| v or scuaamerconagg po (CONTENTS CHAPTER? ma raptors Pages 21 Definitions, ay CHAPTER 0 22 Convex Combination, = CHAPTER 02 pe 33 Soong a ‘Mathematical Preliminaries Ea Exercise on Chapter I 3 “Matrices and Determinants 2 mare I 3 fae j 3 tions of Matrix and Addit and Multipli a. = : es artetnttinsnhtninia® = deo : aa =a QS Minor of order k. 3 32 Slack and Surplus Variables, 35 06 Determinant. sas 3 3 Fenton 20d Notations. % ea ; So a ee i= é ace ‘ ae acre tM BeSeeiemdins 3 O11 Adjoit of a Matrix. 5 38 Optimaliy Conditions, 2 BES si Vp 3) cee . is RS pe i ieee oe ‘Vectors and Vector Spaces Bal “To determine starting BFS. 6 a oo Bh: ee oe eg ie 4 » Saeed = HE Eee mmimmerem ||: 44! RENEE | ae Hs Beas atts Je ae Bo! aoe ee a ne 2 ig © 2 conembeniaetiae 4 ciate meets ert | Selig, Se ceme anemone aes edits tems Ereretiaetarsie arent 0 seme ea : aa ee = ec tenerertaer i ac eat ge eee = Coepiae Toto cara sane Goreme ee | Gree it. (Formulation and Graphical Solution) ‘a rea ‘of Degeneracy ms s eal ae 2 TE napeaig ee Te ER weocenastemngmatze 13. Mathematea! formation of LEE B 43° Method of Resing Degenea. 0 ' 2 ee yo Ro Redes tee 7 15 An Important Theorem. a 45 Selection of the outgoing (departing) vector. 1 1s Some Important Theorems, » 45, Computational Procedare 135 17 Solution ota inest programming problem. B 47 Generalized Simplex Method. 136 18 Geometrical (or graphical) method for the solution of ‘Exercise on Chapter I ~ Exercise on Chapter 1. a CHAPTER S$ Revised Simplex Method Te Scanned with CamScannervat x2 Solution of transportation problem. 3 ig Staion cable somo 4 BS Gotimaity est ey 1d Thee octane of pinay te 15, Compaen Nth or ct Med By j HAD Bopehcracy in Tranepotatin Problems x Ha Beisel tanspraon is BE Benes eet ee Mathemati iminari aR ematical Preliminaries =- 3 §0-1. Introduction Compete Steg) ey Introduction. In this chapter we shall cass some base. ree, 2 concins and techniques of + renters athena ie ma ad HE Giese Game o fucllate the reader in the: dewopment and andetesdag tthe BR. ester inite Games 3 theorsicl and computational spear of facs prosuosnn Buseed 124 Zeon EMME, (or Rectangult) Gases, s inthe sucoceding chapters BS: Rese 3 : 1 Matries an Deterfants. Bg Poms BG $02. Detoons By ram aon 3 f 1, Mati “A mois & o rcensla any of mx 8 number BS Metin ss : 1} eonged in m rows and m coleman os tone Ea er rf Cenpattine gu Som @ | wag PEE Be rene iat sale poi ot AP lon ao an | Gilmnn By 2x mand x 2games 414 a Bu wekaceseal Game. 4 Gm Ona Say ow a fmouigame andiocar y ‘yale the clement ofthe mcd inh row and 2h colina, i 2 Square Mate I'm =, maid is alle a square manic tam Folemeatteoum Gane eon sigs Th) 1 > at cede a. 2 oS, Sa by amex 2 3. Digosal Mati. A agure matric in wich ofiggonat HR Senna lee a lab 3 ements, ay for, ar all aval o sero eled a dagonal mate He Mi sta fanaon eer arable 4. Unt (or Leni) Matric. A (or Idetiy) mati» Bah abe rade conto ors ocion Ef] diagoal mts whose dagoa lene eet ner word SE et ss 2 sare math beled sunt ma Way= Ofer! ej andey™ lor = f-4 unit matic of order m is denoted by Iy 0 simply Z Toe Tus [2 f] omts=]o 2) ce fou 5. Tanja Matis, Aste mati is ead 9 ong mans al oy =0 for >} al og=0 he <) Noll Mate A matics led ana mare als clement are equal tose. Scanned with CamScannerLinear P : BraMing 1. Transposed Matrix. The vanpore of matrix A is a at obtained ty interchanging is rows and cols and is denoted by 1 oa’ iF the mate A is of order mXn then the order of 4% 4, 1. gaat of Matis. Two mats A= [a] and B = ‘are suid to be equal if : (@) they are ofthe same order, and (i). 