This is Part Two in a series – you can read Part One here:
New features are introduced to Azure Machine Learning
Azure machine learning is probably the most popular service in the Azure service stack. So we will start our article with some new features added to it.
Model interpretability is probably the best feature added in MS Build 2019. We build models from a given data set during the development life cycle. The purpose of a model is to simulate a system or predict the behavior of a system. The ability to define a system using a model is called model interpretability. New Azure update allows you to explain machine learning models both globally on all data and locally on a specific point. This service made available through multiple SDK packages. There are several direct and meta explainers available in the SDK. A few of them are SHAP, LIME and Mimic explainers.
Azure Machine Learning now has improved forecasting with Automated ML and GA. Meanwhile, AutomatedML is integrated with Databricks, SQL, CosmosDB, and HDInsight.
Data training with Automated ML
Integrating .Net with machine Learning service is an important development in this update. It’s done through ML.net and NimbusML. ML.Net is a machine learning framework for .net developers. It’s open source and cross-platform. .Net developers use it to train custom machine learning models in common scenarios like Sentiment Analysis, Image Classification, Recommendation. NimbusMl is a python binding for ML.Net which allows data scientists in train machine learning models using ML.Net in Azure Machine Learning service.
Azure has introduced a DevOps extension to Machine Learning service in order to simplify machine learning life cycle. It supports major activities in ML life cycle such as preparing data, training models, packaging models, validation and deploying models.
Another important feature added was FPGAs which is a machine learning inference option. It's based on Brainwave project which is a hardware architecture developed by Microsoft.
Visual Studio Subscription with GitHub Enterprise offering
Now you can buy both GitHub enterprise and Visual Studio subscription at one low price. This is a great opportunity for you to bring open source code and its community to your enterprise applications. Also, integrating Github Enterprise with Visual studio has other advantages such as saving a lot of time of developers to adopt Git.
New series of Azure Bot Framework enhancements now available
Azure bot service let you build intelligent bots that can do various tasks. With Azure bot service, you can build any type of bot from a simple Q&A bot to virtual assistant. Comprehensive, open source SDK allows you to build bots for any targeting device. In addition to that, you can build bots who can speak, listen and understand your users by integrating Azure cognitive service. MS Build 2019 adds few new features to Azure bot framework to make the bot development easier. Some of the newly introduced features are
- Adaptive dialogs to make more sophisticated conversations
- Enhance bot responses through Language Generation package
- An improved emulator is now available for debugging channels.
- New templates are available to make bots with different skill sets
Enhanced Bot responses and improved conversations with Bots.
Azure now supports GitHub identity single sign-on
Azure has done several updates to improve the GitHub integration to Azure projects. Single Sign-on Option is one of them. That means you can have one credential for Azure and GitHub. You can easily take your code from Github and deploy on Azure without delay. Thanks to this new feature, now you can log in to any product from Azure to Office 365. It works anywhere a personal Microsft account does.
You can now sign into Azure with Github.
Azure Search now supports complex search types
Azure search is the most advanced if not the only cloud search service with built-in AI capabilities that has the power to explore and identify relevant content from any type of information. But there are times that you encounter raw data sets with nested substructures that can’t be broken down smoothly into a tabular format such as multiple addresses and contact numbers for a single customer. This was the biggest issue Azure customers faced with Azure search and they were asking a solution for it. Azure has been able to deliver it with MS Build 2019 and now Azure search natively supports complex search types.
User-defined network policy in Azure Kubernetes Service (AKS) is now available
Azure has a fully managed, highly available, secure Kubernetes Service. You can easily manage and deploy containerized applications with it. It has the following features.
- Offers serverless Kubernetes
- Enterprise-grade security and governance
- integrated continuous integration and continuous delivery (CI/CD) experience
Azure has newly added user-defined network policy to AKS in order to secure network segmentation within Kubernetes. This feature also gives cluster operators the ability to control communication among pods.
