You are on page 1of 3

A Non-Convex Combination of Gaussian Distributions

LINDO Systems has recently introduced LINDO API !" and LINGO #!"! $he ne% releases offer a number of enhancements includin& si&nificantly ex'anded nonlinear ca'abilities and im'roved 'erformance on linear and inte&er 'roblems!

Find Better Solutions to Tough Nonlinear Problems


(any nonlinear models are nonconvex )e!&!* they have more than one local o'timum+ and* as a result* not %ell suited to traditional solution techni,ues that rely on local search 'rocedures! Local search solvers are &enerally desi&ned to search only until they have identified a local o'timum! If the model is non-convex* other local o'tima may exist that yield si&nificantly better solutions! $he ne% Global solver and (ultistart feature included in the Nonlinear o'tion can hel' find better solutions to non-convex models!

Determine the Proven Global Optimum


-ather than sto''in& after the first local o'timum is found* the Global solver %ill search until the &lobal o'timum is confirmed! $he Global solver converts the ori&inal non-convex* nonlinear 'roblem into several convex* linear sub'roblems! $hen* it uses the branch-and-bound techni,ue to exhaustively search over these sub'roblems for the &lobal solution!

Use Multistart to Improve Solutions


.hen limited time ma/es searchin& for the &lobal o'timum 'rohibitive* the ne% (ultistart feature can be a 'o%erful tool for findin& &ood solutions more ,uic/ly! $his feature intelli&ently &enerates a set of candidate startin& 'oints in the solution s'ace! $hen* the nonlinear solver intelli&ently selects a subset of these to initiali0e a series of local o'timi0ations! 1or non-convex nonlinear models* the ,uality of the solution returned by the multistart solver %ill be su'erior to that of the &eneral nonlinear solver!

Solve Linear and Integer Models Faster


$he ne% releases of LINDO API and LINGO include enhancements that can si&nificantly im'rove 'erformance on many linear and inte&er models!

Find Faster Solutions Using the Dual Solver


$he im'roved Dual Sim'lex solver in the base versions of LINDO API !" and LINGO #!" delivers substantially better 'erformance! On broad classes of 'roblems* the ne% solver can 'rovide

s'eed im'rovements of u' to 2""3! $he Dual Sim'lex solver is also even more robust due to im'roved handlin& of de&enerate and numerically unstable 'roblems!

Lineari e Models !utomati"all#


Nonlinear solvers that utili0e &radient based al&orithms do not 'erform %ell on models %ith nonsmooth functions! $he ne% Lineari0ation ca'abilities in LINDO API and LINGO can dramatically im'rove 'erformance on models %ith common nonsmooth functions! $he feature automatically converts the nonsmooth functions and o'erators to a series of linear* mathematically e,uivalent ex'ressions! (any nonsmooth models may be entirely lineari0ed! $his allo%s the solver to ,uic/ly find a &lobal solution to %hat %ould have other%ise been an intractable 'roblem! LINDO API can lineari0e the functions absolute value* if* and* or* not* max* and min as %ell as the o'erators 4* 45* 46* 5* 6* and 65! It can also reco&ni0e and lineari0e the 'roduct of a binary )"78+ variable and a variable* or any 'roduct of t%o continuous or discrete variables constrained to be e,ual to* less than* or &reater than 0ero! LINGO functions su''orted by lineari0ation include 9A:S)+* 9(A;)+* 9(IN)+* 9S(A;)+* and 9S(IN)+ alon& %ith any 'roducts of binary and continuous variables! $he feature also lineari0es all of LINGO<s lo&ical o'erators )=L>=* =>?=* =G>=* and =N>=+!

Get $ui"%er Solutions to Integer Problems


$he ne% inte&er solver benefits from a number of enhancements that boost 'erformance on many classes of 'roblems! A 'artial list of ne% features includes@ (ore advanced 'robin&7're-solvin& includin& liftin& cli,ue S'ecial 're-solvin& of ro%s %ith all binary variables Additional and faster cut &eneration Im'roved roundin& heuristic Ne% enumeration solver for 'ure binary models (any ne% user controllable 'arameters Im'roved reduced cost fixin& and bound ti&htenin& %ithin the tree Im'roved 'erformance on mixed inte&er ,uadratic models

!utomati"all# &e"ogni e $uadrati" Models


?uadratic Pro&rammin& )?P+ models are a common class of nonlinear model that is encountered in financial 'ortfolio analysis! An enhancement to LINDO API and LINGO can automatically detect if a nonlinear model has a ,uadratic obAective and7or is ,uadratically constrained! If it is determined that the model is ,uadratic and the license includes the :arrier solver o'tion* the solver %ill automatically ex'loit the ?P structure to solve the model in a fraction of the time re,uired by the &eneral nonlinear solver!

