KATÁBASIS · the descent · how systems find their best by going downhill · kept by VIRGIL (the guide of the descent)

THE LEARNING RATE ◧ 2D · ◍ 3D · ◆ 4D · ◐ shadow · 👶 TAP

One knob rules training: the learning rate, the size of each downhill step. Too small and learning crawls, wasting a thousand steps to inch to the bottom. Too big and each step OVERSHOOTS, bouncing higher and higher until it diverges to infinity. There’s a sharp stability edge between them. Getting it right is most of the art of training. Slide across the edge.

◆ LIT▲ AMBER
◧ THE MEASURE · 2D
◍ THE STEP SIZE · 3D · too big, too small, just right
◆ THE FOURTH · 4D · a tesseract turns
◐ THE SHADOW · one dimension down
👶 THE TODDLER CORNER — one fat tap
learning rate
stability bound
1.00
after 20 steps
behaviour

◆ LIT — exact / checkable

On a quadratic loss L=½a·x² (curvature a), gradient descent updates x ← x(1 − ηa). It converges only if |1 − ηa| < 1, i.e. 0 < η < 2/a — below that it converges (fastest near η=1/a), and ABOVE the bound each step overshoots and the magnitude GROWS geometrically to infinity. For a=2 (loss x²) the bound is η=1. This sharp edge is why learning-rate tuning, schedules, and warmup dominate practical training. A fail-loud self-check throws unless η=0.5 converges and η=1.1 diverges. ◆ real optimization, node-verified.

▲ AMBER — the figure

The exact bound 2/a is for a quadratic; real non-convex losses have varying local curvature, so practitioners use schedules, adaptive rates (Adam), and warmup rather than one fixed η. The stability threshold shown is exact.

KATÁBASIS: to find the lowest point, follow the slope down — but mind the local valleys.  — VIRGIL
David Lee Wise / ROOT0, with AVAN · KATÁBASIS — kept by VIRGIL, the guide of the descent