The Complete Course from ExamCollection industry leading experts to help you prepare and provides the full 360 solution for self prep including AI-900: Microsoft Azure AI Fundamentals Certification Video Training Course, Practice Test Questions and Answers, Study Guide & Exam Dumps.

Describe fundamental principles of machine learning on Azure

6. Machine Learning Techniques

Hi and welcome back. Now in this chapter, I want to go through the categories of the different machine learning techniques that we have. So first, there is supervised learning. So over here, you're going out and using the algorithm that you have, the machine learning algorithm, along with your data, to go ahead and develop a machine learning model. Now, in terms of this model, when it comes to supervised learning, over here, you will already have your labelled data in place. Remember in the earlier chapter when I talked about your data? You had your features, which go ahead and feed into the machine learning model. And when training a model, you only go ahead and enter, let's say in this example, the total sales amount. You only have your labels in place. So this makes it much easier for the machine learning model to go out and develop a relationship between the features and the label. So this is known as supervised learning, but in some cases, you might not have the labels in place. So over here, the algorithm will try to organise the data and then decide on the outcome accordingly. This is known as unsupervised learning. And then finally, we have something known as reinforcement learning. To give you an example of reinforcement learning, imagine you're trying to ensure that a specific robot in this case can navigate via a specific path, a specific route. Over here, you are not giving any sort ofinstructions or you are not giving what is theoutlay or the structure of this course. As such, you want to ensure that the robot is able to go in and learn as it is navigating on this particular course or this particular path. So this is known as reinforcement learning. So, it's very important when it actually comes to autonomous driving. So let's say that when it comes to autonomous driving, the system itself needs to understand how to navigate on the roads based on, basically, what it finds on the roads themselves. It needs to find different obstacles and learn on its own. So over here, let's say that the robot is going ahead and starting on this particular path. Now, if it is following the correct path and not running into any obstacles, you will go ahead and give it something known as a reward, a plus point, just to ensure that it is following the correct path. If it is deviating away from the path, then I should go ahead and give it negative points. So over here, you are trying to go ahead and sway the actual outcome and allow the robot to go ahead and decide what to do next. Because if you put a negative spin on what the robot is doing, the robot will try to change the way it approaches, how it navigates through that specific path to that specific course, and then it will try to stay on that path itself. So over here, they're trying to give an example when it comes to reinforcement learning. So in this chapter, I kind of want to gothrough these different machine learning techniques that you have. So you have supervised learning, unsupervised learning, and reinforcement learning.

7. Machine Learning Data - Features and Labels

Hi. Welcome back. Now in this chapter, I just want to go through two concepts when it comes to your data. So the data that you go out and feed into your machine learning process in order to train your model, in order to develop your model. So remember, when it comes to your model, again, you go ahead and feed in your machine learning algorithm, and you go ahead and feed in your data. Now, when it comes to a classification of machine learning, that's supervised learning. So you'll go ahead and basically feed in your data. So, giving some example data over here, so over here, in terms of this data set, let's say this data set pertains to the sales amount for video courses that are being sold by an online platform. So let's say they have this data in place that is based on a particular unit price and the length of the course. Over here, they only have the total sales amount, and they want to go ahead and predict what could be the total sales amount based on new courses that get published on their platform. So for new courses that are being published on their platform, they would already have the unit price in place for the course that would already be defined and the length in minutes for the course itself. And based on that, they want to go ahead and determine what could be the total sale amount for that particular course. So based on the historical data that they already have for their existing courses, they want to go ahead and feed it along with the machine learning algorithm in order to go ahead and develop that machine learning model. Now, over here, when it comes to what data you're going to feed into the machine learning process, it's very important to understand and to decide what data you're going to use because you cannot go ahead and use all the data that you have. So for example, in this data set, it doesn't make much sense to go ahead and use the course ID when it comes to the data set because when it comes to the prediction for the total sales amount, it won't be dependent on the course ID. So over here, you have to go ahead and select the data that you want to go out and feed into the machine learning process. So over here, we are going to go ahead and feed in the data of the unit price, the length, and the total sales amount to go ahead and train our machine learning model. Now, over here, what we want this model to predict in the end is the total sales amount. So when it comes to your data, this one column could be this one column.Please know that machine learning will go ahead and predict multiple columns as well. This is basically the value of Y in your function. So remember, you have y is equal to the function of x, which has a very simplistic equation when it comes to developing a machine learning model. So over here, you are trying to predict the value of a label when it comes to the data that you feed into the machine learning model, this is known as a label.So over here when it comes to supervised learning for your data, you would only have the existing values for the total sales amount for your existing courses, and this is known as a label. It's very important to know the other columns of data that are feeding in to go ahead and make this prediction. That's basically your values of x; these are known as your features of data. So what are the features that you are using to go ahead and make your prediction? So, in this chapter, I want to kind of go through these very important concepts when it comes to data, when it comes to the machine learning process: one is the features of your data set, and the other is the labels that you can actually attach to a data set.

