{"code":0,"data":{"blobItems":[{"url":"/tsxs-esm/p1.9f17edb1c43a351f.png","width":1956,"height":1046},{"url":"/tsxs-esm/p3.28c4608fa1d5a548.png","width":3600,"height":1044},{"url":"/tsxs-esm/p3.28c4608fa1d5a548.png","width":3600,"height":1044},{"url":"/tsxs-esm/p8c.16c7f5dbb2a4351e.png","width":1322,"height":760},{"url":"/tsxs-esm/p4.a1612722509e36f0.png","width":1600,"height":766},{"url":"/tsxs-esm/p15.04a8917989858457.png","width":2560,"height":2984},{"url":"/tsxs-esm/p2.355c8a98346a2851.png","width":2312,"height":1432},{"url":"/tsxs-esm/p4.a1612722509e36f0.png","width":1600,"height":766},{"url":"data:image/png;base64,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","width":1063,"height":623},{"url":"/tsxs-esm/p7.f9ac0d3e8b62d866.png","width":4290,"height":798},{"url":"/tsxs-esm/p9.fbe7cbd4eb3bed10.png","width":4834,"height":2458},{"url":"/tsxs-esm/p6.2453a123744eaa99.png","width":2462,"height":1278},{"url":"/tsxs-esm/p12.1c6c74c23a2ce0f5.png","width":3898,"height":1716},{"url":"/tsxs-esm/p16.9ed3878666dd10ad.png","width":4170,"height":1564},{"url":"/tsxs-esm/p14.e6403f2b402a9023.png","width":1786,"height":2368},{"url":"/tsxs-esm/p5.de9dcdc1bb78643f.png","width":2466,"height":1326},{"url":"/tsxs-esm/p13.9281a0541fe714be.png","width":3608,"height":484},{"url":"/tsxs-esm/p10.62419684d8e5f58a.png","width":1125,"height":391},{"url":"/tsxs-esm/p11.1650c02590cb9a70.png","width":2234,"height":1242},{"url":"/tsxs-esm/p8.ae7ccc5808bbee7f.png","width":570,"height":515},{"url":"/tsxs-esm/p0.835bbb266490c65a.png","width":638,"height":924}],"renderInfo":{"tocStack":[{"id":"why-write-it","level":1,"text":"为什么写此文"},{"id":"symbolism-and-connectionism","level":1,"text":"什么是『智能』？ 符号主义和联结主义"},{"id":"find-predicted-point","level":1,"text":"进一步的例子：找出点的预测"},{"id":"loss-function","level":1,"text":"理解『拟合』以及评估拟合效果的『损失函数』"},{"id":"func-neural-network-forward-propagation","level":1,"text":"从函数到『神经网络』和『正向传播 』"},{"id":"nonlinear-fitting-and-activation-functions","level":1,"text":"非线性拟合和激活函数"},{"id":"matrix","level":1,"text":"多参数和引入矩阵"},{"id":"over-fitting","level":1,"text":"参数和层是不是越多越好？ 什么叫『过拟合』了？"},{"id":"what-is-training","level":1,"text":"“训练神经网络” 中的训练指的是什么？"},{"id":"gradient-descent","level":1,"text":"寻找损失函数的最小值：梯度下降法"},{"id":"back-propagation","level":1,"text":"反向传播与链式法则"},{"id":"neural-networks-and-llm","level":1,"text":"神经网络是大模型的基础"},{"id":"cnn-rnn","level":1,"text":"前大模型时代的 CNN RNN 架构"},{"id":"tokenize-and-embedding","level":1,"text":"分词与词嵌入"},{"id":"attention-is-all-you-need","level":1,"text":"Attention Is All You Need —— Transformer 结构"},{"id":"self-attention-qkv","level":1,"text":"自注意力机制 (Self-Attention) 与 QKV"},{"id":"multi-head-attention","level":1,"text":"多头注意力"},{"id":"translate","level":1,"text":"Transformer 原始论文的目标：翻译任务"},{"id":"openai-and-gpt","level":1,"text":"OpenAI 与 GPT"},{"id":"ai-words","level":1,"text":"从原理出发看看现在 AI 术语底下的技术含义"},{"id":"token-分词和词嵌入","level":2,"text":"Token 分词和词嵌入"},{"id":"权重-(weights)-和偏置-(bias)-和参数量","level":2,"text":"权重 (Weights) 和偏置 (Bias) 和参数量"},{"id":"大模型的训练","level":2,"text":"大模型的训练"},{"id":"大模型为什么能表现出「智能」","level":2,"text":"大模型为什么能表现出「智能」"},{"id":"rag-是什么？为什么总是提到『向量数据库』","level":2,"text":"RAG 是什么？为什么总是提到『向量数据库』"},{"id":"什么是大模型的『推理』","level":2,"text":"什么是大模型的『推理』"},{"id":"但是，大模型怎么知道什么时候该停下来-？-","level":2,"text":"但是，大模型怎么知道什么时候该停下来 ？ "},{"id":"大模型的-function-calling-和-mcp-","level":2,"text":"大模型的 function calling 和 MCP "},{"id":"多智能体和工作流编排","level":2,"text":"多智能体和工作流编排"},{"id":"怎么写好-prompt-?","level":2,"text":"怎么写好 prompt ?"},{"id":"其他术语","level":2,"text":"其他术语"},{"id":"<|im_end|>","level":1,"text":"<|im_end|>"}]}}}