【转】Google Analytics 团队:背后的故事与创业启示

中文版本:http://tech2ipo.com/59627

英文版本:http://www.attendly.com/the-real-story-on-how-google-analytics-got-started/

 

There’s nothing small about Google Analytics. After all, it’s used by around half of the world’s most popular websites.

While current startups are still dreaming of the day when they can use the phrase “market share,” Google Analytics already has the largest market share in its industry.

But this is far from how it all began. Before being acquired by Google, a small web consultancy team created a product called Urchin. This is the story of how they made it big.

Where it all started

In 1995, the founders of Google Analytics built a company called Web Depot, which provided web development and hosting services to San Diego businesses. It had four co-founders: Paul Muret was both the CEO and the engineering/technical end of the company, Jack Ancone was the CFO and brothers Brett and Scott Crosby took care of business development and sales.

Two years later, Paul wrote the first version of their analytics software, Urchin. It was planned to just be part of their services until he had a “lightbulb moment“:

“One of our large clients was struggling with the fact that it took 24 hours to process a single day’s worth of website tracking results. We tried out our new analytics tool, andit took 15 minutes to process the same data. That’s when the light bulb went off – that Urchin was for real.”

So the team were faced with a choice: keep providing web consultancy services, or pivot and focus on building and selling Urchin?

They chose the latter and slowly moved away from doing hosting and development work. It wasn’t an easy process, but it was a move that paid off in the long run. As Brett Crosby said:

“We built our business around a very scalable product, which allowed us to do things like target hosting companies and get massive numbers of users with one deal, rather than focusing on one very complex deal.”

As they started landing top web hosts as customers, Urchin quickly became the standard analytics software for thousands of websites.

As a result, it didn’t take long for Google to notice them.

From Urchin to Google Analytics

When startup founders attend trade shows, there are some common things they expect to achieve. Get new customers. Generate interest about their product. Or if they’re really lucky, bring in some new investors.

So when Google representatives approached the Urchin team at a trade show in 2004, it’s likely that the encounter was beyond anything they expected at the time. An offer was made, and negotiations began, which would take several months to wrap up.

The deal eventually went through on Brett Crosby’s wedding day on March 2005.

“I was in my tux, literally just about to walk down the aisle when I signed the contract,” herecalled in an interview.

Where Are They Now?

After the acquisition, Urchin’s web-hosted analytics software, Urchin On Demand, became the Google Analytics we’re now familiar with. Their client-hosted version, simply named Urchin, was rebranded as “Urchin from Google”.

But more than just the product names changed after the acquisition.

Paul Muret became the Director of Engineering at Google Analytics. After spending more than a decade working on the product, he gave his first keynote speech during the Google Analytics Summit last October.

Jack Ancone became the Senior Director of New Business Development at Google.

Brett Crosby is now the Director of Product Marketing at Google. He also sits on the board of directors at Euclid Inc., which provides in-store consumer analytics for retailers.

Like the other founders, Scott Crosby also worked for Google as Senior Program Manager after the acquisition. Upon joining Google he made 2 key decisions: to live in San Francisco and not drive to work. So with his brother Brett, he founded SF2G, a community that promotes cycling to work. He left Google in 2010 and is also now the COO of Euclid Inc.

As for Urchin, while its web-hosted version lives on as Google Analytics, the client-hosted version wasn’t so lucky. Though they released 3 more version upgrades since the acquisition, sales for Urchin software were discontinued on March 2012.

And the moral is…

Google Analytics has a history that’s as long and as colorful as the reports it generates. Even if its startup days were over a decade ago, there are a few things up and coming entrepreneurs can learn from this story.

Startup Lesson #1 – A successful idea doesn’t need to be unique or flashy. But it has to be useful.

The Urchin team’s “big idea” wasn’t their development and hosting services, but the analytics tool they developed to process website tracking results.

Analytics software isn’t really the first thing that comes to mind when you try to think of a product idea that could change the world.

But the thing is, they did change the world – at least online. They provided something faster and more user-friendly than what was available at the time, which became the widespread standard.

Startup Lesson #2 – Recognize your opportunities for pivoting.

Sure, Urchin software could have been just another feature in their web hosting and development services, but if the team had persevered with their original business plan, they wouldn’t have realized the full potential of their new product.