95 = by, for alli andj. 0:3. Operations of Matrix Addition and Multiplication, 1. Addition of Matrices. Let A= [og] and B= [oj be two maltices of the same ordey, hes the sum of hese wo macs denoted by A + Bis given by Coat B= [oj j] a pit nde of C=A + B sequal othe order of the matices 4 is important to note that the addition of matrices satisfy the @ Dates (Commutative tiny @ G++c=4404 (Associative @ Gsyeese Q ms 2 Multiplication oa mati ya scalar ij be a matrix of order m Xn and ¢ be a seala, then ‘he product of the matrix A ty c, denoted by cis given by i oak alwng ropes al fw mlion of mai © ramen soa Beno of mais © eas reanen (Ost ‘3. Multiplication of Matrices, tet lag td B= gl be 99 mais of te ode em Note, The multiplication. he nerf cls of ca Aan Bis defined the rows of matix B. natrix A is equal to the number ee ‘Mathematical Preliminaries 3 ‘The propertics of matric muliplcation ae as follows: (4B) C= 4(8Q) (Associative lw) @) A+B C=Act Bc (Distributive law) (i) CA+B)=cA 408 @) 9(AB)=(04)B =A (aR) @) =A=ar (W)_(ABy = Ba", 1 is important to note that the matrix multiplication is sot ‘commutative in genera. iz, ingeneral ABBA. 0-4, Sub-matris. If we cross out ll but rows and s-columas of anm x m matrix A, the resulting rs matrix is called a sub-mmaoi of A 50-5. Minor of order k. For aim xm matrix A if Ris the kth ‘order sub-matrx obtained by deleting all but k-rows and k-cokemas of ‘A. Then |B | is called a kth onder minor of A. §.0°6. Determinant. There is a number associated with every square matric and is calle the determinant ofthe matrix Del, The deeminant of square matrix A= [oj] of order n, writen as| A | is defined tobe the number computed from the following sum involving then? elements i. IAL 5 (2) aya) the sum being taken ovr all permutations of the second subscripts. A term & assigned a plus signif (yjy.or) is an even permutation of (1,2 yom )y nd minus signifi & an edd permutation. e182, 182112939 ~ 22Arre FanMndss + anener= Menem ‘Ths cl ego fh detrital wth espe tthe Ft "50-7. Important properties of determinants, Now we give impntaa Fopenes of We datas (abou eet Thal mathe Scanned with CamScannerLinear Progaming verified from defisiton ofthe determinant 1. Mfevery clement of a row or column of a determinant is zero, the the value of the determinant is zero. 2, IF]B | isthe determinant formed by interchanging any two rows ge columns ofa determinant | 4 |, then 1Bl=~|A| 3. The value ofa determinant is not changed if ts rows and columns are interchanged. 4. Iftworows or columas of a determinant are identical then the vale ofthe determinant is 20. 5 Homyeban dt asmvor cio emia ba sia Then hea ofthe deternan mul © creat aro clan rate Snes he argon cnet a note ow sane ‘the value of the determinant is not changed. i sable mee deme ei ithe deta indy detg tow [4 eer eyeing hon ve l41=/en an oy 4 em. a3, then My minor of element is gen by Mya [es]. on ty + Hema be vee tha lai Mar ~ 012 Mia + 015M, ; fiat 013M 40-9. Cetstors, The cofactor ofthe element ay of determinant 1A | is (— 1)'* times the determinant obtai 04h chem of [4 | ad. deed Gp” ME HOH ie, y= (— aay, an tn ay | Mo fatelen thea Cy, cofactor of — Mathematical Preliminaries min Cue Oe oe 1 may be verified that 14 |= an Cy + aC + a1 Cy $010, Rask ofa Matrix, The ant ofa mati A is if all he minors of order r+ 1 or more ofthe mtx A are zero but at Kast one trior of order is nt 220 §0-11. Adjoint of Matric Let Aiba square matrix of order Then the adjoint of matric isthe wanspos ofthe mari, obtained from A by replacing each clement aj of A by ts cofactor Cy and i denoted by adj-4. Cu Ca. Gt Ca Ca Ga ie, adj A= pe (Ghee Case ‘Singular and Nonsingular Matrices. A square matrix 501. i i its determinant does not vanish ‘Ais sid to be a non-singular mai Gnd singular mai ifs determinant vanish eva is nonsingular if |A | # 0 and singular if | |= $013, Inverse ofa Matric Let A bea square matric If there css square matric B whieh sss the relation B= I= BA then B scaled the inverse of Avan is denoted by A “The inverse ofthe mari Ais given by ae hiedi.a 4 rai . Computation ofthe iverse by parte! Com pore man prone for speed a ites - weld va) and Rare suse submatrices andQ, 0 (0a mart) where £ (uit mat are two matrices. a Ro? exists and is known, sven by then the inverse of the matrix Ais Scanned with CamScanner2. Wetors and Var Sees. Se Bat oy 1. estar. Matrices having a single ow or column are ref, tos vectors Amattc ofa sige rows called a7ow vector, anda matt ofa single cum sealed coum vector = ‘exo wil Be denoed by lowercase bold face type, Typ i Eexsinal column vector | 2] i ao writes 3 6.62 iy the meinen row vers wien 2 (682-0444): Asvitora = (36) eld an n-componet ecto, 1.29mi sido be the th componet of vetor a As (ena) ia poi in ree dimensional space, s the wan ear feat en ‘gaz Thee is 0 frome dtneonBetecn row andes 2 Ualt Vo, A ui vec deat by i vector whos it sis avecor component ny and aloe compen re zero, : or vectors wth componeats thee tem ui vectors 2700-2 OY Ore Dynses= Ora Nall veer, A vector wow al components we ov 4 ml cor Fro ver ands dead by Ota ee 9=00,.0. 4 Sum vector A vector 2 ron ee (hte ah of ote component is scaled is 3 ver defied by ~ be by 09~ by. Se Mathematica! Preliminaries 7 17. Scalar Multiplication of a vector. The product of a vector by a scalar kis defined by . Ha (bay kayo h) 8. Distance between two vectors. The distance between [40 n-component vectors.a and b, denoted by | a—b | is defined to be the postive number : 1” Jab] =v{@—b).@—-b}= 5a 9. Scalar Product of two vectors. If a= (@ydpd5) 2d b=(ybyb,) ate two vectors then the scalar product of @ and b denoted by ab is defined by b= a] |b lease where 0 isthe angle between the two vectors. 1 may be verified that ab=ayby + a3b2 056s ‘Two vectors a and b are said to be orthogonal it ab § 0-15, Euclidean space. An n-dimensional Euclidean space, denoted by” or R ie the collection of alln component vectors (points) iydyrnd,)- For these vectors, operations of addition, rukiplication ofa vector by a scalar are defined by the rules for matrix ‘operations (as above), Furthermore, associated with any two vectors in the collection is a non-negative sumber called the distance between two sectors (defined in 8 of § 0°14), "Now in further discussion each cor willbelongto some Euclidean '$0-16, Linear Dependence and Independence of vectors : “The set of vectors ay 8; 3y fom E” is said 10 be linearly dependent (LD) if there eit sears y= 3,2 yeh, witha leat one 4, #0, such that Siyehaythntthy=d were a eoreoig E Bie oy os ays aati Liye © ve jyndye hr tat cea Wok be dey inp) Scanned with CamScannerRS Linear Progra mm incr Combinaine (LC) oF YEO. Let ayy, vecor 2 given by the expres.” a “" prestng Bia is sour. ay be the vectors from E. Thea th seh that te there dj #12, kare sears led he near combination (.¢) ofthe given set of vectors and belong (0 E". Important pins. A 1 The terms LD. and LI. are mutually exclusive ie, any se sector either LD. of Li : 2 Asset of vectors contig the O vector is LD. 2. Any inf set of vectors said to be LI. ts every suber ‘LA, otherwise itis LD. ‘ 0-18. Spanning set A st of vectors 8,83 84 from Ey $3 to span or generate E" if every vector in E" can be wtitien 354 linear combination of vectors 83 on 8 S019, Basiset. A spanning ct whose vectors are Lis aed basis set or a bas, Thus a set BC E form a basis for £” if @ Bis, andi) Bspan Ee, cach vector in E* of the vectors of 8, : Fer cramp thea B= (eye can be expressed as a LC Soa ono te vectors Ranges gut coratd LC ari eae epesettion ; wique, of any vector in terms of a basis set of tes ba Baa cot ang, ‘Scan replace A eprentg the a be! Property of the new set. [Se nc REN Mathematical Preliminaries ° 4. Any two basis for the same set have the same number of vectors. S. Any set of (n+ 1) oF more vectors of £* will be LD. 6. Any set of LiL. vectors of £" forms a basis of E". 3. Simultaneous Linear Equations. § 0-21. Simultaneous Linear Equations ‘simultaneous linear equations in n (2 m) unknowns x.) written as A gystem of m i, may be aya + ayaty tt ayaa = y.