Available online Journal of Scientific and Engineering Research, 2014, 1(1): Research Article

Similar documents
Establishing Relations among Various Measures by Using Well Known Inequalities

Key words: Fractional difference equation, oscillatory solutions,

Continuous Indexed Variable Systems

Series of New Information Divergences, Properties and Corresponding Series of Metric Spaces

Some Probability Inequalities for Quadratic Forms of Negatively Dependent Subgaussian Random Variables

14. Poisson Processes

The Mean Residual Lifetime of (n k + 1)-out-of-n Systems in Discrete Setting

VARIATIONAL ITERATION METHOD FOR DELAY DIFFERENTIAL-ALGEBRAIC EQUATIONS. Hunan , China,

THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Series A, OF THE ROMANIAN ACADEMY Volume 10, Number 2/2009, pp

JORIND 9(2) December, ISSN

CONTROLLABILITY OF A CLASS OF SINGULAR SYSTEMS

Quantum Mechanics II Lecture 11 Time-dependent perturbation theory. Time-dependent perturbation theory (degenerate or non-degenerate starting state)

The Poisson Process Properties of the Poisson Process

Solution to Some Open Problems on E-super Vertex Magic Total Labeling of Graphs

A Remark on Generalized Free Subgroups. of Generalized HNN Groups

Comparison of the Bayesian and Maximum Likelihood Estimation for Weibull Distribution

The Bernstein Operational Matrix of Integration

On cartesian product of fuzzy primary -ideals in -LAsemigroups

Moments of Order Statistics from Nonidentically Distributed Three Parameters Beta typei and Erlang Truncated Exponential Variables

RATIO ESTIMATORS USING CHARACTERISTICS OF POISSON DISTRIBUTION WITH APPLICATION TO EARTHQUAKE DATA

QR factorization. Let P 1, P 2, P n-1, be matrices such that Pn 1Pn 2... PPA

Complementary Tree Paired Domination in Graphs

The algebraic immunity of a class of correlation immune H Boolean functions

Bianchi Type II Stiff Fluid Tilted Cosmological Model in General Relativity

Fourth Order Runge-Kutta Method Based On Geometric Mean for Hybrid Fuzzy Initial Value Problems

International Journal Of Engineering And Computer Science ISSN: Volume 5 Issue 12 Dec. 2016, Page No.

Asymptotic Regional Boundary Observer in Distributed Parameter Systems via Sensors Structures

COMPARISON OF ESTIMATORS OF PARAMETERS FOR THE RAYLEIGH DISTRIBUTION

Asymptotic Behavior of Solutions of Nonlinear Delay Differential Equations With Impulse

Real-Time Systems. Example: scheduling using EDF. Feasibility analysis for EDF. Example: scheduling using EDF

Solution of Impulsive Differential Equations with Boundary Conditions in Terms of Integral Equations

8. Queueing systems lect08.ppt S Introduction to Teletraffic Theory - Fall

Chapter 8. Simple Linear Regression

MAX-MIN AND MIN-MAX VALUES OF VARIOUS MEASURES OF FUZZY DIVERGENCE

Probability Bracket Notation and Probability Modeling. Xing M. Wang Sherman Visual Lab, Sunnyvale, CA 94087, USA. Abstract

(1) Cov(, ) E[( E( ))( E( ))]

Cyclically Interval Total Colorings of Cycles and Middle Graphs of Cycles

Stability Criterion for BAM Neural Networks of Neutral- Type with Interval Time-Varying Delays

AML710 CAD LECTURE 12 CUBIC SPLINE CURVES. Cubic Splines Matrix formulation Normalised cubic splines Alternate end conditions Parabolic blending

Brownian Motion and Stochastic Calculus. Brownian Motion and Stochastic Calculus

Regression Approach to Parameter Estimation of an Exponential Software Reliability Model

Fully Fuzzy Linear Systems Solving Using MOLP

( 1)u + r2i. f (x2i+1 ) +

A note on Turán number Tk ( 1, kn, )

Optimality of Distributed Control for n n Hyperbolic Systems with an Infinite Number of Variables

Orbital Euclidean stability of the solutions of impulsive equations on the impulsive moments

