{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"2 · 精确搜索：模型给分，评价定义相关","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.md","pages":[{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","unit":"2 · 精确搜索：模型给分，评价定义相关","unitId":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","chapter":"course","number":1}],"aliases":[{"oldId":"b-81d20a3b20543fffdce3f2ee","targetId":"b-81d20a3b20543fffdce3f2ee","page":"reading-opening","operation":"move","beforeSHA256":"1a4de4df51c1a8787dc96ec401888dfe89d250842519319b9af7a931ca765eee","afterSHA256":"1a4de4df51c1a8787dc96ec401888dfe89d250842519319b9af7a931ca765eee","reason":"Same source body in authored semantic page"},{"oldId":"b-ee6ad4e330ef33dd0be09026","targetId":"b-ee6ad4e330ef33dd0be09026","page":"reading-opening","operation":"move","beforeSHA256":"7ef3582cd4ed063f0618142b3a3b3f0d84ffc05cebb3b557cb05a3392db41021","afterSHA256":"7ef3582cd4ed063f0618142b3a3b3f0d84ffc05cebb3b557cb05a3392db41021","reason":"Same source body in authored semantic page"},{"oldId":"b-2854fbf9ecb71c8c0544bae2","targetId":"b-2854fbf9ecb71c8c0544bae2","page":"reading-opening","operation":"move","beforeSHA256":"dc06c1e728a4b5740ba634b145bc9a1da6323031626734cb4555db5a744df806","afterSHA256":"dc06c1e728a4b5740ba634b145bc9a1da6323031626734cb4555db5a744df806","reason":"Same source body in authored semantic page"},{"oldId":"b-37df082b6f18307dcfcdc8c5","targetId":"b-37df082b6f18307dcfcdc8c5","page":"reading-opening","operation":"move","beforeSHA256":"8021c18afc52ba7b6612f43ecfb0fdd4789efe91cc9cd656984357d81a8514de","afterSHA256":"8021c18afc52ba7b6612f43ecfb0fdd4789efe91cc9cd656984357d81a8514de","reason":"Same source body in authored semantic page"},{"oldId":"b-368a9abdd56572cbf19d192d","targetId":"b-368a9abdd56572cbf19d192d","page":"reading-opening","operation":"move","beforeSHA256":"6a6a7b78abfe6364f4ef040e54cd2bf106294e030c2d6a73dcdad1dd69ddfa24","afterSHA256":"6a6a7b78abfe6364f4ef040e54cd2bf106294e030c2d6a73dcdad1dd69ddfa24","reason":"Same source body in authored semantic page"},{"oldId":"b-89c06f127494e049d6b2e431","targetId":"b-89c06f127494e049d6b2e431","page":"reading-opening","operation":"move","beforeSHA256":"a9db9edd976ff657caa5941b4245c0b52971ebde344e66c024f0e5ab01b0401f","afterSHA256":"a9db9edd976ff657caa5941b4245c0b52971ebde344e66c024f0e5ab01b0401f","reason":"Same source body in authored semantic page"},{"oldId":"b-6b1caba7d02f641fb007be33","targetId":"b-6b1caba7d02f641fb007be33","page":"reading-opening","operation":"move","beforeSHA256":"6981901db7db96134fa7095db2336567bb8c37f0f76f8e294f664880fb8c1808","afterSHA256":"6981901db7db96134fa7095db2336567bb8c37f0f76f8e294f664880fb8c1808","reason":"Same source body in authored semantic page"},{"oldId":"b-1bcb5b4c6781a806fbf9b561","targetId":"b-1bcb5b4c6781a806fbf9b561","page":"reading-opening","operation":"move","beforeSHA256":"aa8be1410199fba8c51caa638e575eceda30ec6f40f30bbfa56b0107d8f9e20f","afterSHA256":"aa8be1410199fba8c51caa638e575eceda30ec6f40f30bbfa56b0107d8f9e20f","reason":"Same source body in authored semantic page"},{"oldId":"block-5ba73f2552e945a7a94eb9043c0ac480--2-精确搜索模型给分评价定义相关","targetId":"b-368a9abdd56572cbf19d192d","operation":"retain-source-identity"}],"blocks":[{"id":"b-81d20a3b20543fffdce3f2ee","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-81d20a3b20543fffdce3f2ee","content":"[LAB 12 · CLIP延伸：缓存与检索（近似索引选读）](/course/notebooks/921eeaf784284ce19568c6ca7619b55a.html)","sha256":"1a4de4df51c1a8787dc96ec401888dfe89d250842519319b9af7a931ca765eee","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-ee6ad4e330ef33dd0be09026","prerequisites":[],"relatedBlocks":[]},{"id":"b-ee6ad4e330ef33dd0be09026","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-ee6ad4e330ef33dd0be09026","content":"# 2 · 