8. Lab - Azure Machine Learning - Creating a workspace

Now, we're going to start with machine learning on Azure. Now in Azure, you have the Azure Machine Learning Service that actually provides you a complete cloud-based environment where you can actually go ahead and train your machine learning models. You can go ahead and deploy these models. You can also go ahead and automate the entire machine learning process, and you can also go ahead and manage and track your machine learning models. When we actually go through our labs, you will see how easy it is to go and actually create your machine learning models. And this is good for those people who don't have any sort of prior development experience. So normally, if you want to go ahead and create a machine learning model, you might go ahead and use, let's say, the Python programming language to go ahead and develop that model. But if you have no sort of coding experience in place, you can actually go ahead and make use of the Azure Machine Learning service. This is not to say that you can't use Python as a language in the Azure Machine Learning Service. No, you can do that. But this service also provides the ability for those with no prior coding experience to go ahead and start working with machine learning. and that's great. So even for myself, in terms of these labs, I really don't have to go ahead and perform any sort of coding in order to go ahead and train a machine learning model. So in this service, you are something known as the Azure Machine Learning Studio. So the Azure Machine Learning Studio is where you're going to go ahead and train your model, deploy a model, et cetera. This is a web-based experience that is available when it comes to Azure Machine Learning. Microsoft Azure Machine Learning You have the designer. So over here, you can go ahead and drag and drop your data sets and your modules, and I'll go ahead and explain this. When we go into our labs, you can go ahead and create experiments and pipelines. So an experiment is actually used to go and process your data, analyse your data, and train and test your model. Now we're going to go ahead and go onto Azure, and we're going to go ahead and create something known as an Azure Machine Learning Workspace. This Azure Machine Learning Workspace will actually go ahead and give you that machine learning studio. And we're going to go ahead and use that machine learning studio to go ahead and train our machine learning models. Now, when you go ahead and create a resource based on the Azure Machine Learning Workspace in the background, it's actually going to go ahead and create some other resources as well. This is something that's completely managed by the Asia Machine Learning Workspace. You have to be concerned about the workings of these resources because we are focusing on the machine learning workspace. But I don't tell you what the resources are that get created. So, first, you have an Azure search account. This has a data store for the workspace itself. Then you have the Azure Container. Let's see, this is used for resting Docker-based containers. Then you have Azure application insights. So Azure Application Insights is an application monitoring tool that is available in Azure. So this stores mounting information for your machine learning models. And it also creates an Azure Key vault. So if you have sensitive information that needs to be stored by the computer targets in your workspace, you can go ahead and make use of the Azure Keyword. So let's go ahead and go onto Azure, and let's go ahead and create the Azure Machine Learning Workspace. So here we are in Azure. Now, let me go ahead and click on Add in all resources. So over here, let me go ahead and search for machine learning. So I'll go ahead and choose the service of machine learning. Let me go ahead and hit on Create. Now over here, I'm going to go ahead and choose my subscription. And let's go ahead and choose our learning group, our resource group, which we created earlier. Now we'll create a separate workspace for our machine learning workspace. So let's go ahead and see if this name is available. So it is available. I'll choose the North European region. So automatically, it's going to go ahead and create a storage account and, as your keyboard indicates, an application inside a resource. And I also have to go ahead and give a name for a container registry. So fine. I'll hit save. Let me go on to Next for networking. I'll leave everything as it is. I'll go on to advance; leave everything as it is. Then we go on to review and create. And let's go ahead and create our machine learning workspace. What we have to focus on next is actually the Machine Learning Studio, which comes in the Machine Learning Workspace. So let's come back. This might take around four to five minutes. We can now add the resource once you have the workspace in place. And then over here, we can click on Launch Studio. So this is going to go ahead and launch the Azure Machine Learning Studio. And this is what we're going to work on, right? So let's put it in this chapter and move on to the next one, where we'll take a closer look at your machine learning studio.

