Experiment gallery

Start from demos that reveal
how models learn.

Each example is designed to expose a different learning behavior: non-linear boundaries, convergence speed, noisy errors, gradient flow, or custom CSV training.

Search examples

Open playground →

Classic non-linear split

Medium

XOR compact

Shows why a linear boundary is not enough. A tiny hidden network learns quadrant-based separation.

Arch2 → 4 → 4 → 1
ActivationReLU
Datasetxor
LevelMedium

Fast convergence demo

Easy

Circles clean

A strong demo for decision boundaries. The network learns an enclosed class region quickly.

Arch2 → 4 → 4 → 1
ActivationReLU
Datasetcircles
LevelEasy

Process-view stress test

Hard

Spiral deep

Great for inspecting confidence, misclassifications, gradients, and the limits of tiny networks.

Arch2 → 8 → 8 → 8 → 8 → 1
ActivationTanh
Datasetspiral
LevelHard

One boundary sanity check

Easy

Linear baseline

Proves the system works on linearly separable data before trying harder shapes.

Arch2 → 2 → 1
ActivationLinear
Datasetblobs
LevelEasy

Bring your own data

Variable

CSV import demo

Turns Axon from a playground into a real local experiment lab for 2D binary datasets.

ArchUser selected
ActivationAny
Datasetcustom CSV
LevelVariable