Week 12: Cloud AI- DSA 441 Cloud Computing
0 likes260 views
Cloud AI is a machine learning platform used to create and deploy machine learning models in the cloud. It allows users to streamline workflows including data preparation, model training, evaluation and publishing models for use in other cloud services. The platform provides services like Machine Learning Studio for visual modeling, Data Science Workshop for interactive development, and Elastic Algorithm Service to deploy models as APIs.
Week 12: Cloud AI- DSA 441 Cloud Computing
- 1. DSA 441 – Cloud Computing Week 12: Cloud AI Asst. Prof. Dr. Ferdin Joe John Joseph Faculty of Information Technology Thai-Nichi Institute of Technology, Bangkok
- 2. Cloud AI (Machine Learning Platform for AI)
- 3. Cloud AI • Machine Learning Platform for AI • Used for creating machine learning models in cloud • Deployed models are used capable of using in other cloud products and services Faculty of Information Technology, Thai-Nichi Institute of Technology 3
- 4. Machine Learning Machine learning uses statistical algorithms to train models with a large amount of historical data, and uses the generated models to help you make informed business decisions. Machine learning can be applied to the following scenarios: • Marketing: commodity recommendation, user profiling, and targeted advertising. • Finance: credit risk prediction for loans, financial risk management, stock forecast, and gold price forecast. • Social network: analytics of key opinion leaders and relational networks. • Text processing: news classification, keyword extraction, text summarization, and text analytics. • Unstructured data processing: image classification and text extraction based on optical character recognition (OCR). • Other forecast scenarios: rainfall forecast and football match result forecast. Faculty of Information Technology, Thai-Nichi Institute of Technology 4
- 5. Machine Learning Types Machine learning includes traditional machine learning and deep learning. Traditional machine learning is divided into the following learning modes: • Supervised learning: Each sample has an expected value. You can create a model to map input feature vectors to target values. Supervised learning can be used to solve regression and classification issues. • Unsupervised learning: Samples do not have target values. Unsupervised learning is used to discover potential regular patterns from the sample data. You can use unsupervised learning to solve clustering issues. • Reinforcement learning: This learning mode is complex. A system constantly interacts with the external environment to obtain feedback and determines its own behavior to achieve a long-term optimization of targets. Examples of reinforcement learning are AlphaGo and autonomous driving. Faculty of Information Technology, Thai-Nichi Institute of Technology 5
- 6. Why Machine Learning Platform for AI • Machine Learning Platform for AI is designed to serve business within Alibaba Group, such as Taobao, Alipay, and Amap.com. • It enables developers of Alibaba Group to use AI technologies in an efficient, concise, and standard way. Machine Learning Platform for AI was officially released in 2018. • It has gained tens of thousands of enterprises and individual developers, and has become one of the leading machine learning platforms on the cloud in China. Faculty of Information Technology, Thai-Nichi Institute of Technology 6
- 7. Machine Learning Frameworks • Flink, a stream computing framework. • TensorFlow, an optimized deep learning framework based on open source TensorFlow. • Parameter Server, a computing framework that can process hundreds of billions of samples in parallel. • Spark, PySpark, MapReduce, and other mainstream open source computing frameworks. Faculty of Information Technology, Thai-Nichi Institute of Technology 7
- 8. Services • Machine Learning Studio: a service for visualized modeling and distributed training. • Data Science Workshop (DSW): a Notebook-based service for interactive AI research and development. • AutoLearning: a service for automated modeling. • Elastic Algorithm Service (EAS): a service that allows you deploy models as online prediction services. Faculty of Information Technology, Thai-Nichi Institute of Technology 8
- 9. Benefits • Services provided by Machine Learning Platform for AI can be used separately or in combination. Machine Learning Platform for AI provides an all-in-one platform for machine learning. After training data is prepared in Object Storage Service (OSS) or MaxCompute, you can use Machine Learning Platform for AI to streamline all workflows, including data uploading, data preprocessing, feature engineering, model training, model evaluation, and model publishing (to both online and offline environments). • Machine Learning Platform for AI can be integrated with DataWorks and allows you to process data by using Structured Query Language (SQL), user-defined functions (UDFs), user-defined aggregation functions (UDAFs), and MapReduce. This ensures higher flexibility and efficiency. • Experiments that are used to train and generate models can be scheduled in DataWorks. You can run scheduled tasks in the staging or production environment. This enables data isolation. Faculty of Information Technology, Thai-Nichi Institute of Technology 9