Real-time Classification of Large Data Sets using Binary Knapsack

Redundancy System Fault Sampling Under Imperfect Maintenance

Efficient Estimators for Population Variance using Auxiliary Information

Integral Φ0-Stability of Impulsive Differential Equations

Other Topics in Kernel Method Statistical Inference with Reproducing Kernel Hilbert Space

The Linear Regression Of Weighted Segments

ON TESTING EXPONENTIALITY AGAINST NBARFR LIFE DISTRIBUTIONS

Random Generalized Bi-linear Mixed Variational-like Inequality for Random Fuzzy Mappings Hongxia Dai

Determination of Antoine Equation Parameters. December 4, 2012 PreFEED Corporation Yoshio Kumagae. Introduction

SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β

Least Squares Fitting (LSQF) with a complicated function Theexampleswehavelookedatsofarhavebeenlinearintheparameters

Use of Non-Conventional Measures of Dispersion for Improved Estimation of Population Mean

INTERNATIONAL JOURNAL OF ENGINEERING SCIENCES & RESEARCH TECHNOLOGY

-distributed random variables consisting of n samples each. Determine the asymptotic confidence intervals for

Application of the stochastic self-training procedure for the modelling of extreme floods

Solving fuzzy linear programming problems with piecewise linear membership functions by the determination of a crisp maximizing decision

4. Runge-Kutta Formula For Differential Equations

Convexity Preserving C 2 Rational Quadratic Trigonometric Spline

FALL HOMEWORK NO. 6 - SOLUTION Problem 1.: Use the Storage-Indication Method to route the Input hydrograph tabulated below.

Midterm Exam. Tuesday, September hour, 15 minutes

New approach for numerical solution of Fredholm integral equations system of the second kind by using an expansion method

Fundamentals of Speech Recognition Suggested Project The Hidden Markov Model

Binary Time-Frame Expansion

An Efficient Dual to Ratio and Product Estimator of Population Variance in Sample Surveys

. The set of these sums. be a partition of [ ab, ]. Consider the sum f( x) f( x 1)

4. Runge-Kutta Formula For Differential Equations. A. Euler Formula B. Runge-Kutta Formula C. An Example for Fourth-Order Runge-Kutta Formula

On Metric Dimension of Two Constructed Families from Antiprism Graph

Some Improved Estimators for Population Variance Using Two Auxiliary Variables in Double Sampling

FORCED VIBRATION of MDOF SYSTEMS

Beyond matched pairs and Griliches-type hedonic methods for controlling quality changes in CPI sub-indices

Cyclone. Anti-cyclone

Mixed Integral Equation of Contact Problem in Position and Time

Mathematical and numerical modeling of inverse heat conduction problem

NON-IDEMPOTENT PLONKA FUNCTIONS AND WEAKLY PLONKA SUMS

1 Introduction and main results

An Application of Generalized Entropy Optimization Methods in Survival Data Analysis

Bounds for the Connective Eccentric Index

Interval Estimation. Consider a random variable X with a mean of X. Let X be distributed as X X

Study on one-dimensional consolidation of soil under cyclic loading and with varied compressibility *

The Properties of Probability of Normal Chain

Pricing Asian Options with Fourier Convolution

The textbook expresses the stock price as the present discounted value of the dividend paid and the price of the stock next period.

The textbook expresses the stock price as the present discounted value of the dividend paid and the price of the stock next period.

Quantitative Portfolio Theory & Performance Analysis

Upper Bound For Matrix Operators On Some Sequence Spaces

On generalized fuzzy mean code word lengths. Department of Mathematics, Jaypee University of Engineering and Technology, Guna, Madhya Pradesh, India

To Estimate or to Predict

A Comparison of AdomiansDecomposition Method and Picard Iterations Method in Solving Nonlinear Differential Equations

On Eccentricity Sum Eigenvalue and Eccentricity Sum Energy of a Graph

Solving Fuzzy Equations Using Neural Nets with a New Learning Algorithm

4. THE DENSITY MATRIX

Supplement Material for Inverse Probability Weighted Estimation of Local Average Treatment Effects: A Higher Order MSE Expansion