Some Things Don’t Change

It’s clear that whatever they pursue, the founders of Google Analytics are still passionate about empowering people by giving them easy access to important data. The Crosby brothers are still working in analytics, only now it’s for brick-and-mortar retail interactions. And Paul Muret is still working on the product he’s been building all these years.

While there are lessons to be learned, the history of Google Analytics proves how successful you can be despite hurdles, pivots and taking risks, which is encouraging for those of us in the early chapters of our own startup stories.

Manus AI:跨境电商虚拟运营团队的实战解析 电商自动化已从单点AI工具升级为端到端Agent协作系统。本文围绕‘自治Agent’与‘动态知识图谱’两大核心技术概念,解析其如何将市场感知、选品决策、建站执行、广告冷启动和数据反哺等环节封装为可配置、可审计、可复用的轻量级虚拟团队。区别于依赖大模型幻觉的传统方案,Manus采用专家规则引擎+轻量微调+实时政策同步的混合架构,显著提升决策确定性与业务鲁棒性。适用于独立站创业者、Dropshipping操盘手及中小跨境团队,在不招人、不学代码前提下实现中型团队级响应速度与策略颗粒度。 阅读详情

相关推荐

BI平台选型核心:数据接入、准备、可视化与预测的四大底层能力

商业智能(BI)平台的本质不是图表生成器,而是企业数据决策流的中枢操作系统。其技术价值在于将原始数据转化为可行动洞察的工程化能力——这依赖于四大基础能力:稳定高效的数据接入(支持多源异构、增量同步与语义理解)、灵活可控的数据准备(处理空值、嵌套结构与业务逻辑)、直观可信的可视化交互(支持动态标记、下钻与跨维度关联),以及可解释的预测建模能力。这些能力共同决定了分析响应速度、业务自助程度与IT运维负担。Tableau、SAP Analytics Cloud、Tibco Spotfire和Qlik Sense在

weixin_30653023的博客 1000

成功团队就是汇聚好的人才?谷歌“亚里士多德项目”为你揭开“惊人”谜底

谷歌一直试着将数据科学与管理结合起来。几年前, Google People Analytics 团队就试着让算法参与到人力资源管理中。2012年,为了找出打造完美团队的规律(完美团队就是好的成员的集合吗),他们又启动了代号亚里士多德项目(Project Aristotle)。 读过1989年Stephen Covey《高效能人士的7个习惯》一书的读者可能还记忆犹新,书中谈到,高效团队

简简单单,平平淡淡 1919

移动独行侠:从场景洞察到MVP实战,打造极致移动产品

在移动互联网领域,MVP(最小可行产品)和敏捷开发是应对激烈竞争、实现产品从0到1突破的核心方法论。MVP强调以最小成本快速验证核心价值,而敏捷开发则通过快速迭代响应市场反馈。其技术价值在于,能帮助开发者或小团队在资源有限的情况下,高效试错、降低风险,并聚焦于解决用户真实场景下的痛点。在应用场景上,无论是工具类App、内容产品还是垂直服务,通过深入的用户场景洞察,结合移动设备的原生特性(如传感器、LBS),都能创造出Web或PC端难以复制的体验。本文以“移动独行侠”(Mobile Mavericks)策略为

weixin_30417487的博客 393

谷歌和它的完美团队

谷歌最近的一项研究表明,团队成员间的互动比团队成员的构成更重要。谷歌研究员总结了谷歌的高效率团队都具备的五个特征。安全感、可靠性、目的性、使命感和影响力对团队效率的影响比职位、技能水平和工作地点带来的影响要大。\u0026#xD;\n\u0026#xD;\n在Aristotle这个研究项目里,谷歌收集了来自很多领域的数据。研究员对团队负责人进行采访,把他们对团队效率驱动因素的看法跟自我评价和销售业...

啊啊啊啊2 346

非科班转AI工程师:业务分析师的四阶段工程化跃迁路径

机器学习工程师(MLE)岗位正从纯算法能力转向‘业务-数据-模型-系统’全链路工程能力。对于具备SQL、AB测试、报表分析经验的商业分析师而言,其天然拥有的问题抽象能力、数据敏感度与跨职能沟通带宽,恰是当前LLM应用落地和MLOps实践中最稀缺的复合素养。本文基于真实可复现的十年转型案例,解析如何将业务动作转化为技术资产:从可审计的特征管道构建,到CLI工具与API服务化封装,再到端到端模型交付与提示词工程治理。不依赖学历镀金或刷题突击,而是以‘解决一个具体业务痛点’为最小闭环,持续积累组织可见的技术影响力

dezhen7015的博客 367

AI 生成 CAN 通信测试用例实战

本文分享了AI辅助汽车电子测试用例编写的实战经验。通过某车企CAN网络测试项目案例,对比传统方式(需27天)与AI辅助方式(仅5天)的效率差异,展示81%的效率提升。文章详细介绍了从DBC文件提取信号信息的方法,包括DBC文件结构解析、信号定义说明,并提供了完整的Python解析脚本代码,可将信号信息导出为JSON或Excel格式,为后续AI批量生成测试用例奠定基础。

