Modeling is the art of formulating your application in terms of precisely described, well-understood problems. Proper modeling is the key to applying algorithmic design techniques to any real-world problem. Indeed, proper modeling can eliminate the need to design or even implement algorithms by relating your application to what has been done before. Proper modeling is the key to effectively using the problem catalog in Part II of this book.
Real-world applications involve real-world objects. You might be working on a system to route traffic in a network, to find the best way to schedule classrooms in a university, or to search for patterns in a corporate database. Most algorithms, however, are designed to work on rigorously defined abstract structures such as permutations, graphs, and sets. After all, if you can't define what you want to do, you can't hope to compute it. You must first describe your problem abstractly, in terms of fundamental structures and properties.
Whatever your application is, odds are very good that others before you have stumbled upon the same algorithmic problem, perhaps in substantially different contexts. Therefore, to find out what is known about your particular ``widget optimization problem,'' you can't hope to look in a book under widget. You have to formulate widget optimization in terms of computing properties of abstract structures such as:
These fundamental structures all have associated problems and properties, which are presented in the catalog of Part II. Familiarity with all of these problems is important, because they provide the language we use to model applications. To become fluent in this vocabulary, browse through the catalog and study the input and output pictures for each problem. Understanding all or most of these problems, even at a cartoon/definition level, will enable you to know where to look later when the problem arises in your application.
Examples of successful application modeling will be presented in the war stories spaced throughout this book. However, some words of caution are in order. The act of modeling reduces your application to one of a small number of existing problems and structures. Such a process is inherently constraining, and certain details might not fit easily into the given model. Also, certain problems can be modeled in several different ways, some much better than others.
Modeling is only the first step in designing an algorithm for a problem. Be alert for how the details of your applications differ from a candidate model. But don't be too quick to say that your problem is unique and special. Temporarily ignoring details that don't fit can free the mind to ask whether they were fundamental in the first place.