紫金大厅B(Grand Hall B)
Add ML ops and Kubeflow pipelines to your personal schedule
14:00 ML ops and Kubeflow pipelines Kaz Sato (Google)
Add Atom:Supremind云原生深度学习平台(Atom:A cloud native deep learning platform at Supremind) to your personal schedule
14:00 Atom:Supremind云原生深度学习平台(Atom:A cloud native deep learning platform at Supremind) Chaoguang Li (Qiniu), Bin Fan (Alluxio), Haoyuan Li (Alluxio)
Add 保险中的机器学习实践 to your personal schedule
16:20 保险中的机器学习实践 ju fang (中国人寿研发中心)
多功能厅2(Function Room 2)
Add Decentralized governance of data to your personal schedule
11:15 Decentralized governance of data Roger Chen (Computable)
Add Architecting AI applications to your personal schedule
13:10 Architecting AI applications Mikio Braun (Zalando)
Add AI pipelines on container platforms to your personal schedule
14:50 AI pipelines on container platforms WEIQIANG ZHUANG (IBM), Huaxin Gao (IBM)
多功能厅5A+B(Function Room 5A+B)
Add 自动驾驶技术是如何应用于新潮传媒、新零售行业 to your personal schedule
11:15 自动驾驶技术是如何应用于新潮传媒、新零售行业 LI YUAN (Perceptin 深圳普思英察科技有限公司)
Add 基于目标检测的智能化成矿异常信息提取 to your personal schedule
16:20 基于目标检测的智能化成矿异常信息提取 李苍柏 (中国地质科学院矿产资源研究所)
多功能厅6A+B (Function Room 6A+B)
Add 视频精彩度分析及智能创作 to your personal schedule
11:15 视频精彩度分析及智能创作 刘祁跃 (爱奇艺)
Add 在边缘实现深度学习 to your personal schedule
13:10 在边缘实现深度学习 陈玉荣 (Intel)
08:00 Morning Coffee | Room: 3楼序厅(3rd Floor Foyer)
Add 快速社交 (Speed Networking) to your personal schedule
08:00 快速社交 (Speed Networking) | Room: 3楼序厅(3rd Floor Foyer)
Add Friday opening remarks to your personal schedule
紫金大厅A(Grand Hall A)
08:45 Friday opening remarks Ben Lorica (O'Reilly Media), Roger Chen (Computable), Jason (Jinquan) Dai (Intel)
Add Top AI breakthroughs you need to know about to your personal schedule
08:50 Top AI breakthroughs you need to know about Abigail Hing Wen (Intel Corp.)
Add AI and retail to your personal schedule
09:30 AI and retail Mikio Braun (Zalando)
Add Closing remarks to your personal schedule
10:40 Closing remarks Ben Lorica (O'Reilly Media), Jason (Jinquan) Dai (Intel), Roger Chen (Computable)
10:45 Morning Break | Room: 报告厅序厅 (Auditorium Foyer)
15:30 Afternoon Break | Room: 报告厅序厅 (Auditorium Foyer)
Add 周五午餐主题桌会 (Friday Topic Tables at Lunch) to your personal schedule
11:55 Lunch 周五午餐主题桌会 (Friday Topic Tables at Lunch) | Room: 彩虹厅及国际厅 (Rainbow Room & Ballroom)
11:15-11:55 (40m) 英文讲话 (Presented in English)
Bringing research and production together with PyTorch 1.0
Joseph Spisak (Facebook)
Learn how PyTorch 1.0 enables you to take state-of-the-art research and deploy it quickly at scale in areas from autonomous vehicles to medical imaging. Joseph Spisak dives deep on the latest updates to the PyTorch framework including TorchScript and the JIT compiler, deployment support, and the C++ interface, and explains how Facebook uses PyTorch 1.0 to power AI across its products.
13:10-13:50 (40m) 人工智能对商业及社会的影响 (Impact of AI on Business and Society), 英文讲话 (Presented in English)
Artificial intelligence meets genomics: Accelerating understanding of our genetic makeup and the use of genome editing to revolutionize medicine
Yue Cathy Chang (TutumGene)
Genome editing has been dubbed a top technology that could create trillion-dollar markets. Learn how recent advancements in the application of AI to genomic editing are accelerating transformation of medicine with Yue Cathy Chang as she explores how AI is applied to genome sequencing and editing, the potential to correct mutations, and questions on using genome editing to optimize human health.
