Constrained minimization

Constraints are relations of equality or inequality, which must be exactly obeyed by the unknowns of a model. A familiar example is the mineral abundances in a rock or the end-member proportions in a mixture, which must sum up to unity whatever the errors on the data. 

The derivatives with respect to x produce the set of equations represented by 

Adding the first two equations and subtracting the third twice results in 

Distribution of energy states. According to quantum theory, the energy states g0, i, 2. that atoms in a gas, a liquid or a crystal can reach are distinct and have an equal probability of being taken by an atom. Standard textbooks (e.g., Swalin, 1962) show that the entropy S of a population of N atoms, nf being in the energy state s , is 

The first constraint is a fixed total atoms number 

---

Techniques have been developed within the CASSCF method to characterize the critical points on the excited-state PES. Analytic first and second derivatives mean that minima and saddle points can be located using traditional energy optimization procedures. More importantly, intersections can also be located using constrained minimization [42,43]. Of particular interest for the mechanism of a reaction is the minimum energy path (MEP), defined as the line followed by a classical particle with zero kinetic energy [44-46]. Such paths can be calculated using intrinsic reaction coordinate (IRC) techniques... 

The point where the constraint is satisfied, (x0,yo), may or may not belong to the data set (xj,yj) i=l,...,N. The above constrained minimization problem can be transformed into an unconstrained one by introducing the Lagrange multiplier, to and augmenting the least squares objective function to form the La-grangian,... 

As we mentioned earlier, this is not a typical constrained minimization problem although the development of the solution method is very similar to the material presented in Chapter 9. If we assume that an estimate k(J) is available at the j,h iteration, a better estimate, k(J+l), of the parameter vector is obtained as follows. 

A radically different approach to the steady-state problem was investigated by Hsing (H6). In this approach the steady-state flow problem was formulated as the following constrained minimization problem ... 

To show that this constrained minimization is indeed equivalent to the steady-state formulation, let us adjoin the equality constraints to the objective function to form the Lagrangian function,... 

The design of the network calls for the selection of pipe diameters such that the discharge through each valve attains the maximum (sonic) velocity for an initial transitory period. Since the flare pressure and the process unit pressures are specified, this requirement amounts to the stipulation of a maximum allowable pressure drop over each path Sj (labeled with a roman numeral) from the valve to the flare. The optimal design in this case may be formulated as the following constrained minimization problem ... 

The constrained minimization problem stated above may be transformed into a form well-suited to gradient projection methods of nonlinear programming by making the following substitution ... 

Solution. The reconciled results in Table E16.4 are obtained by solving the optimization problem with the process model as the only set of constraints. Because all constraints are linear, an analytical solution exists to the problem, as given in Equation 16.11. This results in an 89.6% reduction in the sum of the absolute error. Note that all reconciled values are positive and hence feasible. It is not unusual for some reconciled flow rates to go negative, in which case it is necessary to solve the problem using a constrained minimization code such as QP. 

The adjustment of measurements to compensate for random errors involves the resolution of a constrained minimization problem, usually one of constrained least squares. Balance equations are included in the constraints these may be linear but are generally nonlinear. The objective function is usually quadratic with respect to the adjustment of measurements, and it has the covariance matrix of measurements errors as weights. Thus, this matrix is essential in the obtaining of reliable process knowledge. Some efforts have been made to estimate it from measurements (Almasy and Mah, 1984 Darouach et al., 1989 Keller et al., 1992 Chen et al., 1997). The difficulty in the estimation of this matrix is associated with the analysis of the serial and cross correlation of the data. 

Figure 3.13 Constrained minimization the minimum of a function f x) submitted to the constraint g(x)=0 occurs at M on the constraint subspace, here on the curve (x)=0 where Vf(x)+XVg(x)=0. P is the unconstrained minimum of/( ). This principle is the base for the method of Lagrange multipliers.

PVVaik) is therefore the direction of constrained minimization. As in the case of Lagrange multipliers, no progress can be made and search will stop when the (k + l)th minimization direction PV (k+1) is orthogonal to the fcth minimization direction PV inner product of these vectors becomes less than an arbitrarily small value. 

In 1979, an elegant proof of the existence was provided by Levy [10]. He demonstrated that the universal variational functional for the electron-electron repulsion energy of an A -representable trial 1-RDM can be obtained by searching all antisymmetric wavefunctions that yield a fixed D. It was shown that the functional does not require that a trial function for a variational calculation be associated with a ground state of some external potential. Thus the v-representability is not required, only Al-representability. As a result, the 1-RDM functional theories of preceding works were unified. A year later, Valone [19] extended Levy s pure-state constrained search to include all ensemble representable 1-RDMs. He demonstrated that no new constraints are needed in the occupation-number variation of the energy functional. Diverse con-strained-search density functionals by Lieb [20, 21] also afforded insight into this issue. He proved independently that the constrained minimizations exist. 

Constrained minimization of an objective function like the minimal function (Debaerdemaeker and Woolfson, 1983 Delitta et ah, 1994)... 

This problem can be translated into one of linear programming. Introducing the variables s > 0 we first construct an equivalent constrained minimization problem given by... 

In addition, we are interested in functions that are at least twice continuously differentiable. One can draw several such curves satisfying (4.27), and the "smoothest" of them is the one minimizing the integral (4.19). It can be shown that the solution of this constrained minimization problem is a natural cubic spline (ref. 12). We call it smoothing spline. 

Here n is the number of moles of the fcth component. The equilibrium composition is found as the set of nt values (k = 1,..., K) that minimizes the function G, with the constraints of the mass balance of the system. The problem of a constrained minimization can be solved by a number of methods [369] frequently the method of undetermined Lagrange multipliers is used [368]. 

Multiplier Estimates for Constrained Minimization", Math.Programming, 1979, 17(1), 32-60. 

Constrained Minimization. In cases where one has internal degrees of freedom, besides the six associated with position and orientation, the use of constrained minimization procedures becomes a useful technique. 

X. The needed elements of the matrix X are obtained directly from the Cholesky decomposition of X. The constrained minimization of S 9) is performed by successive quadratic programming, as in Chapter 6, but with these multiresponse identities. 

The strategy described in Section 6.4 for constrained minimization of the sum of squares S 9) is readily adapted to the multiresponse objective function S 9) of Eq. (7.2-16). The quadratic programming subroutine GRQP and the line search subroutine GRS2 are used, with the following changes ... 

GREGPLUS in selecting pivots in the normal-equation matrix A for each constrained minimization of S 6). GREGPLUS does not judge a parameter estimable unless its test divisor exceeds ADTOL at pivoting time. 

Once a reasonable candidate for the pharmacophore is available, then the associated distance constraints can be used to examine the remaining analogs active at the receptor under investigation for their ability to assume the pattern. If the question is simply one of the ability to present the pharmacophore, then a constrained minimization can be used. Since there is usually more than one conformer capable of the pharmacophore, however, and volume considerations may dictate which conformer is more likely, a systematic search utilizing the distance constraints from the pharmacophore is preferable. The entire set of active compounds can be screened for consistency with the proposed pharmacophore to test its credibility. 

---