小明的博客 694

对Google内部创业的一些思考

对Google内部创业的一些思考今天看了一篇文章《6天旅程之Google内部创业》,自己有一些感触,说的更具体点,应该是非常激动。记得自己在毕业设计阶段(我的毕业设计是在就职的公司做的),当时特别想给自己做一个实时提醒软件使用使用,无奈没有这样的时间,因为公司没有给我这样的时间。就我个人的观点来说,很喜欢Google这样的内部创业机制,主要因为以下两点: 1 时间掌握很好。Google并没

AaronLu的专栏 551

ICCV 2025 | 哈工大、清华等提出首个双曲学习框架:HLFormer,革新局部相关视频检索

离原点越远,代表的语义越具体(部分)。现有方法大多在传统的欧几里得空间中对视频和文本进行建模,但这往往会导致对视频内在层次结构的错误表示,从而影响检索的准确性。通过UMAP对学习到的帧嵌入进行可视化(如下图),可以观察到,来自同一时刻的视频帧在嵌入空间中聚集在一起,清晰地展示了模型对时序关系的捕捉能力。下图的双曲空间可视化进一步证实了模型的层次建模能力:离原点越近的点,代表的语义层级越高、粒度越粗糙,反之亦然。:既然文本查询描述的是视频的一部分,那么在几何表示上,文本的嵌入向量应该从属于视频的嵌入向量。

我爱计算机视觉 1082

Google推出Mobile Web Specialist认证

文 / Google 网络开发者培训部项目经理 Sarah Clark 如果您是一名网络开发者,就一定知道这是一个鱼龙混杂的市场,而您也想让自己从其他网络开发者中脱颖而出。想要展示您具备构建自适应和灵活网络应用的技能吗? Google Developers 认证团队荣幸地宣布推出 Mobile Web Specialist 认证。 这项新认证基于我们对市场的完全分析,面向需要掌握

谷歌开发者 811

电赛控制-----经验分享

1.赛前准备 先简单介绍一下电赛。电赛是两年一届,单数年是大电赛,全称是全国大学生电子设计大赛,之前由瑞萨电子赞助,所以之前也叫瑞萨杯,从19年开始赞助方变成了TI公司,偶数年是小电赛,全称是“TI杯”模电邀请赛,(这里不得不提TI公司的实力是愈来愈大了),一般来说小电赛的含金量会略低于大电赛,而且对与想拿国奖的同学来说还是要参加大电赛,因为小电赛一般每个学校会限制只有两个队伍进国赛,而大电赛是没有这个限制的,而且我们学校电赛保研的话,大电赛只要求国二,而小电赛只有国一才可以保,这样难度就增加了很多。..

qq_35656655的博客 1万+

使用运放构建电压跟随器

运放,是运算放大器的简称 运算放大器是用模拟电子器件(如晶体管,场效应管,二极管等)构成的模拟集成电路,其特点是有很高的放大倍数和抗干扰能力,因此可以被设计成各种用途的派生电路,如电压比较器、窗口电路、波形发生电路等等,其中也有电压跟随器。 电压跟随器是一种具有100%电压负反馈的放大器电路,其特点是输出电压的幅度和极性都与输入电压相同,所以叫跟随器。典型线路如图所示: 运放跟随器有输入阻抗高,而输出阻抗低的特性,一般来说,输入阻抗可以达到几兆欧姆,而输出阻抗低,通常只有几欧姆,甚至更低。 在电

