Skip to main content
Alnur Ismail - Founder, Advisor, Investor

Normalization in ML Explained

• 1 min
probability normalization

What I worked on

Tried to understand why normalization requires integrating over all possible outcomes in probability. Looked at examples like dice PDFs and compared spiking neural networks and energy-based models.

What I noticed

  • Integration ensures total probability sums to 1
  • PDF for dice would be discrete and flat
  • Spiking networks and energy-based models have very different dynamics

”Aha” Moment

na

What still feels messy

Still unclear how integration maps directly to ML normalization steps.

Next step

Review how energy-based models use normalization constants (partition functions).