Tag: #역전파
Kubernetes, DevOps and MLOps, written from things we ran and measured ourselves · 2 posts
What Convolutional Neural Networks Learn: Filters, Receptive Fields, Backpropagation, and LeNet & AlexNet
We examine "what" Convolutional Neural Networks (CNNs) learn in images, starting from the smallest units. Using numpy examples, we run through 3×3 filters sliding across images multiplying and adding, receptive fields ex
2026-10-17 · 14 min read #비전 모델#합성곱#CNN#역전파#경사 하강ResNet: Why Residual Connections Made Deep Networks Trainable and Batch Normalization
We examine how ResNet solved the "degradation paradox" where stacking more layers actually increases training error. If a block learns only the "residual to add to the input" rather than the entire output, the gradient f
2026-10-17 · 13 min read #비전 모델#ResNet#잔차 연결#배치 정규화#기울기 소실