Skip to main content
← Back to Machine Learning & Deep Learning
Elementary Lab

ML Pattern Finder

Sort creatures into groups, then watch a model discover the hidden rule that tells them apart.

Pattern Finder

You are the teacher! Sort creatures into two groups, then press Find the Pattern — the machine will figure out which features predict each group. Finally, build your own creature and see what it predicts!

Sort the creatures into two groups

Group A (0)

Use the buttons below to add creatures here

Unsorted (16)

col:red
sz:bigsh:circle
col:blue
sz:smallsh:star
col:green
sz:bigsh:square
col:yellow
sz:smallsh:triangle
col:blue
sz:bigsh:circle
col:red
sz:smallsh:square
col:green
sz:smallsh:star
col:yellow
sz:bigsh:circle
col:blue
sz:smallsh:square
col:red
sz:bigsh:triangle
col:green
sz:bigsh:triangle
col:yellow
sz:smallsh:star
col:red
sz:smallsh:circle
col:blue
sz:bigsh:triangle
col:yellow
sz:bigsh:square
col:green
sz:smallsh:circle

Group B (0)

Use the buttons below to add creatures here

Find the Pattern

Sort at least 2 creatures into each group first! Group A needs 2 more. Group B needs 2 more.

Sort more creatures then press Find the Pattern.

How does it actually work?

When you press Find the Pattern, the machine counts how many times each feature value (like blue, or small, or star) appears in each group. It then computes a probability — for example, "60% of blue creatures were in Group A." This is called frequency-based learning, closely related to a classic algorithm called Naïve Bayes.

When it predicts your new creature's group, it multiplies these probabilities together for all three features and picks the group with the higher combined probability. The reasons it shows you are the feature values that were most unevenly split between the two groups — those are the strongest clues.

Real machine learning works the same way at its core: count patterns in labeled examples, then apply those patterns to new data. The more labeled examples you provide, the more reliable the patterns become.