How large a model can I solve with C-WHIZ?
Maximum model size is limited only by the size of your computer. The Standard version of C-WHIZ solves models with as many as 32,767 constraints; the Plus version has no model size limits. C-WHIZ exploits a common feature of LP models known as supersparsity to pack the largest possible matrix in the available RAM. Extremely large model might require Verison 4 C-WHIZ.
Back to the top
How do I control C-WHIZ execution?
C-WHIZ execution control is handled through command line arguments, or through the MPSIII parameter file called the communication region (CR), or a combination of both.
A user interface is provided in both the DOS and Windows versions that lets you set up the execution control with a set of menus and dialog boxes. Chapter 5 of the C-WHIZ manual provides a thorough discusion of C-WHIZ control.
Back to the top
When I run my model there seems to be a long delay between the matrix generator and the reports; how can I tell if everything is all right so that I don’t panic and stop the computer?
Predicting the amount of time it might take to optimize a given model is difficult; the time is a function of many things. Model size is an important factor and so is machine speed, but so are certain intangibles such as the size of the feasible space and how discriminating the coefficients are. If I have solved a model, changed some of the coefficients and resolved it, I would expect it to take approximately the same amount of time, but it might take a much longer time or a much shorter time. Since the optimization is an iterative procedure, we don't know what path it will take or how many iterations it might require.
During the optimization process, you may request the display of optimization log-lines on your screen. You can set the communication region variable XFREQLOG to n so that the log-lines appear every n simplex iterations. Two different messages are used. Before C-WHIZ has found the feasible space the message is:
nnnnn ITERATIONS XFUNCT= xxxxx.xxxxx XNIF= iiiii XSIF= yyyyy.yyyyy
Where:
nnnnn is the number of simplex iterations performed so far,
xxxx.xxxxx is the objective function value, iiiii is the number of those infeasibilities yet to be resolved, and
yyyyy.yyyyy is the sum of infeasibilities
After the feasible space is entered, the message changes to:
nnnnn ITERATIONS XFUNCT= xxxxx.xxxxx
You should be able to see these values progressing.
Back to the top
Can C-WHIZ solve mixed-integer problems?
Yes, The optional mixed-integer component (we call it MIPIII) uses branch and bound and supports a variety of discrete variable types: binary, general integer, semi-continuous, special ordered sets of type 1 and 2, and special unordered sets. MIPIII control is through the command line, the CR, and a set of control tables in a text file. An interactive user interface can be used to supply all of the control information.
Back to the top
Will C-WHIZ do forest management using a GIS?
C-WHIZ is a linear simplex optimizer. Given a system of linear equations, inequalities, and a mathematical expression of a function to minimize (maximize), it finds a solution to that system of equations and inequalities that has a minimum (maximum) value for the given function. C-WHIZ has no idea what these numbers represent or where they came from. Indeed, that is one of the things that makes it valuable. You may be looking at optimization from the point of view of forest planning, but others see it through the eyes of steel making, oil field operation, beer production, global energy supply planning, refinery operation, pizza production, paper production, personnel policy analysis, pipeline operation, power plant coal purchase, etc. ad infinitum. C-WHIZ is the single unchanging element in all of these applications.
Forest management modelers have had very good success with C-WHIZ. Forestry users of C-WHIZ include US Forest Service (with both Forplan and Spectrum models), New Zealand Forest Research Institute, Remsoft (Woodstock model), University of Northern Arizona, University of California at Berkeley, University of Northern British Columbia, and FORCE/Robak.
The key to using C-WHIZ (and any other optimizer, for that matter) is the sentence, above, that begins with "Given a system ...". Your job, as an analyst, is to give the optimizer that system of equations, etc., i.e., the model. Every application of C-WHIZ embodies some scheme for generating the relevant model. These schemes usually involve one or more computer programs that acquire data, edit data, present data to users for review and update, generate the linear programming matrix (the optimizers view of the model), execute the optimizer, analyse the solution, write the solution to report screens, etc.
The US Forest Service has a continuing project with the University of Montana in Missoula, MT to develop a GIS based forest planning system. The product exists and is called MAGIS. It uses C-WHIZ and MIPIII for optimization and our model management software, DATAFORM, for matrix generation and solution control. Other than MAGIS, C-WHIZ has no specific advantage for forest management model except that we have observed a significant speed advantage for most forest management models and for any given model, C-WHIZ requires considerably less RAM to hold the model.
Back to the top
