Problemformulering är design

Reframing the Same Data

~20 min

Deep Insight: "Supervised" and "Unsupervised" are often just different ways of looking at the same data.

Interactive Experiment

Toggle between "With Labels" and "Without Labels" below. Notice that the *points* don't move. Only your *goal* changes.

🔍 Without Labels

Model sees patterns (gray circles show natural groupings) but has NO GOAL. It can cluster, but doesn't know what's "right" or "wrong".

Behavior: Groups similar things together, but can't predict labels.

Try all four modes to unlock the insight →

Two Views of the World

Unsupervised View

"What does this data look like?"

  • • Finds natural clusters
  • • No "right" or "wrong" answers
  • • Good for exploration

Supervised View

"How do I predict X from Y?"

  • • Draws boundaries
  • • Creates a prediction machine
  • • Requires humans to provide labels

💡 The Frame Matters More Than The Algorithm

The most important choice you make isn't "Random Forest vs Neural Net". It's "Do I have labels? What am I trying to predict?"