14:00-14:40 (40m) 实施人工智能 (Implementing AI), 模型与方法 (Models and Methods), 英文讲话 (Presented in English)
ML ops and Kubeflow pipelines
Kaz Sato (Google)
Kaz Sato explains how creating an ML model is just a starting point. To bring the technology into production service, you need to solve various real-world issues such as building a data pipeline for continuous training, automated validation of the model, version control of the model, scalable serving infra, and ongoing operation of the ML infra with monitoring and alerting.
14:50-15:30 (40m)
16:20-17:00 (40m)
11:15-11:55 (40m) 实施人工智能 (Implementing AI)
Analytics Zoo:基于Apache Spark的生产级别分布式TensorFlow(Analytics Zoo: Distributed TensorFlow in production on Apache Spark)
Yang Wang (Intel)
Building a model is fun and exciting; putting it to production is a different story. Yang Wang offers an overview of Analytics Zoo, a unified analytics and AI platform for distributed TensorFlow, Keras, and BigDL on Apache Spark, designed for production environments. See how you can benefit from its easy deployment, high performance, and efficient model serving for deep learning applications.
13:10-13:50 (40m) 实施人工智能 (Implementing AI)
Sparkling: 基于Apache Spark进行一站式机器学习
Yiheng Wang (Tencent)
机器学习项目在企业中实际落地往往涉及到复杂工作流构建和数据管理,以及多种工具的整合。而且随着数据规模的增加,团队规模的扩大,这一任务更具挑战性。Apache Spark是业界流行的大数据框架,被广泛的应用在海量数据的分析处理。本议题将介绍我们在腾讯云上如何基于Apache Spark为客户建立一个一站式机器学习平台的相关工作。主要内容包括多种数据源的接入,构建复杂数据管线,利用数据可视化理解数据,通过可插拔的机制使用各种流行的机器学习框架,以及部署和监控模型。我们也会分享在这一过程中遇到的问题和挑战。听众也可以了解到,通过这种和大数据紧密结合的一站式机器学习,用户可以怎样更加高效的建立和管理他们的机器学习项目,从而加速了机器学习在业务中的落地。
14:00-14:40 (40m) 实施人工智能 (Implementing AI)
Atom:Supremind云原生深度学习平台(Atom:A cloud native deep learning platform at Supremind)
Chaoguang Li (Qiniu), Bin Fan (Alluxio), Haoyuan Li (Alluxio)
The Atlab Lab at Qiniu Cloud focuses on deep learning for computer vision. Chaoguang Li, Haoyuan Li, and Bin Fan lead a deep dive into AVA, a high-performance and cost-effective cloud-based training platform for deep learning, which deeply integrates an open source software stack including TensorFlow, Caffe, Alluxio, and KODO, the company's own cloud object storage.
14:50-15:30 (40m) 实施人工智能 (Implementing AI)
查询地球:Uber的地理空间大数据分析(Query the planet: Geospatial big data analytics at Uber)
Zhenxiao Luo (Twitter)
Locations and trips provide insights that can improve business decisions and better serve users, but geospatial data analysis is particularly challenging. It requires efficiency, usability, and scalability in order to meet user needs and business requirements. Join Zhenxiao Luo to learn how Uber uses artificial intelligence to analyze geospatial big data, one of its distinct challenges.
16:20-17:00 (40m) 实施人工智能 (Implementing AI)
ju fang (中国人寿研发中心)
11:15-11:55 (40m) 英文讲话 (Presented in English)
Decentralized governance of data
Roger Chen (Computable)
Roger Chen details how to enable powerful data lineage properties with decentralized data governance models using blockchain technology. As a result, organizations can easily satisfy growing compliance regulations around data privacy while gaining access to rich external data resources for building AI models.
13:10-13:50 (40m) 实施人工智能 (Implementing AI), 英文讲话 (Presented in English)
Architecting AI applications
Mikio Braun (Zalando)
Mikio Braun takes you through the past 20 years of machine learning research to explore aspects of artificial intelligence, then examines current examples like autonomous cars and chatbots. Together you'll put together a mental model for a reference architecture for artificial intelligence systems.
14:00-14:40 (40m) 与人工智能交互 (Interacting with AI), 实施人工智能 (Implementing AI)
乐天构建数据科学平台最佳实践(Best practices for building a data science platform at Rakuten)
Orchlon Ann (Rakuten), TzuLin Chin (Rakuten)
Orchlon Ann and TzuLin Chin offer an overview of the Data Science Platform, a suite of tools for exploring data, training models, and running GPU/CPU compute jobs in an isolated container environment. Discover it's benefits, including one-click machine learning environment creation, a powerful job scheduler, and a flexible function-as-a-service component.
