{"version":2,"relationshipURLBase":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html","relationshipResolution":"Resolve relative relationship URLs against relationshipURLBase; complete records and absolute URLs are in blocks.","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html","markdown":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.md","pages":[{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","unit":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","unitId":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","chapter":"course","number":1}],"aliases":[{"oldId":"b-1a06676ae65b3f867d2e5510","targetId":"b-1a06676ae65b3f867d2e5510","page":"reading-opening","operation":"move","beforeSHA256":"343aaa499ee36d5531eab9b8b2e10a88939c1d9efb4bb1251925f2c15e332b89","afterSHA256":"343aaa499ee36d5531eab9b8b2e10a88939c1d9efb4bb1251925f2c15e332b89","reason":"Same source body in authored semantic page"},{"oldId":"b-84df94682c54068ee295945b","targetId":"b-84df94682c54068ee295945b","page":"reading-opening","operation":"move","beforeSHA256":"b9f1206df008ff40d381aa8631168880bad50c8f06e787fed2087e311e612392","afterSHA256":"b9f1206df008ff40d381aa8631168880bad50c8f06e787fed2087e311e612392","reason":"Same source body in authored semantic page"},{"oldId":"b-a1e86f723293a37d31c47a82","targetId":"b-a1e86f723293a37d31c47a82","page":"reading-opening","operation":"move","beforeSHA256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","afterSHA256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","reason":"Same source body in authored semantic page"},{"oldId":"b-d5bf27dcf630bdcd45f0401c","targetId":"b-d5bf27dcf630bdcd45f0401c","page":"reading-opening","operation":"move","beforeSHA256":"e5185ec67dda4a18681db2d55dc5cbe0c312c9a5519949e3ded12e96d49c997f","afterSHA256":"e5185ec67dda4a18681db2d55dc5cbe0c312c9a5519949e3ded12e96d49c997f","reason":"Same source body in authored semantic page"},{"oldId":"b-6446c5ab2bd640ff779eef8f","targetId":"b-6446c5ab2bd640ff779eef8f","page":"reading-opening","operation":"move","beforeSHA256":"63be3c105638264d2aae901705b39efdac48b51792ecdcca46eeab48da6849b8","afterSHA256":"63be3c105638264d2aae901705b39efdac48b51792ecdcca46eeab48da6849b8","reason":"Same source body in authored semantic page"},{"oldId":"b-641cd76f90f6efa646f6627c","targetId":"b-641cd76f90f6efa646f6627c","page":"reading-opening","operation":"move","beforeSHA256":"4817fcc914228568a896cf868c7086a259c529d8b0058eafbf7d3c804220921e","afterSHA256":"4817fcc914228568a896cf868c7086a259c529d8b0058eafbf7d3c804220921e","reason":"Same source body in authored semantic page"},{"oldId":"b-5431d4e4f446e593816ee3ff","targetId":"b-5431d4e4f446e593816ee3ff","page":"reading-opening","operation":"move","beforeSHA256":"456a867b4dbb0433e3d8fbc338d2248f469f965ab45799669e1f4f2b2e54c557","afterSHA256":"456a867b4dbb0433e3d8fbc338d2248f469f965ab45799669e1f4f2b2e54c557","reason":"Same source body in authored semantic page"},{"oldId":"b-a5f3822f8d57134ad2ac7f44","targetId":"b-a5f3822f8d57134ad2ac7f44","page":"reading-opening","operation":"move","beforeSHA256":"e18be066685c20bd70873c87261ce3bc6e6c265ff1e95fada9349e79b886a4f0","afterSHA256":"e18be066685c20bd70873c87261ce3bc6e6c265ff1e95fada9349e79b886a4f0","reason":"Same source body in authored semantic page"}],"blocks":[{"id":"b-1a06676ae65b3f867d2e5510","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-1a06676ae65b3f867d2e5510","content":"[CLS 01 · 学习信号、表示与迁移](/course/notebooks/fb4365d604974dd19df492cb22516e44.html)","sha256":"343aaa499ee36d5531eab9b8b2e10a88939c1d9efb4bb1251925f2c15e332b89","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":null,"nextBlock":"b-84df94682c54068ee295945b","prerequisites":[],"relatedBlocks":[]},{"id":"b-84df94682c54068ee295945b","type":"h1","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-84df94682c54068ee295945b","content":"# 