Isotropic Non-Heisenberg Magnet for Spin S=1

IMPROVED PORTFOLIO OPTIMIZATION MODEL WITH TRANSACTION COST AND MINIMAL TRANSACTION LOTS

Transcription:

Avalable ole wwwjsaercom Joural o Scec ad Egeerg Research, 0, ():0-9 Research Arcle ISSN: 39-630 CODEN(USA): JSERBR NEW INFORMATION INEUALITIES ON DIFFERENCE OF GENERALIZED DIVERGENCES AND ITS APPLICATION KC Ja, Praphull Chhabra Deparme o Mahemacs, Malavya Naoal Isue o Techology, Japur- 3007 (Rajasha), INDIA Absrac Dvergece measures are useul or comparg wo probably dsrbuos Depedg o he aure o he problem, he dere dvergeces are suable So s always desrable o creae a ew dvergece measure I hs work, ew ormao equales, correspodg o derece o wo geeralzed - dvergeces, are obaed ad characerzed Secodly, we oba ew dvergece measure correspodg o ew covex uco ad dee he properes Furher, bouds o ew dvergece erms o oher sadard dvergeces are evaluaed Comparso o hs dvergece wh ohers s doe as well Idex erms: New Covex ad ormalzed uco, New dvergece measure, Comparso graph o dvergeces, New ormao equales, Bouds o ew dvergece Mahemacs Subjec Classcao: Prmary 9A7, Secodary 6D5 Iroduco P p, p, p3, p : p 0, p, Le be he se o all complee e dscree probably dsrbuos I we ake p 0 or some,, 3,,, he we have o suppose 0 ha 0 0 0 0 0 Csszar s - dvergece [] s a geeralzed ormao dvergece measure, whch s gve by (), e, p C P, q q () P p E C PC p q q Ad,,, () Smlarly (Ja ad Saraswa [5]) roduced a geeralzed measure o ormao gve by p q S P, q q (3) Joural o Scec ad Egeerg Research 0

Joural o Scec ad Egeerg Research, 0, ():0-9 Where : (0,) R (se o real o) s real, couous ad covex uco ad,,,,,,, P p p p p q q q q 3 3 Γ, where p ad q are probably mass ucos May kow dvergeces ca be obaed rom hese geeralzed measures by suably deg he covex uco Some o hose are as ollows Ch- square dvergece [8] = P, Relave JS dvergece [9] = F p q () q p, p log p q Relave J- dvergece [3] = J p q Relave AG dvergece [0] = G (5) p q, log q R Tragular dscrmao [] = P, J- dvergece [6,7] = J p q (6) p q p q, log p (7) p q p, log (8) p q (9) q We ca see ha JR P, F, P G, P, P, W P,,,,,whereW J J J P R R, pq p q ad s Harmoc mea dvergece Dvergeces rom () o (7) are o- symmerc ad (8), (9) are symmerc, wh respec o probably dsrbuo P, Γ () ad (9) are also kow as Pearso dvergece ad Jereys- Kullback- Lebler dvergece, respecvely Besde hese, Symmerc Ch- square dvergece [] ca be wre as he sum o Ch- square dvergece ad s adjo, e, P,, P P, p q p q (0) pq New Iormao Iequales I hs seco, we roduce ew ormao equales o derece o wo geeralzed - dvergeces Such equales are or sace eeded order o calculae he relave ececy o wo dvergeces Theorem Le : I R R, 0 ad suppose he assumpos: a be wo covex ad ormalzed ucos, e ad are wce dereable o (α, β) where 0, b There exss he real cosas m, M such ha m < M ad, 0, m M, () Joural o Scec ad Egeerg Research

I P,, he we have he equales, Joural o Scec ad Egeerg Research, 0, ():0-9,,,,,, me S E S M E S () Where E P,, S P, Proo: Le us cosder wo ucos ad m are gve by () ad (3) respecvely F m, (3) M F M () Where m ad M are he mmum ad maxmum values o he uco Sce Fm FM ad he ucos ad ad,, 0 0, (5) are wce dereable The vew o (), we have 0, (6) Fm m m 0 (7) FM M M I vew (5), (6) ad (7), we ca say ha he ucos β) Now, wh he help o leary propery, we have, Fm ad FM are ormalzed ad covex o (α, E,,,, F m SF E m m S m E P, me,,, S ms ad E,,,, F M SF E M M S M ME P, E,,, MS S Sce E P, S P,, (8) (9) rom [5], hereore (8) ad (9) ca be wre as he ollowgs E P, S P, me P, S P, 0,,,,, 0 ad M E S E S Or E P, S P, m E P, S P, (0), ad M E P, S P, E P, S P, () (0) ad (), ogeher gve he resul () New Dvergece Measure ad Properes Joural o Scec ad Egeerg Research

