-, )
(- ∞ ∞)

or
あるいは、

(-π/2, π/2) χ = arctan x,
(-π/2、 π/2) ∋ χ =逆正接関数x、

the probability of holding the following inequalities is:
以下の不平等を持つ可能性は、以下の通りです:

x < a
x <: 

P(s|p) = 1/2 + 1/2 × a/  
P(s|p) = 1/2 + 1/2 ×a/ ∞  

or   
あるいは、   

P(s|p) = 1/2 + 1/π  arctan a;
P(s|p)= 1/2 + 1/π逆正接関数;

 

x > a
x > : 

P(s|p) = 1/2 - 1/2 × a/  
P(s|p) = 1/2 - 1/2 ×a/ ∞  

or   
あるいは、   

P(s|p) = 1/2 - 1/π  arctan a;
P(s|p) = 1/2 - 1/π逆正接関数;

 

a0  < x  <  a1
a0  < x< a1: 

P(s|p) = 1/2 × (a1 - a0)/  
P(s|p) = 1/2 ×(a1 - a0)/ ∞  

or   
あるいは、   

P(s|p) = 1/π × (arctan a1 - arctan a0).
P(s|p) = 1/π ×(逆正接関数a1 - 逆正接関数a0)。

When using uniform distributions on
やっている均一な分布を使うとき、

(-, )
(- ∞ ∞)

or
あるいは、

(-π/2, π/2) χ = arctan x
(-π/2、 π/2) ∋ χ =逆正接関数x

corresponding to more sensitive reserves generalizing probabilities, the expected value of the reserve (5) of the inequalities (6) is
可能性を一般化しているより敏感な蓄えと一致して、不平等(6)の予備(5)の期待値は、そうです

M R(x, a0 < x < a1) = (2π)-1 ×
M R(x、a0<は< a1に×印を付けます) =(2π)-1 ×

i=0 1 (((-1)i + sign xs)ailn(1 + 2|ai|)/(1 + 2|ai| + 2ai2) -
∑i=0 1(((-1)i +のサインxs)ailn(1 + 2つの|ミツユビナマケモノ|)/(1 + 2|ミツユビナマケモノ| + 2ai2)-

0.25(1 + (-1)i (ai + 2ai|ai|)/(1 + 2|ai| + 2ai2)) +
0.25(1 +私(-1)(ミツユビナマケモノ+ 2ai|ミツユビナマケモノ|)/(1 + 2つの|ミツユビナマケモノ| + 2ai2))+

(2π)-1(arctan xs)i=0 1 (sign xs - (ai + 2ai|ai|)/(1 + 2|ai| + 2ai2) +
(2π)-1(逆正接関数xs) ∑i=0 1(xsに署名してください - (ミツユビナマケモノ+ 2ai|ミツユビナマケモノ|)/(1 + 2|ミツユビナマケモノ| + 2ai2)+

(4π)-1i=0 1 (1 + ai sign xs + 2|ai|)ln((1 + xs2)/
(4π)-1 ∑i=0 1(1 +ミツユビナマケモノ徴候xs + 2|ミツユビナマケモノ|)ln((1 + xs2)/

(1 + 2|ai| + 2ai2))/(1 + 2|ai| + 2ai2).
(1 + 2つの|ミツユビナマケモノ| + 2ai2))/(1 + 2|ミツユビナマケモノ| + 2ai2)。

Besides that, the usual probability, initial and central moments [10] are not sensitive to the incompleteness of information. For example, each of them gives identical results when one ball is extracted from a box having white and black balls in equal or unknown portions. It is conditioned by the first power of probabilities (for discrete random variables) or of their densities (for continuous ones). Hence one may utilize (usual or normed) initial and centra1 moments for which this power is not equal to 1.
そ れの他に、普通の可能性、イニシャルと中心瞬間[10]は、情報の不完全さに敏感でありません。たとえば、1つのボールが等しいか未知の部分で白くて黒い ボールを備えている箱から引き抜かれるとき、彼らの各々は同一の結果を与えます。それは、可能性(別々の確率変数のために)の、または、彼らの密度(連続 もののために)の最初の力によって条件づけられます。それゆえに、人はこの力が1と等しくない(普通であるかノルム)最初のおよびcentra1瞬間を利 用するかもしれません。

 

 

Bibliography
参考文献

 

[1]       Barford, N. C. Experimental Measurements: Precision, Error, and Truth. Addison-Wesley, 1967
[1]        バーフォード、N. C.実験的な寸法: 精度、エラーと真実。アディソン‐ウェズリー、1967

[2]       Taylor, J. R. An Introduction to Errors Analysis. University Science Books Mill Valley, California, 1982
[2]        エラーにテイラー(J. R)。導入 分析。大学科学本が、谷、カリフォルニア、1982を粉にします

[3]       Cantor, G. Gesammelte Abhandlungen mathematischen und philosophischen Inhalts. Berlin, 1932
[3]        カントール、G. Gesammelte Abhandlungen mathematischen und philosophischen Inhalts。ベルリン、1932

[4]       Hausdorff, F. Grundzüge der Mengenlehre. Leipzig, 1914
[4]        ハウスドルフ、F. Grundz ge der Mengenlehre.ライプツィヒ、1914

[5]       Weierstraß, K. T. W. Mathematische Werke. Berlin, Leipzig, 1894-1927, Bd. 1-7
[5]        Weierstra、K. T. W. Mathematische Werke.ベルリン、ライプツィヒ、1894-1927、Bd. 1-7

[6]       Gelimson, L. G. Generalization of Analytic Methods for Solving Strength Problems [In Russian]. Drukar Publishers, Sumy, 1992
[6]        Gelimson(L. G)。Generalizationの 強さ問題[ロシア語で]を解決する分析的方法。Drukar出版者、Sumy、1992

[7]       Gelimson, L. G. General Strength Theory. Drukar Publishers, Sumy, 1993
[7]        Gelimson、L. G.将軍強さ 理論。Drukar出版者、Sumy、1993‖

[8]       Borel, E. Lecons sur les Séries Divergentes. Paris, 1928
[8]        ボレル、レズS riesに基づくE. Lecons Divergentes.パリ、1928

[9]       Cooke, G. Infinite Matrices and Sequence Spaces. London, 1950
[9]        クック、G.無限のマトリックスとシーケンス スペース。ロンドン(1950)

[10] Fréchet, M. Recherches théoriques modernes sur la théorie des probabilites. Paris, 1937-1938
[10] Fr chet、M. Recherches th oriques ラth orie 12月probabilitesに基づくmodernes。パリ、1937~1938