Dataset
Linear
linearly separable · easy
XOR
non-linear · perceptron fails
Circle
inner vs outer ring
Spirals
two interleaved arms
Moons
crescent shapes
Blobs
3-class clusters
Hidden layers
1
2
3
Neurons / layer
4
8
16
32
Activation
ReLU
Tanh
Sigmoid
Learning rate
0.01
0.1
0.5
1.0
Batch size
16
32
All
Steps / frame
×1
×5
×20
⏸ Pause
↺ Reset
New data
dataset =
Linear
step =
0
loss =
—
acc =
—
arch =
2→8→8→1
act =
ReLU