Code for Coffee 5万+

Google Analytics接入:分析用户行为优化DDColor体验

通过集成Google Analytics,DDColor系统追踪用户在ComfyUI中的操作行为,分析工作流使用、参数调整与失败场景,驱动产品优化。数据揭示默认参数不合理、格式兼容等问题,实现从模型可用到体验好用的跨越。

weixin_42522857的博客 441

01|无代码开发 | 如何使用Zion实现导航网站制作

以上就是我制作导航网站的心得,分享给大家,作为一个小白想要独立搭建一个网站还是比较困难的,而在Zion不仅可以用可视化的方法进行搭建,甚至还可以一键发布,不用考虑服务器域名啥的问题,真是太方便了。

2301_76262110的博客 1390

Google Analytics 入门宝典:数据分析新手的必修课

文章摘要:Google Analytics 4(GA4)作为当前主流的数字营销分析工具,对于优化网站和应用性能至关重要。GA4 提供了深入的用户行为洞察,帮助营销人员做出数据驱动的决策,简化报告流程,并揭示数据背后的故事。文章详细介绍了GA4的设置步骤、关键报告和指标,以及如何避免常见错误,确保数据跟踪的准确性。通过正确使用GA4,企业可以最大化其数字营销活动的效果,保持市场竞争力。

SEO_juper的博客 889

wps js宏获取工作簿、工作表、单元格的值、链接转图片、单元格区域数据转二维数组、保存工作簿的常用Excel操作

wps js宏入门看这里,我写了常用的js操作Excel的方法:获取工作簿对象、获取工作表对象、获取单元格的值、链接转图片、生成条形码图片、单元格区域数据转二维数组方法、保存工作簿、弹出输入框、YesNo选择框,助你从VBA转战JS。

huaermeier的博客 3万+

跨境电商产品信息抓取:使用Python爬虫进行市场分析

跨境电商平台是指为不同国家和地区的买卖双方提供在线交易服务的电子商务平台。亚马逊(Amazon):全球最大的电商平台之一,提供丰富的商品种类和全球配送服务。eBay:全球知名的在线拍卖和购物网站,支持个人和企业卖家。速卖通(AliExpress):阿里巴巴旗下的跨境电商平台,主要面向全球消费者提供中国制造的商品。Wish:以移动端为主的跨境电商平台,主打低价商品和个性化推荐。这些平台上的商品数据包括商品名称、价格、销量、用户评价、库存等信息,对于市场分析具有重要价值。

2201_76125261的博客 1233

Google如何成功整合创业公司(转)

糟糕的并购在科技界俯拾即是,最著名的案例莫过于美国在线(AOL)和时代华纳的失败交易,类似的故事还包括惠普收购Palm、eBay收购Skype。为什么这些联姻会以悲剧收场?操盘者过于膨胀的野心、市场判断的失误、文化难以融合、被并购方人才流失,或者收购本身就是为了毁灭对方,每个失败故事背后的教训各异,后果却是相同:两败俱伤。 在此意义上,谷歌是个奇迹。从2001年起...

weixin_30783913的博客 254

C#串口通信革命:RJCP.DLL.SerialPortStream实战全解析

本文介绍了如何通过RJCP.DLL.SerialPortStream库实现高性能串口通信解决方案。相比.NET自带的SerialPort类,该库提供了完全的异步非阻塞IO、动态缓冲区控制和多线程事件驱动等优势,特别适合工业自动化和物联网场景。文章详细讲解了环境搭建、核心功能对比,并通过代码示例展示了端口配置、异步通信管道搭建等关键技术。最后以Modbus RTU协议为例,演示了高级协议封装方法,为开发者提供了完整的工业级串口通信实现方案。

墨夶的博客 1875

如何向管理层汇报营销成果:工程师必备的终极指南

Marketing-for-Engineers是一个精心策划的营销文章和工具集合,旨在帮助工程师解决实际营销任务,如寻找测试用户、增长初始用户群、零预算推广项目以及扩展营销活动以建立稳定的收入流。本文将为工程师提供一份简单实用的指南,帮助你向管理层清晰、有效地汇报营销成果。 Marketing for Engineers封面图 ## 为什么营销成果汇报很重要? 对于工程师来说,我们通常更擅长

gitblog_01025的博客 623

迈瑞BS系列全自动生化分析仪LIS协议接口手册(内含HL7和ASTM)

迈瑞BS系列全自动生化分析仪LIS协议接口手册,内含HL7和ASTM)

