Microsoft Azure Machine Learning - DP-100

Explore how a workspace provides a logical environment to run experiments, group resources, and track development history, and learn to create it via graphical interface, Python SDK, or CLI.

Learn to create an Azure machine learning workspace through the graphical portal by selecting subscription, resource group, workspace name, region, and edition, then launch the machine learning studio.

Connect visual studio code to your Azure account and create a workspace with the Azure ML SDK, signing in via web, then configure Free Trial subscription, resource group, and location.

Create an Azure Machine Learning workspace from the command line by launching an elevated session, naming the workspace, selecting the resource group, installing the Azure CLI extension, and logging in.

Explore how experiments in Azure ML run scripts or pipelines, track metrics, and preserve run history to compare results across data sets, with a simple random array example.

Connect to your Azure Machine Learning workspace and create a new notebook to run experiments, logging runs and metrics like mean, min, and max for review in the portal.

Learn to run Azure Machine Learning experiments by encapsulating logic in an API and configuring Python environments with run and script configurations, logging metrics and histogram image from script workflow.

This lecture introduces estimators in Azure ML, showing how a single configuration and script run configuration improves flexibility and reduces compute-context dependency, while handling package dependencies with predefined frameworks.

Create and run an estimator-based experiment in azure ml by configuring workspace, scripts, and packages (matplotlib), then monitor results and prepare to implement a classification algorithm with the iris dataset.

Explore a concrete iris dataset classification workflow using logistic regression, including data encoding, train-test split, evaluation with accuracy and confusion matrix, and saving the trained model.

Explore how to pass hyperparameters to estimators, run experiments with regression models, and manage outputs, logs, and model registration in Azure Machine Learning.

Discover how datastores connect your workspace to storage accounts, enabling training data from Azure Blob storage, raw data, and SQL data; create an Azure data source in the graphical interface.

Create a data store in Azure Machine Learning Studio using the graphical interface, selecting your storage account and blob container, and supplying the authentication credentials from the portal.

Create and manage datastores in Azure ML using the SDK by connecting to a workspace, registering a blob data store, setting a default datastore, and uploading data to the datastore.

Register, version, and use datasets in the Azure machine learning workspace via data stores. Create datasets from local files, data sources, or open data, and manage versions.

Learn to pass a dataset as input to an Azure Machine Learning experiment, create an estimator, configure inputs and data stores, and log the dataset's number of rows.

Define and attach custom environments for Azure ML experiments using dot yaml files or conda dependencies, ensuring the correct Python runtime and required packages like matplotlib are available.

Create a compute target using the graphical interface in Azure Machine Learning, naming the instance, selecting a VM size, and launching JupyterLab to connect with your workspace.

Connect your Azure ML workspace via the Python SDK, create a compute target with a cluster size and min/max nodes, then run an estimator-based experiment on it.

Explore the distinction between parameters and hyperparameters, define a hyperparameter search space, and compare grid, random, and baseline sampling methods for tuning model performance.

Learn to tune a classification model’s hyperparameters in Azure Machine Learning using Python SDK, selecting the best regularization coefficient via hyperdrive with grid sampling.

Explore automated machine learning in Azure ML, with algorithm selection and preprocessing such as scaling and handling missing values for classification and time-series forecasting using decision trees and random forests.

Explore automated machine learning (AutoML) with Azure ML by creating and tuning experiments, configuring environments, and managing notebooks, scripts, and dependencies within a workspace.

Learn to use the Azure Machine Learning designer to build a pipeline in a workspace, configure a compute target, and import data with the iris dataset.

Explore building Azure machine learning pipelines using the designer: import iris data, transform features, split data, train a multiclass decision tree, score and evaluate for 93% accuracy.

Deploy your registered machine learning model to production by packaging it with an image, an entry script, and an environment, then create inference and deployment configurations for real-time predictions.

Learn to deploy a registered model as a web service with the Python SDK in Azure ML, create an inference config and environment YAML, and configure an ACI deployment.

Deploy a model as a web service in Azure Machine Learning, verify it's healthy, and perform inferences via the endpoint using JSON data.

Deploy your model with the graphical interface by selecting the model, choosing a deployment service, configuring the container, and providing the entry script and dependencies; then verify the rest endpoint.

Learn to monitor a deployed machine learning model by capturing telemetry with application insights, linking it to your Azure ML endpoint, and analyzing logs and metrics with the ML SDK.