×
Community Blog Redefining Tech with Machine Learning and AI – Part 2

Redefining Tech with Machine Learning and AI – Part 2

Part 2 of this 2-part article introduces the Alibaba Cloud Machine Learning Platform for AI (PAI).

By Shantanu Kaushik

In Part 1, we introduced machine learning and the core methodology to achieve optimal results to facilitate learning. This article introduces the Alibaba Cloud Machine Learning Platform for AI (PAI) and the architecture of this solution.

The Machine Learning Platform for AI (PAI) has more usage and implementation scenarios than other available solutions. Let’s use an example of an organization with a strongly defined DevOps pipeline. The DevOps solution is all about automation and reaching the threshold where application delivery is a highly self-sustaining process.

This process includes the following steps:

  • Core or first build
  • Testing
  • Debugging
  • Testing Again
  • Network Optimization
  • Content Delivery
  • Release
  • Metrics and Feedback Collection
  • Integration
  • Security Practices
  • Future Strategy Building

This application lifecycle is highly regulated with machine learning. Based on different phases, the Machine Learning Platform with AI works with:

The Alibaba Cloud Solution

Alibaba Cloud Machine Learning uses statistic algorithms to create training models based on huge amounts of historical data. These models are applicable in a number of scenarios and can help make informed business decisions. Alibaba Cloud Machine Learning includes traditional machine learning and deep learning.

Alibaba Cloud Machine Learning Platform for AI was designed to serve enterprises to attain an optimal working cycle and a structure that incorporates the core business values and builds on the historical data. Alibaba Cloud launched the Machine Learning Platform for AI in 2018. Since then, it has worked with thousands of enterprises and individual developers to become one of the leading machine learning and AI solutions available.

Alibaba Cloud Machine Learning Platform for AI supports computing frameworks, such as:

  • Flink: A stream computing framework
  • TensorFlow: An open-source and optimized deep learning framework
  • Parameter Server can process hundreds of billions of samples parallelly
  • Spark
  • PySpark
  • MapReduce

Benefits

  1. Regression
  2. Classification
  3. Clustering
  4. Text Analysis
  5. Finance Analysis
  6. Strategy Modeling
  1. Data Uploading
  2. Data Preprocessing
  3. Feature Engineering
  4. Model Training
  5. Model Evaluation
  6. Model Publishing
  • Machine Learning Platform for AI allows you to create a complete workflow for enterprise-level machine learning data modeling and application.
  • The Machine Learning Platform for AI infrastructure relies on Alibaba Cloud distributed computing clusters, enabling it to handle a large number of concurrent algorithm computing tasks.
  • Machine Learning Platform for AI can be integrated with DataWorks.
  • You can process data to provide higher flexibility and efficiency using:
  1. Structured Query Language (SQL)
  2. User-Defined Functions (UDFs)
  3. User-Defined Aggregation Functions (UDAFs)
  4. MapReduce
  • Alibaba Cloud Machine Learning Platform for AI provides statistics components, machine learning components, and network and data analysis components.
  • A superb practice of data isolation by running scheduled tasks (experiments and modeling) in pre-production and production environments

Architecture

Alibaba Cloud Machine Learning Platform for AI uses multiple layers to distribute workflows:

1

1.  The infrastructure layer includes:

 a. CPU and GPU

 b. Field Programmable Gate Array (FPGA)

 c. Neural Network Processing Unit (NPU) resources

2.  The computing engines and container services layer include:

 a. MaxCompute

 b. E-MapReduce (EMR)

 c. Realtime Compute

 d. Alibaba Cloud Container Service for Kubernetes (ACK)

3.  The computing framework layer includes:

 a. Alink

 b. TensorFlow

 c. PyTorch

 d. Caffe

 e. MapReduce

 f. SQL

 g. Message Passing Interface (MPI)

4.  The presentation layer of the Alibaba Cloud Machine Learning Platform for AI streamlines the workflows of machine learning by:

 a. Data Preparation – Smart labeling enables you to label data and manage datasets in multiple scenarios.

 b. Model Creation and Training – Machine Learning Platform for AI provides diverse services to meet different modeling requirements, including:

 i. Machine Learning Studio provides visualized modeling and distributed training.

 ii. Data Science Workshop (DSW) is a notebook-based service for interactive AI research and development.

 iii. AutoLearning provides automated modeling.

 iv. Elastic Algorithm Service (EAS) can deploy models as online prediction services.

 c. Model DeploymentMachine Learning Platform for AI provides models as services:

 i. Elastic Algorithm Service (EAS) is a cloud-native online inference platform

 ii. Blade is a tool to accelerate model inference.

 iii. There is an intelligent marketplace where you can obtain recommended solutions and model algorithms to solve business issues and improve production efficiency.

5.  The business layer of Machine Learning Platform for AI is widely used in the following scenarios:

 a. Finance

 b. Medical Care

 c. Academics and Education

 d. Transportation

 e. Security Sectors

 f. Search Systems

 g. Recommendation Systems (Decision Engines)

Wrapping Up

Alibaba Cloud Machine Learning Platform for AI is a highly intuitive solution intended for enterprises and developers. It is a lightweight machine learning platform created using cloud-native technologies.

It provides an end-to-end modeling service that includes the highly efficient Data Science Workshop (DSW). DSW works with interactive modeling, Machine Learning Studio for a no-code visualized modeling experience, Deep Learning Containers (DLC) that makes a superb choice for distributed model training, and Elastic Algorithm Service (EAS) for online prediction.

Upcoming Articles

  1. Seeding Digital Transformation in Southeast Asia – Part 1: What Is Digital Transformation?
  2. Seeding Digital Transformation in Southeast Asia – Part 2: Challenges and Focus Areas
  3. Seeding Digital Transformation in Southeast Asia – Part 3: An Introduction to Buttercup Spices
  4. Seeding Digital Transformation in Southeast Asia – Part 4: The Strategy
  5. Seeding Digital Transformation in Southeast Asia – Part 5: Application Migration
  6. Seeding Digital Transformation in Southeast Asia – Part 6: Resource Planning and Networking
  7. Seeding Digital Transformation in Southeast Asia – Part 7: Building a Smarter Supply Chain
  8. Seeding Digital Transformation in Southeast Asia – Part 8: Introducing AI for Better Workflow
0 0 0
Share on

Alibaba Clouder

2,599 posts | 758 followers

You may also like

Comments