Skip to content

CIFAR-10 CNN 과제별 실습과 실행 결과 추가 - #19

Draft
Alpenj wants to merge 2 commits into
mainfrom
agent/cifar10-interactive-workbook
Draft

Alpenj wants to merge 2 commits into
mainfrom
agent/cifar10-interactive-workbook

Conversation

@Alpenj

@Alpenj Alpenj commented Aug 18, 2026

Copy link
Copy Markdown
Owner

요약

  • Pytorch CNN 실습.pdf의 P1~P5를 과제별 독립 실행 파일로 구성했습니다.
  • seed 42, 30 epoch × 3 조건의 BN/Dropout 절제 실험을 RTX 5050 Laptop GPU에서 직접 실행했습니다.
  • 피처맵, TensorBoard 태그, 혼동 행렬, 고신뢰 오답을 검증하고 PNG/CSV/JSON 공개 결과를 포함했습니다.
  • 체크포인트와 TensorBoard 이벤트는 재생성 가능한 대용량 파일이라 Git에서 제외했습니다.

실제 결과

  • BN + Dropout: 85.06%
  • No BN: 최저 test loss 시점 79.43%
  • No Dropout: 최저 test loss 시점 87.04%, 과적합 감지 16 epoch
  • baseline 최약 recall: cat 61.8%, 최대 혼동: cat → dog 17.2%

검증

  • ruff format --check: 통과
  • ruff check: 통과
  • pytest: 4 passed
  • P1~P5 CUDA 실행 완료
  • PNG 12개 및 JSON 5개 무결성 확인

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant