This document provides an overview of machine learning and artificial intelligence on Azure. It discusses the importance of AI and ML technologies and how organizations that leverage these technologies can gain a competitive advantage. It also describes different types of machine learning algorithms such as supervised learning, unsupervised learning, and deep learning. Finally, it discusses Microsoft's AI infrastructure and services for building, training, deploying, and consuming machine learning models on Azure.
10. The technology exists today.
Organizations that leverage these technologies
will be able to obtain or sustain a competitive
advantage.
Those that don't will be left behind.
11. Types of Machine Learning Algorithms
Algorithms
Supervised
Labeled Data
Classification
Categorical
Binary
2 categories
Multi-class
>2 categories
Regression
Numerical
Unsupervised
Unlabeled data
Clustering
Grouping
12.
13. Supervised Learning
• Using AI to determine an A -> B (input to response) mapping
• Have a machine “learn” these mappings
A B
Email Spam?
Image Object
Audio Text
Housing data Price
Ads, users Click?
Binary classification
Multiclass Classification
Multiclass classification
Regression
Binary classification
14.
15.
16. What is Deep Learning?
Input Layer
Hidden Layers
Output Layer
Data propagation direction
Error propagation direction
17. Types of deep learning architectures
• Recurrent Neural Network
• Long Short-Term Memory network
• Convolutional Neural Networks
• Recursive Neural Networks
• Deep Belief Networks
• Generative Adversarial Networks
18. When to use what?
Build your own or consume
pre-trained models?
Which Experience do you
want?
Deployment target
What engine(s) do you want
to use?
Microsoft ML & AI
products
Build your own Consume
Azure Machine Learning Cognitive services, bots
Code first Visual tooling
On-premise
ML Server
Cloud
AML Services (Preview)
Cloud
AML Studio
On-premise
Hadoop
SQL
Server
SQL
Server
Spark Hadoop Azure Batch DSVM Azure Container
Service
19. Workshop Build a Machine Learning Workflow and
deploy web service on AML Studio
22. Microsoft AI Infrastructure & Services
Best of Microsoft research
and open source
Comprehensive
ML and AI capabilities for
Data Scientists
Easy to consume
AI for developers
Hyperscale hardware innovation (scale, GPU, FPGA) – Azure Compute
Leading hardware, at scale,
on demand
MS AI Stack chart
*high level service are packaged
*raw infrastructure like GPU or DSVM (spin end series VM to run ML workloads)
*CNTK is DL library. Like tensorflow/caffe same family of tools
*AzureStudio
*APIs if you are not familiar with building from scratch