Joural o Scec ad Egeerg Research, 0, ():0-9 I hs seco, we oba ew dvergece measure or ew covex uco; urher dee he properes o ew covex uco ad ew dvergece Frsly, Le : 0, ad R be a uco deed as 3, 0,, 0,, (3) 3 (3) 3 Properes o uco deed by (3), are as ollows a Sce 0 s a ormalzed uco s a covex uco as well b Sce 0 0, c Sce d a 0 mooocally creasg 0, ad a, s mooocally decreasg 0, ad 00 000 0, 8 0 0, ad 800 600 00 00 6 8 0 Fgure 3: Covex uco Now pu ad (3) ad () respecvely, we ge he ollowg ew dvergece measure,,,, S E S 3 3 p q 3p 3p q p q 6 pq 8q (33) 8p q p q Properes o ew dvergece measure deed (33), are as ollows a S P, s covex ad o- egave he par o probably dsrbuo S P, 0 P or p q (Aas s mmum value) b P, Joural o Scec ad Egeerg Research 3

c Sce S P, S, P S, Joural o Scec ad Egeerg Research, 0, ():0-9 s o- symmerc dvergece measure Applcao o New Iormao Iequales I hs seco, we oba bouds o ew dvergece measure (33) by usg ew equales deed (), erms o sadard dvergeces Proposo Le P,, J P,, J P, ad S, (33) respecvely For P, Γ, we have a I 0 065, he 863 J P,, P J P, S P, R R be deed as (), (6), (9) ad 3 3 max, J P,, P J P, R () b I 065, he 3 J P,, P J P, S P, R 3 J P,, P J P, R () Proo: Le us cosder log, 0,, 0, log ad (3) Sce 0 0 ad 0, so s covex ad ormalzed uco respecvely Now, Pu (3) ad Now, le g Ad (), we ge he ollowgs p q p q S J R q (), log, p p q, log,, (5) E p q J P q p 3 are gve by (3) ad (3) respecvely, where ad 5 6 9 g, g 3 3 3 Joural o Scec ad Egeerg Research

I g 0 069793 065 I s clear ha g () s decreasg (0, 065] ad creasg (065,) Joural o Scec ad Egeerg Research, 0, ():0-9 Also g () has a mmum value a =065, because g 065 30035 0 Now, a I 0 065, he, m g g 065 863 (6) 3 3 M sup g max g, g max,, b I 065, he (7) 3 m g g (8), 3 M sup g g (9), The resuls () ad () are obaed by usg (33), (), (5), (6), (7), (8) ad (9) () Proposo Le P,, F P, ad S, P, Γ, we have a I 0 058, he P F PS 68,,, 3 3 max,, P F, P b I 058, he 3,,, P F PS P F P 3,, Proo: Le us cosder Sce Pu be deed as (), (5) ad (33) respecvely For (0) () log, 0,, 0, ad () 0 0 ad 0, so s covex ad ormalzed uco respecvely Now, (3) ad (), we ge he ollowgs q S q F P p q (3), log, Joural o Scec ad Egeerg Research 5

E Now, le g Ad P, Joural o Scec ad Egeerg Research, 0, ():0-9 p q q q pq q p q p p q p p p p p q q p, P () p 3 9 g, g 9 3 g 0 05773 058 0 I are gve by (3) ad () respecvely, where ad I s clear ha g () s decreasg (0, 058] ad creasg (058,) Also g () has a mmum value a =058, because g 058 38 0 Now, a I 0 058, he, m g g 058 68 (5) 3 3 M sup g max g, g max,, b I 058, he (6) 3 m g g (7), 3 M sup g g (8), The resuls (0) ad () are obaed by usg (33), (3), (), (5), (6), (7) ad (8) () Proposo 3 Le,,, P, Γ, we have a I 0 076, he G J ad S, 698 J P, G, P S P, max,,, b I 076, he 3 3 J G P 3 J P, G, P S P,,, Proo: Le us cosder 3 J G P be deed as (7), (9) ad (33) respecvely For (9) (0) Joural o Scec ad Egeerg Research 6