14:50-15:30 (40m) 企业人工智能 (AI in the Enterprise), 实施人工智能 (Implementing AI), 英文讲话 (Presented in English)
AI pipelines on container platforms
AI pipelines simplify the lifecycle workflow management and enhance reproducibility and collaboration for machine learning and deep learning projects. Cloud native platform solutions offer great portability and scalability. Weiqiang Zhuang and Huaxin Gao show how, by combining strengths, AI pipelines on container platforms can help accelerate AI application development and deployment.
16:20-17:00 (40m) 企业人工智能 (AI in the Enterprise), 实施人工智能 (Implementing AI), 模型与方法 (Models and Methods), 英文讲话 (Presented in English)
Best practices for building enterprise-grade recommendation systems on Azure with Microsoft/Recommenders
Le Zhang (Microsoft), Jianxun Lian (Microsoft)
Enterprises benefit from recommendation systems for revenue and customer engagement, but creating such a system is time-consuming. Le Zhang and Jianxun Lian explore the Microsoft/Recommenders repository, which offers solutions to building recommendation systems. It contains classic and state-of-the-art algorithms from Microsoft and enables enterprise success by leveraging Azure's cloud capability.
11:15-11:55 (40m) 实施人工智能 (Implementing AI)
LI YUAN (Perceptin 深圳普思英察科技有限公司)
13:10-13:50 (40m) 案例研究 (Case Studies)
中国电信如何利用对抗性自动编码器来对抗金融诈骗(How China Telecom combats financial fraud with adversarial autoencoders)
Weisheng Xie (Orange Financial)
Weisheng Xie dives deep into how China Telecom uses adversarial autoencoders (AAEs) for risk factors modeling to fight a special kind of financial fraud. It's just one step in a long path of unsupervised tasks, but it's proved to be efficient and effective in practice.
14:00-14:40 (40m) 与人工智能交互 (Interacting with AI), 实施人工智能 (Implementing AI)
王书浩 (透彻影像)
14:50-15:30 (40m) 人工智能对商业及社会的影响 (Impact of AI on Business and Society)
Hongyu Cui (DataVisor)
AI技术在赋能各个产业的同时,也被网络黑产所利用,使得黑产攻击更加自动化,更加隐蔽,难于检测。 DataVisor在互联网反欺诈领域研究发现,目前黑产的攻击模型呈现以下趋势:攻击方法多样化而变化快,攻击手段趋于模拟正常用户,攻击账号主要来源由大规模注册渐渐转向ATO账号。传统的规则系统和有监督的模型,由于对欺诈案例以及标签数据的强依赖,往往无法及时应对迅速演化的黑产攻击,在反欺诈中一直处于被动防守的状态。DataVisor的无监督算法,通过全局分析,在高维空间聚类,可以在无标签情况下,自动发现大规模关联欺诈团伙。无监督算法在提前预警以及检测快速演变欺诈模式方面体现了显著的优势。
16:20-17:00 (40m) 实施人工智能 (Implementing AI)
李苍柏 (中国地质科学院矿产资源研究所)
11:15-11:55 (40m) 实施人工智能 (Implementing AI)
刘祁跃 (爱奇艺)
对视频进行精彩度分析,有助于筛选优质内容,尤其是冷启动阶段 同时,基于算法对精彩内容的理解,可以辅助创作,如进行标题辅助生成、动态/精彩封面生成、智能拆条等 我们通过对视频、音频、文本等多模态内容分析,同时利用用户交互数据,建立了完备的视频精彩度分析系统,并落地在长/短视频的不同业务场景下,明显提升了业务产出质量和效率
13:10-13:50 (40m) 模型与方法 (Models and Methods)
陈玉荣 (Intel)
14:00-14:40 (40m) 模型与方法 (Models and Methods)
基于知识图谱的可解释性推荐系统(Explainable reasoning over knowledge graphs for recommendation)
Dingxian Wang (eBay)
In recent years, there's been increasing attention on incorporating knowledge graphs into recommender systems. By exploring the interlinks within a knowledge graph, you can discover the connectivity between users and items as paths. Dingxian Wang outlines a new model, knowledge-aware path recurrent network (KPRN), for exploiting knowledge graphs for recommendation.
