Concept library/Architecture
04
6 minute interactive lab

The forward pass

How does an input become an output?

YOU’LL LEAVE KNOWINGInference is a chain of deterministic transformations.
THE EXPERIMENT

Trace every number from input to prediction.

Move the controls. Watch the internal state respond. The explanation follows the behavior—not the other way around.

Runs the real small-scale equationThe displayed values are calculated from the formula shown in this lesson.
INFERENCE TRACEFollow the arithmetic

The parameters stay fixed. Changing an input recomputes every downstream activation in order.

inputsa = 0.70b = 0.30
weighted sumsz₁ = 0.540z₂ = 0.080
ReLUh₁ = 0.540h₂ = 0.080
sigmoid output66.6%z = 0.688
THE MENTAL MODEL

Inference is a chain of deterministic transformations.

A forward pass applies the network’s current weights, biases, and activations in order. Nothing learns during this step: the same input and parameters always produce the same output.

IN DEVELOPER TERMSprediction = layer₂(activation(layer₁(input)))
GO DEEPER

Check the model against the source.

Connectionism optimizes for intuition, then points you to the rigorous treatment.

  1. Dive into Deep LearningForward propagation
  2. Deep Learning bookDeep Feedforward Networks
NEXT CONNECTION

Did the model learn—or memorize?

Overfitting and generalization

Continue to lab 05