← Back

SPSA

technique

1 creator use or recommend it

▲ 1 active in the last 90 days

1 use 1 mention
Last discussed Oct 2, 2026

Simultaneous Perturbation Stochastic Approximation, an optimization algorithm

Recently discussed in

What creators say

Used for: Training a 1 billion parameter LSTM for next-token prediction without backpropagation Praised: Needs no gradients, can optimize 0-1 loss landscapes, and avoids local minima... vs: Performed best among every zero-order method in a textbook; finds the minimum...

"The learning procedure, the update rule that worked the best, is something called SPSA that pretty much no one knows about except James Spall and his students that invented this learning update rule... What's cool about this, the other property is..."

More from Francois Chaubard →
Y Combinator
mentioned
Used for: Alternative learning procedure to backpropagation vs: Alternative to backpropagation for neural network training

"I'm highly interested in SPSA, but if there's alternatives that I'm not aware of, please like recommend them"

More from Y Combinator →

About this data

  • ✓44,000+ mentions extracted from real podcast transcripts
  • ✓Every quote links to the source episode
  • ✓Sponsored mentions are always labeled
  • ✓Critical takes are shown, never filtered out
How Followed works →
Get an email when a new creator recommends it

Follow SPSA

Free email updates