AZ-900 Episode 16 | Azure Artificial Intelligence (AI) Services | Machine Learning Studio & Service

Summary

This video script delves into Azure's artificial intelligence offerings, focusing on machine learning services and the Azure Machine Learning Studio. It explains AI basics, the machine learning process, and how Azure Machine Learning aids in building, training, and deploying models as web services. The script highlights features like Automated ML, visual designer, and pipelines for end-to-end model management, showcasing Azure's comprehensive, cloud-based platform for machine learning model development.

Takeaways

• 🧠 AI is a branch of computer science that simulates human intelligence and capabilities, while Machine Learning (ML) is a subset of AI focused on teaching software to make predictions based on data.
• 🛠 Azure Machine Learning is a key service in Azure for building, training, validating, and deploying ML models as web services.
• 📚 The process of building an ML model involves training, packaging, validating, and deploying, with Azure Machine Learning providing tools to assist in each step.
• 📈 Azure Machine Learning Studio is a web-based interface that allows for the management of the entire Azure ML service, including notebooks, Automated ML, and a visual designer.
• 💻 Notebooks in Azure ML can be written in Python or R, offering a flexible environment for creating and executing ML scripts.
• 🔧 Automated ML (AutoML) simplifies the process by automatically testing various algorithms on the provided data to find the best model.
• 🎨 The Visual Designer in Azure ML enables building ML models through a drag-and-drop interface without writing any code.
• 🔄 Pipelines in Azure ML allow for the orchestration of the entire ML process from training to deployment, streamlining the workflow.
• 🗂️ Asset management in Azure ML includes datasets, experiments, pipelines, models, and endpoints, providing a centralized way to manage all components of the ML lifecycle.
• 🖥️ Compute resources in Azure ML can be managed for training and deploying models, abstracting away the underlying infrastructure concerns.
• 🔗 Azure ML integrates with other Azure services such as Azure Blob Storage and Azure File Share for data storage and management.

Q & A

What is the main focus of the video script?

The main focus of the video script is to discuss artificial intelligence products in Azure, particularly the Azure Machine Learning service and how it can be used to build, train, and deploy machine learning models.

What is the relationship between AI and machine learning as described in the script?

In the script, AI is described as a branch of computer science that simulates human intelligence, while machine learning is a subcategory of AI where software is taught to make predictions and draw conclusions based on data.

What is the role of 'building a model' in the context of machine learning?

Building a model in machine learning involves teaching the software to analyze data and make predictions. This process includes training the model with data, validating its accuracy, and potentially retraining it to improve results.

How does Azure Machine Learning assist in the machine learning process?

Azure Machine Learning provides a set of tools that assist in the entire process of building a machine learning model, including training, packaging, validation, deployment, monitoring, and retraining, as well as managing compute resources and automating the selection of algorithms.

What is the significance of 'AutoML' in Azure Machine Learning?

AutoML in Azure Machine Learning is an automated process that allows users to apply various algorithms to their data to determine which one performs best, and then deploy that algorithm as a designated web service.

What is a 'pipeline' in the context of Azure Machine Learning?

In Azure Machine Learning, a pipeline is an end-to-end solution that allows users to build, manage, and execute the entire process of training, deploying, and managing machine learning models, whether using notebooks, designer, or automated tools.

How does the Azure Machine Learning Studio differ from the older Machine Learning Studio?

The Azure Machine Learning Studio is a web-based visual interface for managing the entire Azure Machine Learning service, offering additional features and an updated experience compared to the older Machine Learning Studio, which is no longer actively developed.

What are some of the tools provided by Azure Machine Learning for building machine learning models?

Azure Machine Learning provides tools such as Jupyter notebooks written in Python or R, a visual designer for drag-and-drop model building, and Automated ML for algorithm selection and parameter tuning.

What is a 'compute target' in Azure Machine Learning, and why is it needed?

A compute target in Azure Machine Learning is a virtual machine or other compute resource that is used to run the machine learning workflow. It is needed to perform tasks such as training and scoring the model without the user having to manage the underlying infrastructure.

How does asset management work in Azure Machine Learning?

Asset management in Azure Machine Learning involves managing various components such as datasets, experiments, pipelines, models, and endpoints. These assets are tied together in the workspace and can be used to track the machine learning process and deployment of models.