Solve De"omposable Models Faster


Some lar&e-scale linear and mixed inte&er 'roblems have constraint matrices that are totally decom'osable into a series of bloc/ structures! Bsin& a ne% feature* the solver can detect and decom'ose the model* solve the inde'endent 'roblems se,uentially* and re'ort a solution for the ori&inal model! $his results in dramatic s'eed im'rovements!

Additional Enhancements LINGO 8.0:


-Solve Quadratic and Quadratically onstrained !odels -- LINGO licenses that

LIN"O A#I $.0


- reate A%%lications to Solve General Nonlinear !odels-- LINDO API is the first full-

include the :arrier o'tion can no% solve models in %hich the obAective function and7or some constraints include ,uadratic terms! :y ta/in& advanta&e of the ,uadratic structure* LINGO can solve these models much more ,uic/ly than usin& the &eneral nonlinear solver! LINGO can even handle ,uadratic models %ith binary and &eneral inte&er restrictions! $hese ne% ,uadratic ca'abilities ma/e LINGO suitable for a''lications such as 'ortfolio o'timi0ation 'roblems* constrained re&ression 'roblems* and certain classes of difficult lo&istics 'roblems )layout 'roblems* fixed-char&e-net%or/ 'roblems %ith ,uadratic obAectives+! -Analy&e In'easi(le and )n(ounded !odels-- Determinin& %hy your model is infeasible or unbounded no lon&er needs to be a dauntin& tas/! LINGO #!" includes a ne% set of tools that allo% you to 'in'oint %hat is causin& a model to be infeasible or unbounded! $he tools isolate a 'ortion of the ori&inal model as the source of the 'roblem! $his allo%s you to focus your attention on a relatively small subsection of the model to loo/ for formulation or data entry errors! -*uild !ulti-client and +e( A%%lications -LINGO #!" has been re-en&ineered to be threadsafe for linear* inte&er* and ,uadratic models allo%in& one instance of the 'ro&ram to simultaneously %or/ on solvin& multi'le models! $his ma/es LINGO #!" ideal for multiclient and internet a''lications!

featured callable solver to offer &eneral nonlinear and nonlinear7inte&er ca'abilities! $his uni,ue feature allo%s develo'ers to incor'orate a nonlinear solver into their custom a''lications! As %ith its linear and inte&er ca'abilities* LINDO API 'rovides the user %ith a com'rehensive set of routines for formulatin&* solvin&* and modifyin& nonlinear models! $he Nonlinear o'tion is re,uired in order to utili0e the nonlinear ca'abilities %ith LINDO API! -,un LIN"O A#I 'rom a +e( *ro-ser -LINDO API !" includes Cava Native Interface )CNI+ su''ort for .indo%s* Solaris* and Linux 'latforms! $his ne% feature allo%s users to call LINDO API in a''lets runnin& from a bro%ser! -E.%anded !A/LA* Inter'ace -- $he (atlab interface has been ex'anded to su''ort all LINDO API functions! Bsin& (A$LA:<s modelin& and 'ro&rammin& environment* you can build and solve linear* nonlinear* ,uadratic* and inte&er models and create custom al&orithms based u'on LINDO API<s routines and solvers! -Generate the "ual In'ormation -- 1or some 'roblem ty'es* it can be advanta&eous to %or/ %ith the dual formulation! LINDO API no% has the ability to ta/e a 'rimal LP and automatically &enerate the dual 'roblem!

For more in'ormation on LINGO or LINDO !PI( please see the produ"t pages) *ou "an also do+nload a demo version 'rom our do+nload page or order a 'ull blo+n version dire"tl# 'rom our order page)

You might also like