CIFAR-10 residual-network test error improves monotonically from 8.75% at 20 layers to a best-run 6.43% at 110 layers while parameter count rises from 0.27M to 1.7M; the much larger 1202-layer model reverses this trend at 7.93%. Table 6 also reports Maxout, NIN, DSN, FitNet, and Highway comparison results.

CIFAR-10 residual-network test error improves monotonically from 8.75% at 20 layers to a best-run 6.43% at 110 layers while parameter count rises from 0.27M to 1.7M; the much larger 1202-layer model reverses this trend at 7.93%. Table 6 also reports Maxout, NIN, DSN, FitNet, and Highway comparison results. · CiteArk

测量结果

已评估 0/25 项

获得支持 0有争议或冲突 0尚未评估 25

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
论文报告	最近可用实测	证据判断
9.38%	—	尚未评估
8.81%	—	尚未评估
8.22%	—	尚未评估
2.5 million parameters	—	尚未评估
8.39%	—	尚未评估
2.3 million parameters	—	尚未评估
7.54%	—	尚未评估
7.72%	—	尚未评估
0.16 percentage points	—	尚未评估
1.25 million parameters	—	尚未评估
8.8%	—	尚未评估
0.27 million parameters	—	尚未评估

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果