精确搜索：模型给分，评价定义相关","sha256":"7ef3582cd4ed063f0618142b3a3b3f0d84ffc05cebb3b557cb05a3392db41021","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-81d20a3b20543fffdce3f2ee","nextBlock":"b-2854fbf9ecb71c8c0544bae2","prerequisites":[{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-2854fbf9ecb71c8c0544bae2","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-2854fbf9ecb71c8c0544bae2","content":"并入相关学习路径","sha256":"dc06c1e728a4b5740ba634b145bc9a1da6323031626734cb4555db5a744df806","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-ee6ad4e330ef33dd0be09026","nextBlock":"b-37df082b6f18307dcfcdc8c5","prerequisites":[{"id":"b-ee6ad4e330ef33dd0be09026","url":"#b-ee6ad4e330ef33dd0be09026","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-37df082b6f18307dcfcdc8c5","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-37df082b6f18307dcfcdc8c5","content":"同一CLIP缓存、exact/IVF搜索、延迟/扫描量、区域交互基线","sha256":"8021c18afc52ba7b6612f43ecfb0fdd4789efe91cc9cd656984357d81a8514de","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-2854fbf9ecb71c8c0544bae2","nextBlock":"b-368a9abdd56572cbf19d192d","prerequisites":[{"id":"b-ee6ad4e330ef33dd0be09026","url":"#b-ee6ad4e330ef33dd0be09026","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-368a9abdd56572cbf19d192d","type":"h2","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-368a9abdd56572cbf19d192d","content":"## 2 · 精确搜索：模型给分，评价定义相关","sha256":"6a6a7b78abfe6364f4ef040e54cd2bf106294e030c2d6a73dcdad1dd69ddfa24","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-37df082b6f18307dcfcdc8c5","nextBlock":"b-89c06f127494e049d6b2e431","prerequisites":[{"id":"b-ee6ad4e330ef33dd0be09026","url":"#b-ee6ad4e330ef33dd0be09026","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-89c06f127494e049d6b2e431","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-89c06f127494e049d6b2e431","content":"对于一个文本查询 $q$，精确检索逐个计算 $s_i=q^\\top v_i$ 并取最大的 K 项。归一化不等于校准：cosine 只是排序分数，不是“该图片正确的概率”。同一矩阵转置即可图搜文；不需要另训练一个模型，但两个方向的候选集合和评价单位不同。","sha256":"a9db9edd976ff657caa5941b4245c0b52971ebde344e66c024f0e5ab01b0401f","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"q","definitionContext":"2 · 精确搜索：模型给分，评价定义相关"},{"latex":"s_i=q^\\top v_i","definitionContext":"2 · 精确搜索：模型给分，评价定义相关"}],"previousBlock":"b-368a9abdd56572cbf19d192d","nextBlock":"b-6b1caba7d02f641fb007be33","prerequisites":[{"id":"b-368a9abdd56572cbf19d192d","url":"#b-368a9abdd56572cbf19d192d","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6b1caba7d02f641fb007be33","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-6b1caba7d02f641fb007be33","content":"本册规定：一个查询对应源类别的四个视图都算相关。**语义 Hit@K** 为“前 K 中至少有一个相关图”的查询比例；它不等于召回全部四个相关图的比例。**ANN top-3 overlap** 为 $|A_3(q)\\cap E_3(q)|/3$ 的平均值，评价近似索引是否找回精确搜索的邻居，完全不使用类别标签。这两项必须分开：模型的精确排序可能错，ANN 也可能完美复制这个错误。","sha256":"6981901db7db96134fa7095db2336567bb8c37f0f76f8e294f664880fb8c1808","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[{"latex":"|A_3(q)\\cap E_3(q)|/3","definitionContext":"2 · 精确搜索：模型给分，评价定义相关"}],"previousBlock":"b-89c06f127494e049d6b2e431","nextBlock":"b-1bcb5b4c6781a806fbf9b561","prerequisites":[{"id":"b-368a9abdd56572cbf19d192d","url":"#b-368a9abdd56572cbf19d192d","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-1bcb5b4c6781a806fbf9b561","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480.html#b-1bcb5b4c6781a806fbf9b561","content":"当前小集 exact 文搜图 Hit@3=1.0、图搜文 Hit@1=1.0。它只能说明这些照片和查询能被本模型匹配，不能支持未知照片或复杂关系的泛化结论。LAB 04 的左右关系反例正好显示：整图对象匹配成功不自动推出关系绑定正确。","sha256":"aa8be1410199fba8c51caa638e575eceda30ec6f40f30bbfa56b0107d8f9e20f","chapter":{"id":"course","title":"2 · 精确搜索：模型给分，评价定义相关"},"unit":{"id":"notebook-blocks/5ba73f2552e945a7a94eb9043c0ac480","title":"2 · 精确搜索：模型给分，评价定义相关"},"page":{"id":"reading-opening","title":"2 · 精确搜索：模型给分，评价定义相关","number":1},"sectionTitle":"2 · 精确搜索：模型给分，评价定义相关","currentQuestion":"2 · 精确搜索：模型给分，评价定义相关","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6b1caba7d02f641fb007be33","nextBlock":null,"prerequisites":[{"id":"b-368a9abdd56572cbf19d192d","url":"#b-368a9abdd56572cbf19d192d","role":"section-title"},{"id":"b-81d20a3b20543fffdce3f2ee","url":"#b-81d20a3b20543fffdce3f2ee","role":"problem-context"}],"relatedBlocks":[]}]}