# Finding an incorrect.

• There's a test. 20 questions. N people passes the test one time and receives a 0-100% report. Let's say 50 people. The challenge is to find the algorithm as far as possible to identify the wrong answers in the test.

I mean, people's questions are shaft, and they can get all right, and they can make mistakes. So far, apart from the deletion of the questions from those groups that have given up 100 per cent of their way to find others. An example is that one person has 80 per cent of the results means about two questions wrong, how to find them if we know what questions people have given and what their final score.

upd: 50 people pass the test, 20 different questions each, 100 questions from the general bank of questions, i.e., one of 50 people will cross the test, 50 people get a percentage of the correct answers.0-100%. What question we know and how he answered, and the timelines for these people and their answers are available. 20, 21 questions are answered.

• Targeted linear programming. The equations look like a linear equation system, only variables. `Aij` limited to 0 and 1.

``````A00  +       A02 = Sum0
A10  + A11       = Sum1
0 <= Aij <= 1
``````

If the solution (some ILP Solver-) exists and is the only one that gets. `Aij=1` for the right answer of the laser `i` Question `j`0 for the wrong.

The existence of the only solution in question is questionable.

2

2

2

2

2

2

2

2

2

2

2

2

2

2

2