%, + ankg + tant = Be emit + gata where aj,bj are known constants, In matrix form the system can be written as follows A o where 2] aod b= | fare In bn at a Bn Mac Asc the cient mats on an ox on ‘The matiica b= [Ab] = Any set of vaso 40 which ty all the equations of the seem ele th son ofthe stem. pte of equation is TE PERE Eco if re cist ston of the bute there The cecsary and uct conn for» stem of equations to be content tat 74) = 48 Cae it rqa rd d=n, the sumber of uknowns. This ips tat= Scanned with CamScannero Liner Powanny When m=, then [| #0. S0 e obtain @ unig Solution : Mion Beg ty, ‘Sn 0 We can find a real number 7 (negative in ease (7) and positive in ease (ii) such that : rE earo or Sataro or Be (+74) oz = @ ‘Now multiplying (3) by and adding to (2), we bave e fx +74 2 i) ai Which implies that (mtn y i ry 4 rhs + Pett Fly 00D forall aes ois also solution ofthe system Ax = Ite choose 7, st ted 20 me a Denk reoH mae Peer rants }=0 fats and Thus +74; 2 0for in. (_# 0 meee ” (> 0) He, ‘Also we have Mx (_#) o ‘which implies if we chooser such Min [-#}>° (<0 interval «that interval given by (2) 2 #0N-eRPLY Scanned with CamScannera Linear Programming Hence whenslies in the non-empty interval the solution given by seat gies: cade (mtn—® bai rg Aygo bed Oe] is also aS. of the sistem Ax= b ana rom (8), this soltion gives vale 2" of Z which is rater than 2" (optimal value or greatest value). Which is a contad Zehi=OorZ'=Z int ie. x’ given by (6)is also an optimal solution. In this way from (derived) a new optimal soluti non-zero variables. This solution is optimal B.ES. if vectors associated to non-zero variables i eet een) fon ere ner of times we will certainly get an Iyutem, Hence the theorem itroe, Mate fa the ES the variable, wil Become er if a Max. [Ai 2 subject Y=) sisk |z|% 43-5, Toobtain BES. fom FS. iotem renee : Mae Zwenstde = O30, Actin x (mtn) matrix Of Coefficients given smn) maf nent sen by here A= Gay sEmen > st one bese oc [Meerut 95 (BP), 97 (BP), 98 (¢ Proof, Letx=leetnvim on BEA ible sola eun( wished) ex ane y aa ae oro. We xe seer st k components of gmt these sea LP BLE act compan censning tne eet Lr ae a ‘Simplex Method & 3 nae b. (0) Now there are two possbilt 1 it the ect op inSyevey are Lil If the vectors eyty theorem istrue non-degenerate iC =m Be Wthe vectors yay 6 are LD. This is the case when >. Inthis case we can reduc the sunber of nonzero variables step by step, s0 that the feasible solution become BFS. Proceeding asin § 3:4, can reduce the BS. Xm ftp oth Qos) (nta-¥) to.a new basic feasible solution, (m+n=8) more than (t~ 1) non-zero components. ing less FS. from the given FS. con number of non-zero (positive) variables. ‘Ifthe columa vectors associated to non-zero variable cen this sof is B.FS. and hence the theore ‘J vectors are no Ll then this solu the above process again and ot_more than his way for finite BES. of the new sol” are Lil, the But if these associate G2 non-ero va umber of mcs we Sysem, Hence the theorem ise Tone eh hoores Stand Sia combined form may sated follows nape TT vn max, Zo cx, sh Atm b 220 Me ota ed eat one ofthe BE sion il DOP! ea ee jon of the LRP feasible sol i Ex A fy = 2a 39> LOE Mar, Z="+22* st 2 tan tagell Scanned with CamScannera line Papen sen find BES. (Mert. L340, Sol. The given LPP. can be writtes a Man Penta es ‘igeaee wtowes= [Fore [evel Here m (umber of con a, ) = 2, 40 the BFS. ofthis non-rero variables The given feasible soletions, = 2; ay In order to reduce the F: 7 varlable vero, Fo sociated with the non-zero variables ate LD, therefore one of thete secon ‘expressed as a LC. of the remaining (wo. ‘ ae coy un ag ., "alist: Ay Ady Dand Ay + $25 = Solving, we have dg = ~2and 2, @y= 24; + 1a, or 0, +24,=0=0 or Brae ated = dead ea Now va, Mar () ay +42) 4,452) = @ isis3 lz, = Max, 1 = Max. (3.3 iables xy should be ze ‘simpler Method 6 Substiteting the value of from (2) in (3), we have 5 o 10.0, + Say 0orZ; ~ cj < Oholds, and least one y)> 0,i = 1,2, ..9n,then It is possible to obtain a new 'S. by replacing one of the columns in Bbyajand ifthe new value 1¢ objective function is Z' then Z' = Z. If the given B. FS. is generate then Z’ > Z. Proof, Let xg = bipy tp2,--»Fpqh be a B. F.S. of the given isisals x Ars bx20 where A = (yap oosegy) and the basis matrix B= By By Bm) Bay=d ~@ GEA, st. ye Since Py By Bn frm the basis of 4, therefore a ean be essed asthe LC ofthese vectors. 25 Ein Bay + Baty + + Pts -@) iy, 0, then we can replace in Bby cand sil Bis a basis trix, assuming that, # 0;from (1), we have hug Prana Scanned with CamScannerFrom (1), have Bites * Patino + Bet t+ Bhim =b termed a8 simpler is relation (9) a B, hich ie to be deleted san tor selecting. the vee rived) from the basis matrix B, aie important {0 note that inthe new B.FS. the variable ih position is different from that in the previous B.FS, whi "fier postions there are the same variables, usually with or Bi Bint hy =e Men fee - Subsitting the value of, rom 3) in (3), we have z AforBie]ittace.o. | [apni eit et tin BA eS em outa a6 Wig vo ace oe Comparing with (4, the new base sltion 48 = Bia ge ta teil vgn BFS degenerate BES. by i es cory prove that Z'=Z We have ‘a a ae ¥ ol a 205 coisn and 2'= 3 cnn whee’ x'g)=x9;~ Heyy, = 1,2,...m 02 r f Siemaahe «Ch where, c’pi, are the scoefficients of the basic varaibles and” xy 2 (= 12s) im the objective function Z. chee Ts obvi at he bi to’ be ute Paes aI a al 8 : , a Bg daz cmins gh ed emg a ih Now using (6), we have wt Bee Sisson ret Sine tg WBE Sot be LR R's! yo Rea Fam 3 ‘Thus we see that (7) and (8) are satisfied only if y,, > 0 and] = Fenix — Bee | +6 x , 980 C= EDs mie ey. hE a attend aie My > 0 and yy >, then (7) and (8) are satisfied only if 3 wm ed yi ag) 4g 2 Beem Be] eo “ sis eo}, Thus we cn Otain dnt BF Spam ip removing the cur vector Bo the bstree iil B. FS. by Te isto Be telced such tha By oy tfyy>0 Scanned with CamScanneruN we 0-2 whee inet. 2jconespond to he vol a x zezerg a) d Thu te new val ofthe jective fection i gui evel a hsbc asa ee Now Bez itg—z)20 Since veo ZezZ onlyite-Z20 or Z-9<0. Thus by choosing the vectra; for whch 6, ~Z,> Oand een ‘one y)> 0, obtain new ba solution with 2 2, 1 the original BLES. was non-degenerate then v>0 and this cate we te Fez. Hone the theorem is proned Note tv 0 thn ithe vga! BES deen eee ic tom ve hee 9 =ypfer ad ry = 8 The A ie th ahes of he vas conmigo Solutions (ld and new) do not change Ss bi 3 20 level Gtr pasion) Hee degenerate Bis, POH a 2110p <0 free hich 820 eh gil 25 hen «vii this ease and the new BLES, ee be 0 Te hc foe eed erring tes vet ote areas © le he {ato the basis is called the entering (o eonagd, itroduced emest vy (in mart 4) i ealed the hey ne © and the Sinton taco nee ihe bass 8 and remove soa. tsumed fom a newb then on ne sr ° shall isis the case it we inset ano for wich" Now we Forth we prove the flowing cores EVE, F =e sinpete Method and the new variable n Theorem. If for any basic feasible solution of @ LPP, there one totum 3, 6 bat et B for thick GB gs ay 2, sm), then the problem has an *unbounded solution Mteerut 94, 96, 97 (BP), 971 Proof Let xy [saison -ntagh be a BLES. of the given oe Zax st. Akad x20 25 Bxy=e or Zam tn ~Q) Let us insert the column aj in B, where ais in A but not in dg, ~ 2) > Oand yj $07 1,2... scalar then adding and subtracting 4a in (1), we Bent —teyttayen my Since ajEA st. aj €B. © Biv Pay --sByy form the basis of A, therefore a; ean be ‘tPressed as the LC. of these vectors _ aE hay o whey AE By Substituting in (3), we have Spats 7a Baty aa, aiB.—ax) 6, +30; = . When A>Othen rare Sa ae “rn ae oF Cece aieny i=in v0 nee Spina oe Scanned with CamScanner