Experiment 2 · about 12 min

Tune the neuron

0/3 tasks

In this garden, tomatoes ripen as they turn red. Softness doesn’t matter at all.

Actually Ripe
Actually Unripe
The neuron says "Ripe"
The neuron says "Unripe"
× Got wrong

Accuracy

50%

7/14 fruits correct

Importance of color-1.0Importance of softness0.5Threshold0.0

Inside the neuron

score = color × -1.0 + softness × 0.5

if score ≥ 0.0 → Ripe

Tap a fruit on the map to see what the neuron calculates for it.

Training log

No training yet. Press the "Let it learn" button.

Tasks

In garden 1, reach 100% accuracy by changing the settings yourself (waiting)
In garden 2, watch the neuron learn by itself all the way (waiting)
In garden 3, discover why one neuron isn’t enough (waiting)

What’s going on?

An artificial neuron calculates a score for each fruit: it multiplies the color by one weight and the softness by another, then adds them up. If the score passes the threshold, it says "Ripe".

Learning means changing these weights. The neuron finds a fruit it gets wrong and nudges the weights a little toward that fruit. It repeats this until there are no mistakes left.

The line on the map splits the neuron’s decision. One neuron can only draw one straight line. For trickier jobs, neurons are combined in layers: this is called deep learning.

Related lessons

Lab