Sce Pu log, 0,, 0, log ad Joural o Scec ad Egeerg Research, 0, ():0-9 () 0 0 ad 0, so s covex ad ormalzed uco respecvely Now, (3) ad Now, le g Ad (), we ge he ollowgs p q p q S G P q (), log, p E p q J q (3), log, 3 3 g, g 3 I g 0 07598 076 0 are gve by (3) ad () respecvely, where ad I s clear ha g () s decreasg (0, 076] ad creasg (076,) Also g () has a mmum value a =076, because g 076 796 0 Now, a I 0 076, he, m g g 076 698 () 3 3 sup max, max, M g g g, b I 076, he (5) 3 m g g, (6) 3 M sup g g (7), The resuls (9) ad (0) are obaed by usg (33), (), (3), (), (5), (6) ad (7) () Proposo Le P,, P, ad S, P, Γ, we have 3, P U P, P, S P, 3, P U P, P, be deed as (), (8) ad (33) respecvely For (8) Joural o Scec ad Egeerg Research 7

Proo: Le us cosder Sce Pu, 0,, 0, Joural o Scec ad Egeerg Research, 0, ():0-9 ad (9) 3 0 0 ad 0, so s covex ad ormalzed uco respecvely Now, (3) ad (), we ge he ollowgs p q (30) S P, P, p q E P, where U P, p q p q p q p q q p p p =, P U P, (3) p q q p 3 Now, le g 3 ad g 0 0 ad (9), respecvely I s clear ha g () s creasg 0,, so, are gve by (3), where ad m g g 3 (3), M sup g g 3 (33) The resul (8) s obaed by usg (33), (30), (3), (3) ad (33) () Fgure shows he behavor o S P,, P,, J P,, P, ad P, cosdered p a, a, q a, a, where 0, dvergece S P, has a seeper slope ha P,, J P,, P, ad P, We have a I s clear rom gure ha he ew Joural o Scec ad Egeerg Research 8

Joural o Scec ad Egeerg Research, 0, ():0-9 Reereces Csszar, I, Iormao ype measures o dereces o probably dsrbuo ad drec observaos, Suda Mah Hugarca, Vol, pp 99-38, 967 Dacuha- Caselle D, Ecole d Ee de Probables de, Sa-Flour VII-977, Berl, Hedelberg, New York: Sprger, 978 3 Dragomr SS, Gluscevc V, Pearce CEM, Approxmao or he Csszar -dvergece va mdpo equales, equaly heory ad applcaos - YJ Cho, JK Km ad SS Dragomr (Eds), Nova Scece Publshers, Ic, Hugo, New York, Vol, 00, pp 39-5 Dragomr SS, Sude J ad Buse C, New equales or Jereys dvergece measure, Tamus Oxord Joural o Mahemacal Sceces, 6() (000), 95-309 5 Ja K C, Saraswa RN, Some ew ormao equales ad s applcaos ormao heory, Ieraoal joural o mahemacs research, Volume, Number 3 (0), pp 95-307 6 Jereys: A vara orm or he pror probably esmao problem Proc Roy Soc Lo Ser A, 86(96), 53-6 7 Kullback S ad Lebler RA, O Iormao ad Sucecy, A Mah Sas, (95), 79-86 8 Pearso K, O he Crero ha a gve sysem o devaos rom he probable he case o correlaed sysem o varables s such ha ca be reasoable supposed o have arse rom radom samplg, Phl Mag, 50(900), 57-7 9 Sbso R, Iormao radus, Z Wahrs Udverw Geb, () (969),9-60 0 Taeja IJ, New developmes geeralzed ormao measures, Chaper : Advaces Imagg ad Elecro Physcs, Ed PW Hawkes, 9(995), 37-35 Joural o Scec ad Egeerg Research 9