LLM AtlasObserver · 0 XP
Pricing
Neural Networks

Layer 2 · Mechanics

Neural Networks

The forward pass step by step, common activation functions, and how architecture choices shape what a network can learn.

8 min read70 XP
ReLU, sigmoid, and tanh: the three activation functions you'll see constantly.

  1. z = Wx + b

    linear combination

  2. a = f(z)

    nonlinear activation

  3. pass a to next layer

One layer's forward computation, repeated for every layer in the network.
Go deeper: L3 Code