New features for Azure Machine Learning are now in preview
Veröffentlicht am Montag, 6. Mai 2019
- Open Datasets - Open Datasets are a collection of datasets from the public domain to accelerate the development of machine learning models built in Azure. Open Datasets integrates with Machine Leaning Studio or can be accessed from python notebooks in Azure Machine Learning Service. Azure Open Datasets offer good quality data from the public domain, which is often hard to find and expensive to curate. Data scientists will be more productive focusing on model building rather than data preparation.
- Visual interface : The new visual interface for Azure Machine Learning adds drag-and-drop workflow capabilities to Azure Machine Learning service. It simplifies the process of building, testing, and deploying machine learning models for customers who prefer a visual versus coding experience. This integration brings the best from ML Studio and AML service together. The drag-n-drop experience let any data scientist can quickly build a model without coding. The tool also gives enough flexibility for data scientist to fine tune its model. The AML service as the backend platform offer all the scalability, security, debuggability … that ML studio can’t give. The easy deployment capability in visual interface enables easy generation of score.py file and creating images. With a few clicks, a trained model can be deployed to any AKS cluster associated with AML service.
- Automated ML – UX :
- Deploy as web services to predict on new data
- Get the best model for classification, regression, or forecasting problems with a few clicks of a button
- Analyze the generated models
- Citizen data scientists: Generate ML models without needing to write Python code (or any type if code) . Data scientists: Explore and generate hundreds of models quickly, then continue to optimize the best ones in Jupyter notebook
- Notebook VMs: Azure Machine Learning will go into private preview with a hosted notebook service in mid April, and we expect to take this to public preview in May. Hosted notebooks provide a code-first experience where users can perform every operation supported by the Azure Machine Learning Python SDK using a familiar Jupyter notebook. Hosted notebooks simplify the getting-started process by providing a secure, enterprise-ready environment for ML practitioners. In the private preview customers will be able to: access a notebook integrated into the Azure ML workspace, use preconfigured Azure ML notebooks with no set up required, fully customize their notebook VMs including having the ability to add packages and drivers.
Now you can use MLflow with your Azure Machine Learning workspace to log metrics and artifacts from your training runs in a centralized, secure, scalable location. MLflow tracking can be done from your local machine, a virtual machine or a remote compute environment.
- Data Box Edge with FPGA: FPGAs are a machine learning inferencing option, based on Project Brainwave, a hardware architecture from Microsoft. Data scientists and developers can use FPGAs to accelerate real-time AI calculations. These Hardware Accelerated Models are now generally available in the cloud, along with a preview of models deployed to Data Box Edge. FPGAs offer performance, flexibility, and scale and are available only through Azure Machine Learning. They make it possible to achieve low latency for real-time inferencing requests, mitigating the need for asynchronous requests (batching).