14:50-15:30 (40m) 与人工智能交互 (Interacting with AI), 实施人工智能 (Implementing AI)
PAI张量加速器和优化器:又一个深度学习编译器(PAI tensor accelerator and optimizer: Yet another deep learning compiler)
杨军 (阿里巴巴), 龙国平 (Alibaba)
本次演讲会介绍阿里计算平台PAI团队过去一年多时间里在深度学习编译器领域的技术工作进展----PAI TAO(Tensor Accelerator and Optimizer)。PAI-TAO采用通用编译优化技术,来解决PAI平台所承载的多样性AI workload面临的训练及推理需求的性能优化问题,在部分workload上获得了20%到4X不等的显著加速效果,并且基本作到用户层全透明,在显著提升平台效率性能的同时也有效照顾了用户的使用惯性。目前PAI-TAO已经先后用于支持阿里内部搜索、推荐、图像、文本等多个业务场景的日常训练及推理需求。
16:20-17:00 (40m) 实施人工智能 (Implementing AI)
Intel架构的低精度推断(Low-precision inference on Intel architecture)
Lei Xia (Intel)
Vector neural network instructions (VNNI) is the new Intel instruction set for low-precision AI inference inside the next-generation Xeon platform. Lei Xia offers an overview of the VNNI and Intel software tools, helping you use this new instruction set to accelerate inference with INT8.
08:00-08:45 (45m)
Break: Morning Coffee
08:00-08:30 (30m)
快速社交 (Speed Networking)
08:45-08:50 (5m)
Friday opening remarks
Ben Lorica (O'Reilly Media), Roger Chen (Computable), Jason (Jinquan) Dai (Intel)
Program chairs Ben Lorica, Jason Dai, and Roger Chen open the second day of keynotes.
08:50-09:00 (10m) 英文讲话 (Presented in English)
Top AI breakthroughs you need to know about
Abigail Hing Wen (Intel Corp.)
Abigail Hing Wen catches you up on some of the most exciting recent breakthroughs in the industry, including natural language processing strong enough to generate sentences indistinguishable from a human’s, highly accurate 3D protein structure prediction to fight disease, and leaps forward in reinforcement learning, a more natural but very complex alternative to other forms of machine learning.
09:00-09:15 (15m) 英文讲话 (Presented in English)
Data orchestration for AI, big data, and the cloud
Haoyuan Li (Alluxio)
Haoyuan Li offers an overview of a data orchestration layer that provides a unified data access and caching layer for single cloud, hybrid, and multicloud deployments. It enables distributed compute engines like Presto, TensorFlow, and PyTorch to transparently access data from various storage systems while actively leveraging an in-memory cache to accelerate data access.
09:15-09:30 (15m)
自驾驶技术与未来自动化车辆仓到仓运输(Self-driving technology and the future autonomous depot-to-depot transport)
Hao Zheng (PlusAI)
PlusAI is developing a full stack self-driving technology to enable large-scale autonomous commercial fleets. Hao Zheng examines some of the unique challenges across different layers of the technology stack of building an autonomous truck that's both safe and efficient and dives into how PlusAI is addressing them.
09:30-09:45 (15m) 英文讲话 (Presented in English)
AI and retail
Mikio Braun (Zalando)
What do your customers want? What are the current and upcoming trends? Mikio Braun takes a look at Zalando and the retail industry to explore how AI is redefining the way ecommerce sites interact with customers to create a personalized experience that strives to make sure customers find what they want when they need it.
09:45-10:00 (15m) 英文讲话 (Presented in English)
为什么说人工智能和云计算乃天作之合?(Why do we say AI Should be Cloud Native?)
Yangqing Jia (Alibaba Group)
The recent years of AI has grown out of labs and created a transformative power for a vast range of industries. But, while we take it for granted that AI and Cloud come hand in hand, I'll show you an argument one step further: AI should be Cloud Native.
10:00-10:20 (20m) 英文讲话 (Presented in English)
Designing computer hardware for artificial intelligence
Michael James (Cerebras)
Artificial intelligence is defining a new generation of computer technology with applications that blur the boundaries between intuition, art, and science. Michael James examines the fundamental drivers of computer technology, surveys the landscape of AI hardware solutions, and explores the limits of what's possible as new computer platforms emerge.
10:20-10:40 (20m) 英文讲话 (Presented in English)
Toward learned algorithms, data structures, and systems
Tim Kraska (MIT)
Systems and applications are composed from basic data structures and algorithms. Most of these have been around since the beginnings of CS and form every intro lecture. Yet, we might soon face an inflection point. Tim Kraska outlines different ways to build learned algorithms and data structures to achieve instance optimality and unprecedented performance for a wide range of applications.
10:40-10:45 (5m)
Closing remarks
Ben Lorica (O'Reilly Media), Jason (Jinquan) Dai (Intel), Roger Chen (Computable)
Program chairs Ben Lorica, Jason (Jinquan) Dai, and Roger Chen close the second day of keynotes.
10:45-11:15 (30m)
Break: Morning Break
15:30-16:20 (50m)
Break: Afternoon Break
11:55-13:10 (1h 15m)
周五午餐主题桌会 (Friday Topic Tables at Lunch)