{"resolved":true,"match_type":"uuid","source_type":"original_book","book_id":"fa2211c7-60b5-424b-a841-a97c7196aed1","title":"数据资本时代","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787508694313","requested_by":"printed_page","book":{"source_type":"original_book","book_id":"fa2211c7-60b5-424b-a841-a97c7196aed1","id":"fa2211c7-60b5-424b-a841-a97c7196aed1","uuid":"fa2211c7-60b5-424b-a841-a97c7196aed1","title":"数据资本时代","normalized_title":"数据资本时代","author":null,"edition":null,"publisher":null,"year":null,"isbn":"9787508694313","isbns":["9787508694313","9783642038211"],"parsed":true,"ai_readable":true,"document_type":"book","page_count":278,"navigation_url":"https://workshop.closeclaw.top/ai/book/fa2211c7-60b5-424b-a841-a97c7196aed1","title_url":"https://workshop.closeclaw.top/ai/title/%E6%95%B0%E6%8D%AE%E8%B5%84%E6%9C%AC%E6%97%B6%E4%BB%A3","fixed_page_url_template":"https://workshop.closeclaw.top/ai/book/fa2211c7-60b5-424b-a841-a97c7196aed1/page/{printed_page}","path_api_url_template":"https://workshop.closeclaw.top/api/ai/book/fa2211c7-60b5-424b-a841-a97c7196aed1/page/{printed_page}","read_url":"https://workshop.closeclaw.top/api/ai/read/fa2211c7-60b5-424b-a841-a97c7196aed1","page_url":"https://workshop.closeclaw.top/api/ai/books/fa2211c7-60b5-424b-a841-a97c7196aed1/page","search_url":"https://workshop.closeclaw.top/api/ai/books/fa2211c7-60b5-424b-a841-a97c7196aed1/search"},"page":{"printed_page":77,"pdf_page":97,"physical_index":97,"physical_index_base":0,"printed_page_status":"recognized","page_sha256":"937ca9062c722177a5af80f7099287cc91501198a3b0a8fd3c08b314d1051c54"},"chapter":"第3章","content":"要使这样一个机器学习方法有效运行，机器学习系统必须满足两个条件。首先，机器学习系统一开始需要大量的数据来训练自己，并且要搞清楚嵌入数据里的内容。②例如，谷歌利用其Web（万维网)上的所有文本，来找出谷歌翻译工具的单词使用概率模式。其次，机器学习系统必须得到频繁的反馈，以便随着时间的推移，它能够根据具体的情况变化进行自我调整，以超越最初的训练成绩。新一代机器学习系统在数据中寻找的不仅是模式—它们以一种更微妙、更具有识别作用的方式来使用反馈数据。例如，它们可以像人类的记忆一样降低旧数据的价值。\n\n第4章 市场的复兴\n\n①实际上，推荐系统并不需要真正了解消费者的偏好。例如消费者甲买了商品A，然后系统发现买了商品A的乙和丙都买了商品B，于是系统就把商品B推荐给甲。这种通过发现和目标消费者购买行为相似的消费者，并把他们购买过而目标消费者没有购买的商品推荐给目标消费者，或者推荐经常与目标消费者购物车里的商品一起被购买的商品，就是推荐系统中最主流的算法之一—协同过滤(collaborativefiltering)。在该算法中，系统不需要真正了解消费者的偏好或者为什么要买某商品。亚马逊自身就是协同过滤的先驱研究者，读者可以参考著名论文 G. Linden, B. Smith, J. York, Amazon. com recommendations:Item-to-itemcollaborative filtering, IEEE Internet Computing7(1)(2003)76.——译者注\n\n2这个条件不必然，它取决于算法。例如谷歌的翻译算法更多的是发现模式而非挖掘内容。一译者注\n\n077","previous_page":76,"next_page":78}