Machine Learning in Microsoft Ecosystem

A presentation describing different ways to do Machine Learning in Microsoft EcoSystem.

Transcript

1. Machine Learning in Microsoft Ecosystem BY MOHAMMED OSMAN @COGNITIVEOSMAN MOHAMMED.OSMAN@SMARTERCODE.IO
2. About Me .NET/Azure developer and Data Scientist Running a blog with a message “Making your code smart and your career smarter” ◦https://www.smartercode.io/blog Pluralsight Author Do not miss the Email course ◦ https://www.smartercode.io/email-course-survey PUBG Lover!
3. Agenda ➢First things first: Understanding the Machine Learning Pipeline ➢Method 1: Machine Learning using Azure AI Cognitive Services ➢Method 2: Machine Learning using ML.NET ➢Method 3: Machine Learning using Azure ML ➢3 Key Take Aways
4. Machine learning is the study of computer algorithms that improve automatically through experience Wikipedia
5. Applications of ML Sales Forecasting Customer Segmentation Object Detection Spam Detection
6. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring Operationalization
7. For more details https://www.smartercode.io /building-machine-learning- solution
8. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring We define our business problem ➔ Helps to choose the right data and ML method
9. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring Collect data from different sources Prepare the data Separate train and test
10. For more details on data preparation https://www.smartercode.i o/exploratory-data- analysis-aws-machine- learning
11. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring Spot-Checking the right ML model Evaluating the model based on the problem type
12. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring Taking our models to production for usage Monitoring model performance
13. It is still not the end!  We have to make sure that our models continue doing well even after deployment!
14. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring Reiterating to make sure that the ML model adapts to the new environment
16. Azure Cognitive Services DOING AI THROUGH RESTFUL CALLS
17. Machine Learning Pipeline (Simplified View) Problem Definition Data Sourcing Data Preparation Data Segregation Model Training Model Evaluation Model Deployment Model Monitoring You only need to mind deciding on the problem definition and how to source the data Analytics Data
18. What is that black boxed API? Application AI Service Data Analytics Image Recognition Sentiment Analysis Forms Analysis Transcription
20. Interested to know more? https://www.smartercode.io/AI-as-Service
21. ML.NET PYTHON IN C#?
22. What is ML.NET? ➢Open source and cross machine learning framework ➢Important aspects ◦ Built for .NET developers ◦ Custom ML made easy with tools ◦ Extended with TensorFlow and more ◦ Trusted, proven and scale ➢Supports AutoML with CLI and tool (no ML knowledge required) ➢ML Jupyer Notebooks
23. Problems you can solve with ML.NET Source: https://dotnet.microsoft.com/apps/machinelearning-ai/ml-dotnet
24. How to use ➢ML.NET API ➢Model Builder Tool oInstallation Guide: https://dotnet.microsoft.com/learn/ml-dotnet/get- started-tutorial/install ➢ML.NET CLI
25. https://github.com/dotnet/machinelearning-samples/tree/master/samples/csharp/getting- started/Regression_TaxiFarePrediction
26. Source: https://dotnet.microsoft.com/apps/machinelearning-ai/ml-dotnet
27. Azure ML ML PIPELINE IN CLOUD
28. What is Azure ML ➢Azure Based Machine Learning service ➢Full “MLOps” From Problem Definition to Operationalization ➢Enjoys Power of Cloud ➢Suitable for large scale data
29. Azure ML Notebooks
30. Azure ML Designer Source: https://bit.ly/3hvNO7j
31. Azure Automated ML Source: https://bit.ly/2UQ9xgp
32. When to Choose What? My data should be on premise ML.NET Azure Cognitive Services on Containers
33. When to Choose What? I do not have much ML knowledge ML.NET Model Builder Azure Cognitive Services Azure ML model designer
34. When to Choose What? No Programming Knowledge Azure ML designer
35. When to Choose What? I want to customize my models ML.NET Azure ML
36. When to Choose What? My data should be on premise ML.NET Azure Cognitive Services on Containers I do not have much ML knowledge ML.NET Model Builder Azure Cognitive Services Azure ML Auto ML No Programming Knowledge Azure ML designer I want to customize my models ML.NET Azure ML
37. Key Take Aways ➢Machine Learning requires many laborious steps if done by fundamental principles ➢Use the right framework/tool for your specific situation ➢Think about Azure ML if you need to keep close eye on model performance in prod
38. Thank you very much! You can reach me at mohammed.osman@smartercode.io and @cognitiveosman