Neural Network Learning Lab – Gradient Descent visuell erklärt

Data

Select dataset Noise 0.05

Model

Hidden neurons Hidden layers Learning rate 0.080

Speed

Training speed 4

Didactics

Visualization

Main plot

Show decision boundary Show weights Show gradients Animate backpropagation Highlight misclassified points Show teaching hints

2D classification and decision boundary

Probability heatmap for class 1

Neural network

Weights, gradients, and signal flow

Das langfristig gelernte Wissen eines neuronalen Netzes steckt nicht in einem einzelnen Neuron, sondern verteilt in Gewichten und Biases. Dickere Verbindungen zeigen stärkere gelernte Kopplungen.

Compression Short fine-tuning

Loss chart

Binary cross entropy: training and test

Training status

Live metrics

Steps0

Test Loss0.0000

Train Acc0.0%

Test Acc0.0%

Gap0.0 pp

More metrics

Train Loss0.0000

Learning rate0.080

Train errors0

Test errors0

Hidden layers1

Neurons per layer6

ArchArchitecture: 2 → 6 → 1

Params0

Math & Formulas

ŷŷ = f(x; W, b)

LL = -[y log(ŷ) + (1-y) log(1-ŷ)]

WW ← W - η ∂J/∂W

bb ← b - η ∂J/∂b

Test preview

Unseen data – not used for training

Confusion Matrix – Test Data

Predicted 0Predicted 1 Actual 000 Actual 100

Precision: 0.0% · Recall: 0.0% · F1: 0.0%

Bias-Variance Indicator

Observe

Trade-off not yet clear. Continue training or adjust model complexity.

What is happening?

XOR

Ready

Start training or the guided explanation. The network begins with random weights and learns a decision function.

The gradient points in the direction of the steepest loss increase. Gradient descent moves in the opposite direction.