9. Lab - Building a Classification Machine Learning Pipeline - Your Dataset

So in the last chapter, we had gone ahead and created our Azure Machine Learning workspace, and we have now gone ahead and launched the Azure Machine Learning Studio. So the Azure Machine Learning Studio you can go aheadand train your machine learning model models and you don'tneed to have any sort of coding experience when itcomes to building your machine learning models. So let me go through some of the aspects that are available in your machine learning studio. So firstly, in the author section, which is available over here, the same thing is available over here as well. So you can go ahead and invoke the designer, and in the designer, you can start building the machine learning or the flow for your machine learning model. Then you can also go ahead and automate the entire machine learning process when it comes to automated machine learning. You can also go ahead and quickly create notebooks. So notebooks are an interactive way to go out and interact with your data. So over here, you need some coding experience when it comes to working with notebooks. When it comes to this particular exam, you don't have to go through the notebooks when it comes to your Machine Learning Studio. Then you can go ahead and actually create your data sets. So over here, these are the data sets that help to train and test your machine learning model. You could also go ahead and restore your models and maybe actually go out and work with the designer. We'll actually be creating pipelines and experiments over there. Then you'll go ahead and use endpoints when you want to go ahead and consume your machine learning model. When you want to go ahead and train and deploy your models, you have to go ahead and ensure that you have the right compute infrastructure in place because in the end, the entire process is going to run on the underlying physical machines. You don't have data stores. So data stores can basically be used to go ahead and contain your data sets. So these are the different aspects that are available when it comes to the Azure Machine Learning Studio. So let me go ahead and click "Create New" and then over here. Let's go ahead and create a new pipeline. So we're going to go ahead and create a completely new pipeline, and in this pipeline we are going to see how to create a machine learning model that is based on the classification technique when it comes to machine learning. So over here, you have a lot of prebuilt assets that are already in place, and what I mean by these assets is that firstly, we have inbuilt data sets that are already in place. So if you don't have any data with you and you want to go out and learn about machine learning, you can actually go ahead and use these sample data sets, and that is what we are going to use in this particular course. So this will ensure that anyone can actually go ahead and start learning machine learning. When it comes to Azure itself, we can go ahead and make use of these sample data sets. You also have various modules in place. So for example, in order to go ahead and train your model, you can go ahead and use the train model module, which is over here. Apart from that, over here you have your different machine learning algorithms in place. So you can actually go ahead and use these machine learning algorithms to go ahead and train your machine learning model. So over here, you have a lot of modules in place. Now all of these modules can be dragged over onto this empty canvas over here. So in this canvas, we are going to go ahead and actually build a pipeline. This pipeline will be used for training our machine learning model. Then, on the right hand side, we have settings for the pipeline itself. So firstly, let's go ahead and give a name to our pipeline. So just give a simple name. Now let me go ahead and just close the setting screens so that we can easily go ahead and see the canvas over here. Now I said we were going to go ahead and use the sample data sets that are available in Azure Machine Learning Studio. Now I'm going to be using this Adult Census Income Binary Classification Data Set. So I want to firstly go through the classification model or the classification training that we can use for building a machine-learning model. So we're going to go ahead and use this data set, so we can go ahead and easily drag the data set onto the canvas over here. Now if you want to go ahead and look at details about this data set or any data set that you actually drag onto this canvas, you can right-click on the data set. You can go ahead and click on Visualize, and you can go ahead and choose Data Set Output. And over here, you'll first see the number of rows that we have in the data set and the number of columns that we have in the data set. So over here we have the different columns for the data set. So when it comes to the people that define this data set, you have data such as the age, what is the current work class, what is the education, et cetera. So you have string values in place, you have numbers in place, and all of them are giving a description about the data if you click on any column. So, over here, we look at the first row itself. So we have one person with a 39-year-old age of 39.Now when you click on that age column over here, it's actually giving you an overview of how the data is being distributed across this entire column. So this is good. In this collection of rows, it's actually giving you stats about the different values in this particular column. So over here, it's telling you that the mean age across all the values over here in this particular column is 38, 58, and 116. It's telling you that in this particular column of data, there is a person who has an age of 17. So this is the minimum age that has been recorded inthis particular data column and the maximum age is 90. Now, all of this information is very important. Information also suggests the number of missing values is also important. Now why is this important? So for example, when it comes to machine learning, the data that you actually feed into your model needs to be accurate. So this is important because this data is going to be fed to build your machine learning model, and your data needs to be accurate. For example, if you had a minimum age of, let's say, two, that wouldn't make sense. In terms of income, you would not have a person your age. So that data would be wrong even if you had a maximum age of, let's say, 200. doesn't make sense, right? So this is important when it comes to understanding your data. So this entire screen is actually giving you a lot of information about the data itself. Also, it should not have missing values because if it doesn't have a value, then it doesn't make sense to use that column for building the machine learning model. Remember the machine learning model. Remember that the algorithm along with your data is building the machine learning model. And that's why your data needs to be complete. It needs to be accurate. So over here, for each column, you can actually go ahead and look at all of the information about the data itself. Over here, if you go ahead and scroll down, it shows you the distribution of the data as well. So over here, if I look at all of this information, what is all of this information? All of these are features. These are the features of our data set. So over here, we then have our label. So all of this information over here basically tells us that for a person that has this data, this feature data, is the income of that person less than or equal to $60,000 per year, or will it be more than $50,000 per year? So over here, we are looking at binary classification. So in the label, we are just trying to predict, based on the data that we provide for the person, would the income be less than or equal to $50,000 or would it be greater than $50,000? So if you actually go through the data values over here, you will see data in places where the value is greater than 50K. So we already have some information in place. So this information could have been collected via a survey. This information is something that you already have in place.So you now have your features in place, as well as your label. So all of this data is going to be fed into this machine learning process, into this pipeline that we're going to create. And this data is going to be used to go ahead and build your machine learning model. Over here, this is one of the most important concepts that I want to explain about the data set itself. Over here, you can look at the different aspects of your data. This data set is mainly based forcreating a machine learning that's based onthe binary classification learning technique. So over here, remember, we are just trying to predict what the income of a person will be, whether it will be less than or equal to $50,000 or whether it will be greater than $50,000. So let us mark and move on to the next chapter. We'll again go through this entire pipeline. Let me go ahead and hit "close" and let's go on to the next video. Then let's go ahead and expand our pipeline.

