Prototype hill-climbing learner #154

New Issue

2026-09-12T11:37:24+02:00

SirStone commented

2026-09-12 11:37:24 +02:00

Question

Add the first learning algorithm: hill-climbing / evolutionary weight optimization. Using the learning loop structure decided earlier, implement: random weight mutation, evaluation by aiming error, keep-best selection. The debug overlay should show error dropping over episodes. This is the "fast path" to proving the network topology can learn to aim.

Parent map: #147

Blocked by: #150, #152

## Question Add the first learning algorithm: hill-climbing / evolutionary weight optimization. Using the learning loop structure decided earlier, implement: random weight mutation, evaluation by aiming error, keep-best selection. The debug overlay should show error dropping over episodes. This is the "fast path" to proving the network topology can learn to aim. Parent map: #147 **Blocked by:** #150, #152

SirStone commented

2026-09-12 12:07:10 +02:00