Team:MoWestern Davidson/project mathmodel

Square Diagrams

We used Wolfram Mathematica to visualize 2-SAT problems and particularly to analyze unsatisfiable problems with varying numbers of clauses.

The clauses of 2-SAT problems can be graphed as squares on a grid, like the one below, forming a "B2 bomber" shape. The following grid shows all of the 4-variable 2-SAT problems. For example, the square completely to the bottom left corresponds to the clause (a' OR b).

This problem is unsatisfiable because choosing both literals of one of the variables present is required to solve the problem. For example, the input (a, b, c, d) satisfies the first three clauses but fails to satisfy the fourth because both a and a' or both b and b' cannot be in an input together.

Block Diagrams

We used Mathematica to construct block diagrams in order to visualize 3-SAT, as well as to graphically represent the clauses not satisfied by a particular input.

Just as the 2-SAT clauses can be graphed as squares, the 3-SAT clauses can be graphed as blocks. The following right-handed coordinate system displays all of the 3-variable 3-SAT problems.

The x- and y-axes are labeled just as they were in the square diagrams, with one difference. If there are n variables in a given 3-SAT scenario to be visualized, there will be n - 1 variables on the x- and y- axes. So, for this scenario, which has three variables, the x-axis is scaled a', a, b', b, and the y-axis is scaled a, a', b, b'. The z-axis always has the remaining variables, starting from the negative of the last variable at the bottom and ending with c at the top. Therefore, the z-axis in this scenario is scaled c', c. As another example, if there were four variables in a 3-SAT scenario, then the x-axis would be scaled a', a, b', b, c', c, the y-axis would be scaled a, a', b, b', c, c', and the z-axis would be scaled d', d, c', c.

We developed an algorithm that highlights all of the clauses not satisfied by a particular input. For example, consider the input (a, b, c, d).

Rough and Fine Distributions

To classify SAT problems, we began looking at the rough distribution of each problem, meaning how many inputs satisfied a certain number of clauses in a problem. There are ₃C₂ ×2² = 12 clauses for 3-variable 2-SAT. There are ₁₂C₃ = 220 problems with 3 clauses. The table below gives the number of clauses satisfied by each input for a few select 2-SAT problems.

The rough distribution for the red problem is 0143. That distribution means 0 inputs satisfied 0 clauses, 1 input satisfied 1 clause, 4 inputs satisfied 2 clauses and 3 inputs satisfied all 3 clauses.

The following pie chart shows 2-SAT 3-clause problems grouped by rough distribution, with all shades of a given color representing one rough distribution.

We also outlined rough distributions on Mathematica.

Fine distributions look more deeply at how each clause was satisfied. If an input satisfies one or two literals in a clause, that clause is satisfied singly or doubly respectively. For example, the input abc’ satisfies the clause (a OR b’) singly because it satisfies only one literal in the clause. However, the same input abc’ satisfies the clause (b OR c’) doubly because it satisfies two literals in the clause.

The fine distribution for the red problem from the table above is 0011212100. This distribution means that 0 inputs satisfied 0 clauses, 0 inputs satisfied 1 clause singly, 1 input satisfied 1 clause doubly, 1 input satisfied 2 clauses with 2 singles, 2 inputs satisfied 2 clauses with 1 single and 1 double, 1 input satisfied 2 clauses with 2 doubles, 2 inputs satisfied 3 clauses with 3 singles, 1 input satisfied 3 clauses with 2 singles and 1 double, 0 inputs satisfied 3 clauses with 1 single and 2 doubles, and 0 inputs satisfied 3 clauses with 3 doubles.

The following pie chart shows 2-SAT, 3-clause problems grouped by fine distribution, with each shade of a given color representing one fine distribution.