VC中sqlite的使用

在VC中对sqlite的使用,实现数据库大批数据的快速插入,读取数据库中的数据在ListCtrl中显示

团队协作编程平台与AI编程助手深度集成实战指南

团队协作编程平台是支撑现代软件工程的结构化工作流中枢,涵盖分支管理、代码审查、CI/CD和跨系统上下文集成;AI编程助手则依托大模型实现自然语言到可执行代码的语义转化。二者融合的本质,是将隐性协作规范(如PR模板、Jira字段、安全策略)转化为AI可感知、可内化、可反馈的显性规则。这种‘协作链路穿透力’决定了AI能否真正嵌入开发毛细血管——而非停留在补全行级代码的表层。本文聚焦真实团队场景,解析如何通过上下文锚点、Prompt Library和质量门禁,让AI读懂Wiki里的37条守则,并在订单分享、微服务

weixin_30315723的博客 331

深度学习工具包 DeepLearningToolbox

深度学习工具包Deprecation notice.-----This toolbox is outdated and no longer maintained.There are much better tools available for deep learning than this toolbox, e.g. [Theano](http://deeplearning.net/software/theano/), [torch](http://torch.ch/) or [tensorflow](http://www.tensorflow.org/)I would suggest you use one of the tools mentioned above rather than use this toolbox.Best, Rasmus.DeepLearnToolbox================A Matlab toolbox for Deep Learning.Deep Learning is a new subfield of machine learning that focuses on learning deep hierarchical models of data.It is inspired by the human brain's apparent deep (layered, hierarchical) architecture.A good overview of the theory of Deep Learning theory is[Learning Deep Architectures for AI](http://www.iro.umontreal.ca/~bengioy/papers/ftml_book.pdf)For a more informal introduction, see the following videos by Geoffrey Hinton and Andrew Ng.* [The Next Generation of Neural Networks](http://www.youtube.com/watch?v=AyzOUbkUf3M) (Hinton, 2007)* [Recent Developments in Deep Learning](http://www.youtube.com/watch?v=VdIURAu1-aU) (Hinton, 2010)* [Unsupervised Feature Learning and Deep Learning](http://www.youtube.com/watch?v=ZmNOAtZIgIk) (Ng, 2011)If you use this toolbox in your research please cite [Prediction as a candidate for learning deep hierarchical models of data](http://www2.imm.dtu.dk/pubdb/views/publication_details.php?id=6284)```@MASTERSTHESIS\{IMM2012-06284, author = "R. B. Palm", title = "Prediction as a candidate for learning deep hierarchical models of data", year = "2012",}```Contact: rasmusbergpalm at gmail dot comDirectories included in the toolbox-----------------------------------`NN/` - A library for Feedforward Backpropagation Neural Networks`CNN/` - A library for Convolutional Neural Networks`DBN/` - A library for Deep Belief Networks`SAE/` - A library for Stacked Auto-Encoders`CAE/` - A library for Convolutional Auto-Encoders`util/` - Utility functions used by the libraries`data/` - Data used by the examples`tests/` - unit tests to verify toolbox is workingFor references on each library check REFS.mdSetup-----1. Download.2. addpath(genpath('DeepLearnToolbox'));Example: Deep Belief Network---------------------```matlabfunction test_example_DBNload mnist_uint8;train_x = double(train_x) / 255;test_x = double(test_x) / 255;train_y = double(train_y);test_y = double(test_y);%% ex1 train a 100 hidden unit RBM and visualize its weightsrand('state',0)dbn.sizes = [100];opts.numepochs = 1;opts.batchsize = 100;opts.momentum = 0;opts.alpha = 1;dbn = dbnsetup(dbn, train_x, opts);dbn = dbntrain(dbn, train_x, opts);figure; visualize(dbn.rbm{1}.W'); % Visualize the RBM weights%% ex2 train a 100-100 hidden unit DBN and use its weights to initialize a NNrand('state',0)%train dbndbn.sizes = [100 100];opts.numepochs = 1;opts.batchsize = 100;opts.momentum = 0;opts.alpha = 1;dbn = dbnsetup(dbn, train_x, opts);dbn = dbntrain(dbn, train_x, opts);%unfold dbn to nnnn = dbnunfoldtonn(dbn, 10);nn.activation_function = 'sigm';%train nnopts.numepochs = 1;opts.batchsize = 100;nn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.10, 'Too big error');```Example: Stacked Auto-Encoders---------------------```matlabfunction test_example_SAEload mnist_uint8;train_x = double(train_x)/255;test_x = double(test_x)/255;train_y = double(train_y);test_y = double(test_y);%% ex1 train a 100 hidden unit SDAE and use it to initialize a FFNN% Setup