Applies to:
Evolver 5.x and Up
RISKOptimizer in @RISK 5.x and Up, Industrial Edition

 

Making an Adjustable Cell Matrix of Ones and Zeroes (Rectangular Case)

This problem can be reformulated as a permutation / assignment problem to greatly reduce the number of adjustable cells and eliminate the need for explicit constraints.

Square matrix (review)

If you have an n × n matrix with exactly one 1 in each row and column, you only need one adjustable cell per row.
Each adjustable cell contains the column number where the 1 appears in that row.

• Instead of 200×200 binary cells, use 200 adjustable cells with values 1–200.

• Use the Order method.

• Every trial is automatically valid (one 1 per row and column).

• Optimization is much faster because no invalid trials are generated.

Interpretation:
If adjustable cell 55 = 187, then row 55 has a 1 in column 187 and zeroes elsewhere.

Rectangular matrix (200 rows × 60 columns)

You can still use one adjustable cell per row, but you cannot use the Order method because columns can have multiple 1s and you don’t know the final column frequencies in advance.

Instead, use the Grouping method:

1. Create 200 adjustable cells (one per row).

2. Each cell contains a number from 1 to 60 (the column index).

3. Initial values can be any mix of 1–60, as long as:

   • Every adjustable cell has a value (no blanks or zeroes).

   • Each group (1–60) occurs at least once if you want all columns used.

During optimization:

• The optimizer varies both the assignment and the frequency of column usage.

• Each trial is automatically valid: exactly one 1 per row, any number per column.

• No explicit constraints are required.

Important option:
If you check “All Groups Must Be Used”, then every column will have at least one 1 in every trial.
If unchecked, some columns may have zero assignments.

More complex constraints

If you have additional rules (for example, certain rows must be assigned to the same column), consider using the Schedule method, which can enforce more structured assignment logic.