You have two options to implement Network policy. You have to choose which option you would use when you create the AKS cluster and it can’t be changed later. The two options available to you are Azure native plug-in and Community project Calico.
Azure Active Directory B2C in China
China is going through a digital revolution. Chinese are enthusiastically embracing the internet. As a result of that, more and more companies are looking to bring their businesses into China. Supporting Customer Identity Access is an important requirement for these businesses when working with Chinese customers. That’s why Azure has decided to bring Azure Active Directory B2C, with all of its capabilities, to China. There are few things to be noted though.
- All customer data will stay in China.
- You need to be compliant to use only Azure China cloud services
- Enjoy a faster performance as you are closer to your organization’s location and customers.
- You can use popular Chinese identity providers like QQ, Weibo and WeChat.
Rules engine for Azure Content Delivery Network is now in preview
Azure Content delivery was giving the following services so far.
- Reliable and secure global content delivery.
- Integrated into Azure services with massive scalability
- Easy setup and pay according to how much you use it.
With MS Build 2019, Azure has decided to add Rules engine functionality to Azure CDN. You can customize CDN behavior or more precisely, how HTTP requests are handled using this new feature. Redirecting URLs based on matching conditions, modifying an HTTP header, defining a caching policy and blocking the delivery of certain content types are few examples for what you can do with the new feature. All these things help you to optimize the delivery of your content to the end user.
Azure Blockchain Service is now available in preview
Azure has introduced its own blockchain service. Azure blockchain service simplifies the building process of consortium blockchain networks and allows you to deploy them with few simple clicks. It has a feature called offload network management. Therefore you can focus on application logic without worrying too much about management. It has built-in governance that can automatically perform governing tasks such as adding new members, monitoring node health and setting permissions. Another advantage of Azure Blockchains service is you can build blockchain applications in an open, flexible platform which gives you the opportunity to use your favorite dev tools.
Azure Cognitive Services capabilities that are now generally available
You can find the following Azure cognitive capabilities available to you now.
Azure cognitive service now has the capability of producing speeches out of natural texts. We call this Natural Text to Speech and has the power of making voices which are so close to recordings of people. Azure has applied the latest innovations in deep learning for that.
Computer Vision read is another exciting capability of Cognitive service that can extract information from common file types like PDF and TIFF. It also has the ability to read multi-page documents.
NER, which stands for Names Entity Recognition, is another wonderful feature that excites many developers. NER is the ability to recognize entities like people, locations and organizations in a free-form text. In the latest Azure update, NER supports English and Spanish.
The following image shows what can improved cognitive services do:
Public Preview of tighter integration of Azure Data Explorer with Python and Spark
Azure Data Explorer is a fully managed real-time data analytics service. It’s fast and can work on large volumes of streaming data. It has many advantages such as
- Ability to quickly identify patterns and anomalies in the data.
- Explore data by asking questions and get answers quickly.
- Optimized query structure allows you to run as many queries as possible.
- Read data from Data Explorer into a Spark data frame and ingest data from data frame into Data Explorer using Spark connector available to you.
- Python and R plugins are available to you to use Python and R codes in your queries.
Notebook support for all Azure Cosmos DB APIs is in preview
Azure Cosmos DB is a fully managed, globally distributed database service. It's popular for the following features.
- Turnkey global distribution
- Transparent multi-master replication
- Single-digit millisecond latency at the 99th percentile
- Elastic and unlimited scalability
- API endpoints for MongoDB, Cassandra, Gremlin, SQL, Etcd, and Table to access and query data.
Azure has added Notebook support for Cosmos DB to enhance the developer experience. Thanks to this enhancement, analyzing and transaction processing are now possible on the same data in a given database. Developers have a few more advantages due to this notebook and Cosmos integration such as being able to interact with stored data, build and train machine learning models and perform interference on the data through Jupiter Notebook experience while inside the Azure portal.