and train a stacked denoising autoencoder (SDAE)rand('state',0)sae = saesetup([784 100]);sae.ae{1}.activation_function = 'sigm';sae.ae{1}.learningRate = 1;sae.ae{1}.inputZeroMaskedFraction = 0.5;opts.numepochs = 1;opts.batchsize = 100;sae = saetrain(sae, train_x, opts);visualize(sae.ae{1}.W{1}(:,2:end)')% Use the SDAE to initialize a FFNNnn = nnsetup([784 100 10]);nn.activation_function = 'sigm';nn.learningRate = 1;nn.W{1} = sae.ae{1}.W{1};% Train the FFNNopts.numepochs = 1;opts.batchsize = 100;nn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.16, 'Too big error');```Example: Convolutional Neural Nets---------------------```matlabfunction test_example_CNNload mnist_uint8;train_x = double(reshape(train_x',28,28,60000))/255;test_x = double(reshape(test_x',28,28,10000))/255;train_y = double(train_y');test_y = double(test_y');%% ex1 Train a 6c-2s-12c-2s Convolutional neural network %will run 1 epoch in about 200 second and get around 11% error. %With 100 epochs you'll get around 1.2% errorrand('state',0)cnn.layers = { struct('type', 'i') %input layer struct('type', 'c', 'outputmaps', 6, 'kernelsize', 5) %convolution layer struct('type', 's', 'scale', 2) %sub sampling layer struct('type', 'c', 'outputmaps', 12, 'kernelsize', 5) %convolution layer struct('type', 's', 'scale', 2) %subsampling layer};cnn = cnnsetup(cnn, train_x, train_y);opts.alpha = 1;opts.batchsize = 50;opts.numepochs = 1;cnn = cnntrain(cnn, train_x, train_y, opts);[er, bad] = cnntest(cnn, test_x, test_y);%plot mean squared errorfigure; plot(cnn.rL);assert(er<0.12, 'Too big error');```Example: Neural Networks---------------------```matlabfunction test_example_NNload mnist_uint8;train_x = double(train_x) / 255;test_x = double(test_x) / 255;train_y = double(train_y);test_y = double(test_y);% normalize[train_x, mu, sigma] = zscore(train_x);test_x = normalize(test_x, mu, sigma);%% ex1 vanilla neural netrand('state',0)nn = nnsetup([784 100 10]);opts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samples[nn, L] = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.08, 'Too big error');%% ex2 neural net with L2 weight decayrand('state',0)nn = nnsetup([784 100 10]);nn.weightPenaltyL2 = 1e-4; % L2 weight decayopts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex3 neural net with dropoutrand('state',0)nn = nnsetup([784 100 10]);nn.dropoutFraction = 0.5; % Dropout fraction opts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex4 neural net with sigmoid activation functionrand('state',0)nn = nnsetup([784 100 10]);nn.activation_function = 'sigm'; % Sigmoid activation functionnn.learningRate = 1; % Sigm require a lower learning rateopts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex5 plotting functionalityrand('state',0)nn = nnsetup([784 20 10]);opts.numepochs = 5; % Number of full sweeps through datann.output = 'softmax'; % use softmax outputopts.batchsize = 1000; % Take a mean gradient step over this many samplesopts.plot = 1; % enable plottingnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex6 neural net with sigmoid activation and plotting of validation and training error% split training data into training and validation datavx = train_x(1:10000,:);tx = train_x(10001:end,:);vy = train_y(1:10000,:);ty = train_y(10001:end,:);rand('state',0)nn = nnsetup([784 20 10]); nn.output = 'softmax'; % use softmax outputopts.numepochs = 5; % Number of full sweeps through dataopts.batchsize = 1000; % Take a mean gradient step over this many samplesopts.plot = 1; % enable plottingnn = nntrain(nn, tx, ty, opts, vx, vy); % nntrain takes validation set as last two arguments (optionally)[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');```[![Bitdeli Badge](https://d2weczhvl823v0.cloudfront.net/rasmusbergpalm/deeplearntoolbox/trend.png)](https://bitdeli.com/free "Bitdeli Badge")

