diff options
Diffstat (limited to 'ROADMAP.md')
| -rw-r--r-- | ROADMAP.md | 16 |
1 files changed, 9 insertions, 7 deletions
@@ -335,13 +335,15 @@ claim is explicitly computational rather than a fit to cortical data. ## Oral bar A: standard deep architectures -**Status: mechanics and gates frozen; no image/ResNet accuracy observed yet.** -`ORAL_A.md` fixes a validation-only ResNet-20 development funnel followed, only -if it passes, by an untouched five-seed ResNet-20/32/56 test panel. The standard -BatchNorm local Jacobian and BN-coupled batch-objective perturbation estimator -both match exact numerical derivatives below `1e-6`. The alternative -translation-shared state-gated vectorizer reduces ResNet-56 feedback parameters -131x and is included prospectively alongside the full spatial template. +**Status: A3 failed; A4 was not opened and all confirmation test seeds remain untouched.** +`ORAL_A.md` froze a validation-only ResNet-20 development funnel. A1 passed with +`91.62%` BP validation accuracy. A2b selected channel-gated SDIL at `41.98%`, +ahead of tuned DFA at `37.16%`, on the 10k/20-epoch screen. In full A3, DFA +ended finite at `33.06%`; SDIL became nonfinite at epoch 90 and ended at +`10.00%`, failing alignment, accuracy, and finiteness gates. The MAC gate alone +passed. Per the stop rule, the five-seed ResNet-20/32/56 test panel was never +run. The exact BatchNorm/local-gradient mechanics remain verified, but the +current training recipe does not establish standard-network scaling. Prepare convolutional local-update primitives and ResNet-20/32/56 protocols early. Queue frozen runs opportunistically on authorized idle GPUs. Because BurstCCN already reports CIFAR-10 and |