The 1202-layer CIFAR-10 ResNet optimizes to below 0.1% training error but has 7.93% test error, worse than ResNet-110 despite similar training error. The authors attribute this generalization gap to overfitting by a 19.4M-parameter model on the small dataset and note that they did not use maxout or dropout.

The 1202-layer CIFAR-10 ResNet optimizes to below 0.1% training error but has 7.93% test error, worse than ResNet-110 despite similar training error. The authors attribute this generalization gap to overfitting by a 19.4M-parameter model on the small dataset and note that they did not use maxout or dropout. · CiteArk

Measurement results

0/5 assessed

0 supported0 disputed or conflicting5 not assessed

The complete result group stays with its claim. Assessed rows come first; their judgments do not replace the whole-claim conclusion. Expand a row for conditions, sources, and evidence history.

Each row is a measurement with its reported value, latest available observation, and cumulative evidence state.
Reported	Latest observation	Evidence judgment
0.1%	—	Not assessed
7.93%	—	Not assessed
19.4 million parameters	—	Not assessed
6.43%	—	Not assessed
6.61%	—	Not assessed

Reported and observed measurements

Assessments (0)

Runs (0)

Measurement results