⚡ Dropout Visualizer

Watch neurons spark, then randomly die — and see dropout vanish completely in eval mode.

📋 How To Use This Lab
  1. Set Dropout Probability (p) — that's each neuron's chance of dying on any given pass.
  2. Press the big ▶ Run Forward Pass button every hidden neuron sparks, then survives or dies.
  3. Watch survivors get scaled up (divided by 1−p) to keep the layer's total output balanced.
  4. Flip to Eval Mode press it again — nothing dies anymore, no matter what p is set to.
p = 0.5
TRAIN MODE
Input Layer Hidden Layer — Dropout Applied Here Output Layer
Train Mode
Eval Mode
⚡ SPARKYPress ▶ Run Forward Pass — I'll spark every hidden neuron, then randomly decide who survives.
🧑‍🏫 PROF. TORCHDropout is a TRAINING-only trick. Call model.eval() and every neuron survives, every single time.

🎛️ Dropout Probability