Microsoft Ignite | The Tour Hong Kong将于2019年2月20日至21日来到香港。这是一项面向全体开发人员和技术专业人士的免费技术培训活动,它将把Microsoft Ignite中的最优质内容直接传递给全球的技术从业者。在此,您可以充分利用微软的云工具和技术,拓展现有技能、搭建新的平台。不要错过这个活动,期待与您相遇!
点击链接,尽快注册:https://www.microsoft.com/en-hk/ignite-the-tour/hong-kong
Explore the latest cloud technologies and learn how to put your skills to work in new areas. Whether you’re developing innovative apps or delivering optimized solutions, these two info-packed days will help you evolve your skills, deepen your expertise, and prepare you to face new challenges.
Build your best conference experience with custom learning paths – you pick the path, we’ll guide you to the modules that meet your learning goals. Each path progresses along five expert-led modules that explore their topics in depth. Plus, you can customize your learning even further – pick up new skills with additional sessions that complement your path. There’s something for everyone, so check out the learning paths on our website here <insert tracking link to city page on website>.
• Azure fundamentals
• Building and maintaining your Azure hybrid environment
• Building your applications for the cloud
• Deploying your modern desktop
• Developing Microsoft 365 applications and integrations
• Getting the most out of your data
• Migrating applications to the cloud
• Operating applications and infrastructure in the cloud
• Optimizing teamwork in your organization
• Securing your organization
And there’s even more:
• Explore the future of cloud development, data, IT, and business intelligence at more than 100+ deep-dive sessions and hands-on workshops.
• Go one-on-one with Microsoft engineers and get answers to your toughest questions.
• Immerse yourself in our vibrant community at networking events, product theaters, attendee hangouts, user groups, meetups, and more.
Don’t miss out, reserve your free seat today. We hope to see you there!
点击链接,尽快注册:https://www.microsoft.com/en-hk/ignite-the-tour/hong-kong
好文章,需要你的鼓励
Fractal AI Research实验室开发了Fathom-DeepResearch智能搜索系统,该系统由两个4B参数模型组成,能够进行20多轮深度网络搜索并生成结构化报告。研究团队创新了DUETQA数据集、RAPO训练方法和认知行为奖励机制,解决了AI搜索中的浅层化、重复性和缺乏综合能力等问题,在多项基准测试中显著超越现有开源系统,为AI助手向专业研究工具转变奠定了基础。
快手科技与清华大学合作发现当前AI语言模型训练中存在严重的权重分配不平衡问题,提出了非对称重要性采样策略优化(ASPO)方法。该方法通过翻转正面样本的重要性权重,让模型把更多注意力放在需要改进的部分而非已经表现良好的部分,显著提升了数学推理和编程任务的性能,并改善了训练稳定性。