NeuroNomixer
  • Home
  • Blog
  • Visual Guides
  • Authors
  • Contact
Sign InSign Up
HomeBlogAuthorsContactPrivacy Policy

© 2026 NeuroNomixer — Built with Next.js & Tailwind CSS

Visual Guides/Optimizers Race: SGD, Momentum, RMSProp & Adam
Deep Learning

Optimizers Race: SGD, Momentum, RMSProp & Adam

Watch SGD, Momentum, RMSProp, and Adam descend the same loss landscape simultaneously. See why adaptive optimizers converge faster on non-uniform loss surfaces, and tune learning rates to explore their trade-offs.

Exploration Progress0%
○ Watch race to step 30○ Adjust a learning rate○ Toggle an optimizer

Sign in to save progress

The Race

startmin-4-2024-3-1.501.53xy
Step: 0 / 40
Speed:
Practitioner Insight

Adam is the default choice for most deep learning today, but it can generalize slightly worse than SGD+Momentum on some tasks. Many practitioners train with Adam then fine-tune with SGD for better final performance.

Tune Learning Rates

Change a learning rate and the trajectories recompute instantly. Watch how each optimizer responds differently.

SGD
0.100
0.001|0.5
SGD + Momentum
0.100
0.001|0.5
RMSProp
0.050
0.001|0.5
Adam
0.050
0.001|0.5

Loss Surface

The landscape is L(x,y) = 0.1x² + 2y², an elongated bowl. At the default learning rate (0.1), SGD settles the steep y-direction fast (y shrinks 40% per step) but crawls along the shallow x-axis (x shrinks only 2% per step): that slow x-progress is what loses the race. Push SGD's learning rate above 0.25 and the steep y-axis starts to oscillate; above 0.5 it diverges. Adaptive optimizers rescale each axis, so they make even progress in both directions.

Gold star = global minimum (0, 0)
White dot = starting point (−3, 2)
ellipsesContour lines of equal loss

Current Loss

SGD
8.9000
SGD + Momentum
8.9000
RMSProp
8.9000
Adam
8.9000
← All GuidesNext Guide →