Learn the best cut-off

This is the training step of the lesson's first program, for any data: find the cut-off that makes the fewest mistakes.

What to build

A program that tries every sensible cut-off and reports the best rule and its score.

Requirements

  • The rule predicts 1 when value ≥ cut-off.
  • Try each different value as the cut-off, in increasing order, and then the largest value plus 1.
  • If several cut-offs are equally good, choose the smallest.
  • Print pass if value >= C, then R of N correct.

Input

Line 1: the values (integers). Line 2: their labels, 0 or 1.

Output

Two lines.

Constraints

1 to 1000 values between −10⁶ and 10⁶.

Example

Input:

1 2 3 4 5 6 7 8
0 0 0 1 0 1 1 1

Output:

pass if value >= 4
7 of 8 correct

Why it works

Only the data's own values can change which examples a rule calls positive, so trying each of them (and one rule that rejects everything) covers every different rule.

Examples

Example 1 · Input
1 2 3 4 5 6 7 8
0 0 0 1 0 1 1 1
Expected output
pass if value >= 4
7 of 8 correct
Example 2 · Input
3 1 2
1 0 1
Expected output
pass if value >= 2
3 of 3 correct

Prompts inside input() are ignored when checking, so input("n: ") is fine.

Hints

Hint 1

The candidates are the different values, smallest first, and then the largest value plus 1 (a rule that never says 1).

Hint 2

Count the right predictions for every candidate and keep a candidate only if it is strictly better than the best so far.

Solution

Show one correct solution (try it yourself first)
values = list(map(int, input().split()))
labels = list(map(int, input().split()))
best_cutoff, best_correct = None, -1
for cutoff in sorted(set(values)) + [max(values) + 1]:
    correct = sum(1 for v, y in zip(values, labels) if (v >= cutoff) == (y == 1))
    if correct > best_correct:
        best_cutoff, best_correct = cutoff, correct
print(f"pass if value >= {best_cutoff}")
print(f"{best_correct} of {len(values)} correct")

Lesson: What machine learning is →