- 10. Architecture Faculty of Information Technology, Thai-Nichi Institute of Technology 10
- 11. Architecture • Infrastructure layer: includes CPU, GPU, Field Programmable Gate Array (FPGA), and Neural network Processing Unit (NPU) resources. • Computing engines and container services layer: includes MaxCompute, E-MapReduce (EMR), Realtime Compute, and Alibaba Cloud Container Service for Kubernete (ACK). • Computing framework layer: includes Alink, TensorFlow, PyTorch, Caffe, MapReduce, SQL, and Message Passing Interface (MPI). You can run distributed computing tasks in these frameworks. Faculty of Information Technology, Thai-Nichi Institute of Technology 11
- 12. Architecture (cont’d) Machine Learning Platform for AI streamlines the workflows of machine learning, including data preparation, model creation and training, and model deployment. • Data preparation: Smart labeling of Machine Learning Platform for AI allows you to label data and manage datasets in multiple scenarios. • Model creation and training: Machine Learning Platform for AI provides diverse services to meet different modeling requirements. These services are Machine Learning Studio, Data Science Workshop (DSW), Deep Learning Containers (DLC), and AutoLearning. Machine Learning Studio is a service for visualized modeling. DSW allows you to create models by interactive programming. DLC is a cloud-native platform for training deep learning models. AutoLearning is a service for end-to-end automated model creation. • Model deployment: Machine Learning Platform for AI provides Elastic Algorithm Service (EAS) and Blade to help you deploy models as services. EAS is a cloud-native online inference platform and Blade is a tool used to accelerate model inference. Machine Learning Platform for AI also provides an intelligent marketplace where you can obtain recommended solutions and model algorithms to solve business issues and improve production efficiency. Faculty of Information Technology, Thai-Nichi Institute of Technology 12
- 13. Architecture (cont’d) • Business layer: Machine Learning Platform for AI is widely used in finance, medical care, education, transportation, and security sectors. Search systems, recommendation systems, and financial service systems of Alibaba Group all use Machine Learning Platform for AI to explore data values for making informed business decisions. Faculty of Information Technology, Thai-Nichi Institute of Technology 13
- 14. Gartner statistics Faculty of Information Technology, Thai-Nichi Institute of Technology 14
- 15. Benefits Rich machine learning algorithms • The algorithms of Machine Learning Platform for AI have been tested by business within Alibaba Group for many years. Machine Learning Platform for AI supports basic algorithms such as clustering and regression, and complex algorithms such as text analysis and feature engineering. Compatibility with Alibaba Cloud services • Models trained by Machine Learning Platform for AI are stored in MaxCompute and can be used with other Alibaba Cloud services. Faculty of Information Technology, Thai-Nichi Institute of Technology 15
- 16. Benefits All-in-one machine learning platform • Machine Learning Platform for AI allows you to streamline workflows including data uploading, data preprocessing, feature engineering, model training, and model evaluation and publishing. Mainstream deep learning frameworks • Machine Learning Platform for AI supports mainstream deep learning frameworks such as TensorFlow, Caffe, and MXNet. Faculty of Information Technology, Thai-Nichi Institute of Technology 16
- 17. Benefits Visualized modeling methods • Machine Learning Platform for AI is developed with classic machine learning algorithms. These algorithms provide the following benefits: • You can drag and drop components to create machine learning experiments. • You can use the built-in Automated Machine Learning (AutoML) module to tune parameters. AutoML can automatically explore model parameters, evaluate models, pass down generated models to downstream nodes, and optimize models. Faculty of Information Technology, Thai-Nichi Institute of Technology 17
- 18. Benefits Quick model deployment • Machine Learning Platform for AI allows you to deploy models trained by Machine Learning Studio, Data Science Workshop (DSW), and AutoLearning as Restful APIs. • This way, you can use models by calling their APIs to meet your requirements. Faculty of Information Technology, Thai-Nichi Institute of Technology 18
- 19. Machine Learning Studio Faculty of Information Technology, Thai-Nichi Institute of Technology 19
- 20. DSW Faculty of Information Technology, Thai-Nichi Institute of Technology 20
- 21. EAS Faculty of Information Technology, Thai-Nichi Institute of Technology 21
- 22. DSW Demonstration Faculty of Information Technology, Thai-Nichi Institute of Technology 22
- 23. Next Week • Capstone Project Preparation Faculty of Information Technology, Thai-Nichi Institute of Technology 23