从零到一:海外独立站搭建七步实战指南与避坑策略

在数字化商业时代,拥有一个自主可控的线上销售渠道是品牌出海和业务增长的关键。其核心原理在于摆脱对第三方平台的依赖,通过自建官网直接触达并沉淀用户,从而构建品牌资产和长期竞争力。从技术实现角度看,这通常涉及域名注册、主机选择、平台配置、支付物流集成等一系列工程实践。对于跨境电商卖家而言,独立站的价值在于掌握用户数据、实现营销自主和提升品牌溢价。其典型应用场景包括DTC品牌销售、垂直品类深耕以及多渠道业务拓展。本文聚焦于海外独立站搭建,详细拆解了从战略筹备到正式上线的七个核心步骤,并针对 **Shopify平台

weixin_33858485的博客 427

锂电池充电器电路设计 里面有电路

锂电池充电器电路设计 里面有电路 3.7V升压到5V电路设计。

2026年团队级AI编程助手选型指南:协作逻辑优先

AI编程助手已从个人效率工具演进为团队协作基础设施。其核心价值不在于代码生成速度,而在于能否理解并执行团队特有的架构规范、知识约定与协作流程。协作型AI的本质是构建可审计、可追溯、可验证的‘团队认知上下文’,通过私有化知识图谱、意图识别引擎与上下文注入管道,将Confluence文档、ADR决策、Jira需求、CI日志等异构信息转化为实时生效的编码约束。它支撑新人快速合规交付、保障跨职能开发一致性、降低架构漂移风险,并在微服务、多云、混合技术栈等复杂场景中成为事实上的‘第二大脑’。本文聚焦真实落地的8款20

cuanwei9356的博客 385

如何用RedditVideoMakerBot快速生成热门Reddit视频:新手完整指南

你是否曾经羡慕那些在TikTok、YouTube和Instagram上获得数百万观看量的Reddit故事视频?这些视频看似简单,但背后需要大量的编辑和素材整理工作。现在,有了RedditVideoMakerBot这个神奇工具,你只需一个命令就能自动化整个视频制作流程! RedditVideoMakerBot是一个开源项目,专门用于自动创建Reddit视频内容。它通过编程魔法,让你无需任何视频编辑

gitblog_00876的博客 422

企业社媒启动难?破解管理层‘不’字背后的三大组织断层

社交媒体运营在企业中常被误读为传播或营销工具,实则本质是组织信任机制与知识流转效率的显性化载体。其底层逻辑在于将隐性经验转化为可验证、可复用、可度量的业务信号——这涉及关系资产计量、工作流重构与员工能力基座建设三大技术命题。当管理者说‘没时间’‘没ROI’‘怕失控’,反映的并非抗拒新渠道,而是现有管理体系尚未适配数字时代的信息协同范式。本文聚焦制造业、医疗、工业品等B2B场景,结合17个真实转型案例,拆解如何用最小成本产出首个‘业务可感知’的内容切片,并建立穿透决策链、前置问题解决、驱动跨部门协同的信任指标

weixin_34384557的博客 478

谷歌SEO优化与独立站:外贸企业的双引擎战略与避坑指南

本文探讨了外贸企业在数字化出海时代如何有效结合谷歌SEO与独立站,构建高效的海外营销体系。文章指出,SEO是获取精准流量的核心渠道,而独立站是承载转化和建立信任的终极阵地,两者相辅相成。作者提出四大协同运营关键事项:战略定位统一、技术基建扎实、内容质量优先、数据驱动迭代,并给出从启动期到成熟期的实用工作流程建议,强调以用户需求为中心、坚持长期优化的重要性。文章认为,只有将SEO与独立站有机结合,外贸企业才能在激烈的国际竞争中建立稳固的线上优势。

2601_96764108的博客 833

10个终极数据可视化工具:图表和仪表板设计的最佳实践指南

在当今数据驱动的世界中,**数据可视化工具**已成为设计师和产品团队的必备技能。Awesome Design Tools项目汇集了众多优秀的**图表设计插件**和**数据可视化工具**,帮助设计师将复杂数据转化为直观易懂的视觉呈现。无论你是数据分析师、产品经理还是UI/UX设计师,掌握这些工具都能让你的数据故事更加生动有力。 ## 📊 为什么数据可视化如此重要? **数据可视化**不仅仅是美

gitblog_00112的博客 886
上一篇: 【转】32 Key Performance Indicators (KPIs) for Ecommerce
下一篇: 【转】扁平化设计原则
wayne_ren
博客等级 码龄21年 60粉丝 300原创
评论
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付:10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值