What are the key features of Azure Machine Learning service mentioned in the script?

The key features of Azure Machine Learning service mentioned in the script include notebooks, Automated ML, a visual designer for building machine learning pipelines without coding, management of data and compute resources, and integration into machine learning pipelines for orchestration of tasks.

Outlines

00:00

🤖 Introduction to Azure AI and Machine Learning

This paragraph introduces the topic of artificial intelligence (AI) products in Azure, focusing on the machine learning service and studio. It explains AI as a branch of computer science that simulates human intelligence and machine learning as a subcategory that enables software to make predictions based on data. The key process in machine learning is building models, which is facilitated by Azure Machine Learning. The paragraph outlines the process of training, validating, deploying, and monitoring machine learning models, and highlights the tools provided by Azure, such as notebooks, a visual designer, and AutoML for automated algorithm selection. It also introduces the concept of pipelines for end-to-end machine learning model development.

05:02

📊 End-to-End Machine Learning with Azure

The second paragraph delves into the practical aspects of using Azure Machine Learning Studio for creating, managing, and publishing machine learning models. It describes the Azure Machine Learning workspace as a central resource that ties together compute resources, permissions, runs, pipelines, experiments, and model deployments. The paragraph also distinguishes the new Azure Machine Learning service from the older 'Machine Learning Studio,' which is no longer actively developed. Key features highlighted include the use of notebooks for Python or R, Automated ML for algorithm selection and parameter tuning, and the visual designer for building machine learning pipelines without coding. The paragraph concludes with a mention of asset management and other features for managing data and compute resources, as well as orchestrating model training and deployment through machine learning pipelines.

Keywords

💡Artificial Intelligence (AI)

Artificial Intelligence, or AI, is a branch of computer science that focuses on creating software capable of simulating human intelligence and cognitive capabilities. In the context of the video, AI serves as the overarching theme, with the episode specifically focusing on AI products within Azure, Microsoft's cloud computing service platform. The script discusses how AI can be utilized to perform tasks that typically require human intelligence, such as learning and problem-solving.

💡Machine Learning

Machine Learning is a subset of AI that involves teaching software to learn from and make predictions or decisions based on data. The script emphasizes that machine learning is integral to AI, where the software is trained to 'draw conclusions' without being explicitly programmed for every possible scenario. The process of training the software is referred to as 'building a model' in the video.

💡Azure Machine Learning

Azure Machine Learning is a key service within Microsoft Azure that assists in the process of building machine learning models. The script describes it as providing a set of tools to facilitate the entire machine learning process, from training models based on data to deploying them as web services. It is central to the episode's discussion on how to manage and streamline machine learning workflows in the cloud.

💡Model Training

Model training refers to the process of teaching a machine learning model to make accurate predictions based on input data. The video script explains that this involves using data to train the model, validating its performance, and then potentially deploying it as a web service if the results are satisfactory. It's a fundamental step in the machine learning process highlighted in the script.

💡Web Services

In the context of the video, web services are the deployed versions of machine learning models that can be accessed over the internet. The script mentions that once a model is trained and validated, it can be deployed as a web service, allowing for broader access and application of the model's predictive capabilities.

💡Automated Machine Learning (AutoML)

AutoML is a feature of Azure Machine Learning that automates the process of model selection and training. The script explains that AutoML allows users to 'drop random algorithms at our data' to determine which algorithm performs best. This feature is designed to simplify the machine learning process by reducing the need for manual tuning and selection of algorithms.

💡Machine Learning Studio

Machine Learning Studio is a web-based visual interface for managing the Azure Machine Learning service. The script describes it as a tool that provides an end-to-end solution for building machine learning models, with features like notebooks, automated ML, and a visual designer. It is a central component in the Azure Machine Learning platform.

💡Notebooks

Notebooks, in the context of the video, are interactive computational environments that allow users to write and execute scripts, typically in Python or R. The script mentions that Azure Machine Learning provides notebooks as a tool for creating machine learning models, offering a flexible way to experiment with and train models.

💡Visual Designer

The Visual Designer is a feature within Azure Machine Learning Studio that allows users to build machine learning models using a drag-and-drop interface. The script highlights this as a user-friendly way to construct models without writing code, making the machine learning process more accessible to a broader audience.

