{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.md","pages":[{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","unit":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","unitId":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","chapter":"course","number":1}],"aliases":[{"oldId":"b-625cb03f7ec2eccd9d69cf82","targetId":"b-625cb03f7ec2eccd9d69cf82","page":"reading-opening","operation":"move","beforeSHA256":"7d57f9e628741b07c2704a32d4b4c2e5ca44d512d696e3562c18003cb33df7d1","afterSHA256":"7d57f9e628741b07c2704a32d4b4c2e5ca44d512d696e3562c18003cb33df7d1","reason":"Same source body in authored semantic page"},{"oldId":"b-329753b51cbe40be5672c43e","targetId":"b-329753b51cbe40be5672c43e","page":"reading-opening","operation":"move","beforeSHA256":"bf80604448c3da5790cc05643279a18850ecb31e2362838d98b8de81d4734791","afterSHA256":"bf80604448c3da5790cc05643279a18850ecb31e2362838d98b8de81d4734791","reason":"Same source body in authored semantic page"},{"oldId":"b-852a88e49740b5ec984c592f","targetId":"b-852a88e49740b5ec984c592f","page":"reading-opening","operation":"move","beforeSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","afterSHA256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","reason":"Same source body in authored semantic page"},{"oldId":"b-f6ee668e7318c880fc055640","targetId":"b-f6ee668e7318c880fc055640","page":"reading-opening","operation":"move","beforeSHA256":"75bc3d13109b5a891b82a4daefd668c001abde4ddf12e5ad4e20e6fe674663c6","afterSHA256":"75bc3d13109b5a891b82a4daefd668c001abde4ddf12e5ad4e20e6fe674663c6","reason":"Same source body in authored semantic page"},{"oldId":"b-f57ba85d7c600c45d58333d9","targetId":"b-f57ba85d7c600c45d58333d9","page":"reading-opening","operation":"move","beforeSHA256":"2e8bcdfb77f7a43c5411ad505709cfa540d0684eb52c7d85a5aaaacaede11f95","afterSHA256":"2e8bcdfb77f7a43c5411ad505709cfa540d0684eb52c7d85a5aaaacaede11f95","reason":"Same source body in authored semantic page"},{"oldId":"b-381abe845231afc57e269ccc","targetId":"b-381abe845231afc57e269ccc","page":"reading-opening","operation":"move","beforeSHA256":"b9e822539ccd14346c56f0133a31f40d09f6260e87ee416a08d197b5f2016ca7","afterSHA256":"b9e822539ccd14346c56f0133a31f40d09f6260e87ee416a08d197b5f2016ca7","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-625cb03f7ec2eccd9d69cf82","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-625cb03f7ec2eccd9d69cf82","content":"[LAB 01 · 固定表示：标签预算与读出](/course/notebooks/4dd8605ed9904820919ea9e13d9e7d7c.html)","sha256":"7d57f9e628741b07c2704a32d4b4c2e5ca44d512d696e3562c18003cb33df7d1","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-329753b51cbe40be5672c43e","prerequisites":[],"relatedBlocks":[]},{"id":"b-329753b51cbe40be5672c43e","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-329753b51cbe40be5672c43e","content":"# 从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sha256":"bf80604448c3da5790cc05643279a18850ecb31e2362838d98b8de81d4734791","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-625cb03f7ec2eccd9d69cf82","nextBlock":"b-852a88e49740b5ec984c592f","prerequisites":[{"id":"b-625cb03f7ec2eccd9d69cf82","url":"#b-625cb03f7ec2eccd9d69cf82","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-852a88e49740b5ec984c592f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-852a88e49740b5ec984c592f","content":"按问题继续","sha256":"e91f24e4e4ca40e2099a0d81d30db6568f4bf2074b499eeee6e3959617bd97ac","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-329753b51cbe40be5672c43e","nextBlock":"b-f6ee668e7318c880fc055640","prerequisites":[{"id":"b-329753b51cbe40be5672c43e","url":"#b-329753b51cbe40be5672c43e","role":"section-title"},{"id":"b-625cb03f7ec2eccd9d69cf82","url":"#b-625cb03f7ec2eccd9d69cf82","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-f6ee668e7318c880fc055640","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-f6ee668e7318c880fc055640","content":"固定CNN与缓存特征、标签预算、kNN/linear和独立评价","sha256":"75bc3d13109b5a891b82a4daefd668c001abde4ddf12e5ad4e20e6fe674663c6","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-852a88e49740b5ec984c592f","nextBlock":"b-f57ba85d7c600c45d58333d9","prerequisites":[{"id":"b-329753b51cbe40be5672c43e","url":"#b-329753b51cbe40be5672c43e","role":"section-title"},{"id":"b-625cb03f7ec2eccd9d69cf82","url":"#b-625cb03f7ec2eccd9d69cf82","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-f57ba85d7c600c45d58333d9","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-f57ba85d7c600c45d58333d9","content":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头。**编码器参数全部冻结**；kNN 储存目标训练标签，线性读出用目标训练标签拟合权重。因此“冻结表示评价”并不等于“评价过程不需要标签”。","sha256":"2e8bcdfb77f7a43c5411ad505709cfa540d0684eb52c7d85a5aaaacaede11f95","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f6ee668e7318c880fc055640","nextBlock":"b-381abe845231afc57e269ccc","prerequisites":[{"id":"b-329753b51cbe40be5672c43e","url":"#b-329753b51cbe40be5672c43e","role":"section-title"},{"id":"b-625cb03f7ec2eccd9d69cf82","url":"#b-625cb03f7ec2eccd9d69cf82","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-381abe845231afc57e269ccc","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/484cd0a1452d4c14857fdf073dc58b31.html#b-381abe845231afc57e269ccc","content":"对同一标签预算比较 kNN/linear，是改变读出；对同一读出比较不同预算，是改变标签数量。共享图像、划分、预处理和源模型排除了这些因素的变动，但一个小任务、一个 seed 和固定超参数仍不足以给出普遍方法排名。","sha256":"b9e822539ccd14346c56f0133a31f40d09f6260e87ee416a08d197b5f2016ca7","chapter":{"id":"course","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"unit":{"id":"notebook-blocks/484cd0a1452d4c14857fdf073dc58b31","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头"},"page":{"id":"reading-opening","title":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","number":1},"sectionTitle":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","currentQuestion":"从源 CNN 中取最后一个 32 维 ReLU 特征，丢弃原十分类头","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-f57ba85d7c600c45d58333d9","nextBlock":null,"prerequisites":[{"id":"b-329753b51cbe40be5672c43e","url":"#b-329753b51cbe40be5672c43e","role":"section-title"},{"id":"b-625cb03f7ec2eccd9d69cf82","url":"#b-625cb03f7ec2eccd9d69cf82","role":"problem-context"}],"relatedBlocks":[]}]}