第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sha256":"b9f1206df008ff40d381aa8631168880bad50c8f06e787fed2087e311e612392","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-1a06676ae65b3f867d2e5510","nextBlock":"b-a1e86f723293a37d31c47a82","prerequisites":[{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-a1e86f723293a37d31c47a82","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-a1e86f723293a37d31c47a82","content":"主线阅读与优先实验","sha256":"42cba31deccba2a350b9b23bd0ca1dcaa4c4c5cb9d35c773e1fcb89124788ccd","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-84df94682c54068ee295945b","nextBlock":"b-d5bf27dcf630bdcd45f0401c","prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-d5bf27dcf630bdcd45f0401c","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-d5bf27dcf630bdcd45f0401c","content":"标签供给、文本目标接口、视图/队列/MAE图与表示迁移证据","sha256":"e5185ec67dda4a18681db2d55dc5cbe0c312c9a5519949e3ded12e96d49c997f","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-a1e86f723293a37d31c47a82","nextBlock":"b-6446c5ab2bd640ff779eef8f","prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-6446c5ab2bd640ff779eef8f","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-6446c5ab2bd640ff779eef8f","content":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新。ImageNet 分类预训练让同一个网络读过大量图像，并从分类错误中调整表示。今天先保留这个闭环，把注意力放到标签：更多图片，是否意味着我们已经获得更多合适的训练题？","sha256":"63be3c105638264d2aae901705b39efdac48b51792ecdcca46eeab48da6849b8","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-d5bf27dcf630bdcd45f0401c","nextBlock":"b-641cd76f90f6efa646f6627c","prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-641cd76f90f6efa646f6627c","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-641cd76f90f6efa646f6627c","content":"一张桌面图可以被标为“室内”，也可以要求框出两个杯子，还可以要求逐像素区分杯与桌面。这三种答案需要不同的判断与标注劳动。把粗类别集合拼在一起也不会自动得到细粒度答案：只标“杯子”的数据无法直接告诉我们哪一个是红杯。","sha256":"4817fcc914228568a896cf868c7086a259c529d8b0058eafbf7d3c804220921e","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-6446c5ab2bd640ff779eef8f","nextBlock":"b-5431d4e4f446e593816ee3ff","prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-5431d4e4f446e593816ee3ff","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-5431d4e4f446e593816ee3ff","content":"2016 年的同期访谈中，Scale 共同创始人 **Lucy Guo 与 Alexandr Wang** 介绍了将人工任务、人员组织与质量控制包装成可调用服务的做法；任务包括分类、数据提取等。这段经历说明，监督供给也有组织与工程问题。[同期访谈](https://www.ycombinator.com/blog/scale)","sha256":"456a867b4dbb0433e3d8fbc338d2248f469f965ab45799669e1f4f2b2e54c557","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-641cd76f90f6efa646f6627c","nextBlock":"b-a5f3822f8d57134ad2ac7f44","prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]},{"id":"b-a5f3822f8d57134ad2ac7f44","type":"p","scope":"document","url":"https://codingai-lec03.pages.dev/course/notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54.html#b-a5f3822f8d57134ad2ac7f44","content":"**一条路线是把人的判断组织得更有效；另一条路线是重新设计训练任务，让已有数据提供一部分答案。我们先借文字看看后一条路线怎样成立。**","sha256":"e18be066685c20bd70873c87261ce3bc6e6c265ff1e95fada9349e79b886a4f0","chapter":{"id":"course","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"unit":{"id":"notebook-blocks/40a5f51d7b11433ea6cce79aa2904a54","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新"},"page":{"id":"reading-opening","title":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","number":1},"sectionTitle":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","currentQuestion":"第二讲用一个具体闭环连接图像、网络、类别预测、人工标签、损失和参数更新","sourceSlides":[],"knowledgeBlock":null,"formulas":[],"previousBlock":"b-5431d4e4f446e593816ee3ff","nextBlock":null,"prerequisites":[{"id":"b-84df94682c54068ee295945b","url":"#b-84df94682c54068ee295945b","role":"section-title"},{"id":"b-1a06676ae65b3f867d2e5510","url":"#b-1a06676ae65b3f867d2e5510","role":"problem-context"}],"relatedBlocks":[]}]}