💡Pipelines

In the script, pipelines refer to the end-to-end process of building, training, and deploying machine learning models. They are a key feature of Azure Machine Learning that allows for the orchestration of various steps in the machine learning workflow, from data preparation to model deployment.

💡Asset Management

Asset management in the context of the video pertains to the organization and control of various components involved in machine learning, such as datasets, experiments, pipelines, and models. The script explains that Azure Machine Learning provides tools for managing these assets, which is crucial for maintaining an orderly and efficient machine learning process.

💡Compute Resources

Compute resources are the hardware and software components required to perform computations, such as training machine learning models. The script mentions that Azure Machine Learning manages compute resources, allowing users to focus on the machine learning process without worrying about the underlying infrastructure.

💡Data Stores

Data stores are locations where data is stored and managed, such as Azure Blob Storage and Azure File Share mentioned in the script. They are essential for connecting to data sources within Azure, which is a critical step in the machine learning process as the models are trained and validated using this data.

Highlights

Introduction to artificial intelligence products in Azure, focusing on machine learning service and studio.

AI is a branch of computer science that simulates human intelligence, with machine learning as a subcategory for predictive analysis.

Azure Machine Learning is a key service for building machine learning models through training, validation, and deployment.

The process of building a machine learning model includes training, packaging, validating, and deploying as web services.

Azure Machine Learning provides tools such as Python or R notebooks and a visual designer for model building.

Automated Machine Learning (AutoML) allows for algorithm selection and parameter tuning to find the best model.

Machine learning models can be managed through Azure's compute resources without worrying about infrastructure.

Introduction to Azure Machine Learning pipelines for end-to-end model building processes.

Demonstration of navigating to the Azure Machine Learning workspace and launching the studio.

Explanation of using notebooks for creating scripts or executing provided samples for model building.

Overview of AutomatedML for experimenting with different algorithms and parameters to achieve the best model score.

Introduction to the Visual Designer for building machine learning models with a drag-and-drop interface.

Asset management in Azure Machine Learning includes datasets, experiments, pipelines, and model endpoints.

Management of compute resources for training and deploying models, and data stores for connecting to data sources.

Live demonstration of building a machine learning model using the Visual Designer with a simple workflow.

Explanation of creating a compute target for running the workflow and selecting it for model training.

Real-time UI demonstration showing the model building process and evaluation results.

Summary of Azure Machine Learning as an end-to-end cloud-based platform for managing ML models.

Differentiation between the old Machine Learning Studio and the new Azure Machine Learning service.

Key features of Azure Machine Learning Service including notebooks, AutomatedML, and Visual Designer.

Management of data and compute resources, and orchestration of model training and deployment through pipelines.

Information about where to find materials for the episode and a teaser for the next episode on serverless computing in Azure.

Transcripts

hello everyone welcome back again after

big data we can move to artificial

intelligence products in azure

and that's the focus of our today's

episode stay tuned

[Music]

for the ai today we'll be talking about

machine learning service and machine

learning studio

but before we move to those solutions

let's talk about

ai in general ai is a branch of computer

science where we use our software

to simulate human intelligence and

capabilities

whereas machine learning is a

subcategory of ai

where we use that software and we teach

it to draw some conclusions and make

some predictions based on

our data and the process of teaching the

software to do that

is called building a model and the key

service to do that in azure is called

azure machine learning

typically the process of building a

machine learning model consists of

training the model based on our data

then packaging and validating that model

if we are happy with the results we can

deploy those models as a web services

then monitoring those web services and

retraining the model

to get even better results and azure

machine learning is here to help us with

this entire process by providing us with

set of tools

tools like notebooks written in python

or r

or a visual designer which allows us to

build machine learning models

using a simple drag and drop experience

directly in our browsers

additionally machine learning models

allows us to manage all the compute

resources where we train

package validate and deploy those models

so that we don't have to worry about

azure

infrastructure and underlying resources

ourselves

additionally azure machine learning

comes with something called automl

this automated process allows us to drop

random algorithms at our data

and see which one scores the best and

deploy that as our designated web

service

and lastly there is also a feature of

pipelines which allows us to build this

entire process

end-to-end whether we are using those

notebooks designer or auto tools

this is complete end-to-end solution for

building machine learning models

let me quickly present to you how azure

machine learning works

first i will navigate to my resource

group called az900

ml inside of this resource group i have

my machine learning workspace resource

which allows me to manage machine

learning service and all its components

in azure portal there's not a lot of

things we can do there are very few

blades when it comes to management of

machine learning workspace

in order to go to machine learning

workspace we need to hit on the launch

studio button

or hit this url here above so let's

click on launch studio button which will

open an

azure machine learning studio this is a

web-based

visual interface for management of

entire azure machine learning service

for example in an outdoor section you

have notebooks automated ml and

visual designer but let's go to the

notebook first

this is a workspace where you can create

your own scripts

or instead you can try out some samples

provided by microsoft

navigate to any notebook that you like

read the tutorial and

execute the provided samples and you

will build your machine learning model

in no time

or instead of writing scripts try

automatedml which allows you to throw

random algorithms at your data

tweak some parameters and see which

model scores the best

and deploy that as a web service and if

this wasn't enough you also have

designer

designer is a visual way of building

machine learning models

with a drag and drop features and

machine learning has also a lot of

features

around asset management assets are

things like data sets

experiments pipelines models that you've

built

and endpoints that you deploy those

models to there's also plenty of other

features which allow you to manage your

compute resources where you train and

deploy your models

data stores where you connect to your

data sources in azure for example azure

blob storage and azure file share

but for now let me quickly go back to

designer where i will build my own

machine learning model using a visual

experience

and before i begin i will create a

compute target so a simple virtual

machine that will run my workflow

so let me simply hit create new and use

predefined size

give it a name like my compute

and hit save after the machine has been

provisioned

i can simply select it hit save and

start building my workflow

i will speed this part up because for

the fundamentals training the process

itself is not important but i want to

show you how easy it is

simply drag and drop your data then

select the cons that will be used

to build machine learning model then

clean your data using another step

and split your data so that you can grab

part of the data for the training and

part for the evaluation of your model

next simply train your model to forecast

the price using linear regression

if you're happy with it simply drag and

drop another block to score the model

and the last step to evaluate the

results and we're done

if we are happy with our pipeline we can

submit it to run

by either choosing existing or new

experiment

let's create new one let's give it a

name

i will call my experiment demo and

submit it to run

depending on the complexity of your

pipeline and the machine that you pick

to train the model

this process might take from few minutes

to even half an hour or even

couple of hours so i've sped this up by

a lot

just to show you that ui very nicely

shows you how your model is being built

in real time and now if you are a data

scientist you can either check the

evaluation results so you can check all

the logs all the outputs that we're

generating during this pipeline check

all the details

or by going to score model where they

can visualize their data sets

where they can see a lot of extra

information about the data set they use

for

training of this model including the

price that they forecasted

and the scores so to summarize azure

machine learning service

is an end-to-end cloud-based platform

for creating managing and publishing

of our machine learning models this is

of course a platform as a service

offering in azure

a top level resource in azure machine

learning service

is called machine learning workspace

think of it like this workspace ties

everything together all the computer

resources all the permissions

all the runs pipelines experiments

history

connection to external services like

azure storage accounts container

registry deployments of your model

literally everything is managed through

azure machine learning

workspace and machine learning studio is

this web portal that we were using for

our end-to-end management of the

workspace

a quick note that i have for you here is

that in the past there was another

service called

machine learning studio this is

completely different from the one that

we're talking about today

because that service is no longer

actively being developed and new

customers are being encouraged to use

new machine learning because it also has

the studio experience

with some additional features and the

key features that you get by using azure

machine learning service are of course

notebooks

written in python or r automated ml

where you can throw

a lot of algorithms and tweak some

parameters to find the best algorithm

and to build the best model for your

data

and with visual designer you can build

your machine learning pipelines without

writing a single line of code

you can also manage your data and

compute resources

so all the resources to manage and store

your data

as well as process those models and

deploy them as a web services

and lastly everything is nicely tied

into other machine learning pipelines

which allow you to orchestrate this

model training deployment and all the

management tasks

as always all the materials for this

episode are found

on my website under episode 16. and

that's it for this one but don't go away

because next episode is all about

serverless computing in azure

so if you like my work support the

channel by subscribing liking and

commenting and

see in the next one