10. Lab - Building a Classification Machine Learning Pipeline - Splitting data

Hi and welcome back. So in the last chapter, I explained how to analyse the data that you had. So we had gone ahead and used one of these sample data sets, which are available in machine learning. in the service itself. That data set is dragged. Remember, this is data set onto the canvas over here. Now let's go ahead and expand this pipeline. This is part of a pipeline in the Azure Machine Learning Service. We are currently in the Azure Machine Learning Designer. We are going to author a pipeline, and this pipeline is going to be used to train a machine learning model that is based on the binary classification machine learning technique. Now, the next thing we want to do is to go ahead and split our data. So remember, in one of our earlier chapters, I told you that when you want to go ahead and train your machine learning model, you'll actually go ahead and split your data. So you will have training data. That data will be used to go ahead and train the machine learning model. And then you'll go ahead and have test data that will be used to go ahead and test the machine learning model in terms of accuracy. So there is a lot that you can actually do prior to that with your data set. But, since we're starting with a machine learning pipeline, we'll go ahead and start splitting our data. Now in order to go ahead and split our data in the designer, we can actually go ahead and search for a module. So over here, this is the split data module. You can go ahead and search for datasets and for modules over here itself.This particular module will also be available under Data Transformation. So over here in data transformation, I should be able to see the split data module. So I can see it over here. So we can go ahead and take the splitdata module and drag it onto the canvas. So now over here we are seeing the properties of this module. So on the canvas, there are two things that you can actually drag. First is your data set, and the other is the module. So why am I eying creating this? Because understanding what you can actually drag onto the designer as your machine learning is important for exams, Now this module has properties. Now over here, it's going to go ahead and split the rows of data. So, in terms of the splitting mode, over here. So remember, over here we are trying to split the rows of data over here.By default, it's going to go ahead and split 50% of the data. So it's going to split the data into two halves. But I want to go ahead and split the data into the percentages that we normally split into. So that's 70% for your training and 30% for testing. So let me go ahead and leave this as it is, and let me go ahead and hit close. If you go ahead and again click on the module, you can confirm whether the setting has been made. So for anything, if you want to see the properties of eitheryour data set or if you want to see the properties ofa module, you can just go ahead and click on it. and you can see the properties over here. Now I'm going to go ahead and join the data set. So you can just go ahead and click on the data set and join it to the split data. So over here, we are connecting the data set into the module in the designer over here. So currently, with no coding background and with no coding knowledge, this designer is going to go ahead and split our data set if we were using Python, right? So we would need to write Python code to go ahead and first import the data set. And if you want to have that visualisation of data, where you can see the mean, where you can see the minimum age, where they can see missing values, you have to code all of that. However, everything is done automatically for you in the machine learning designer. It's all built into the service. Even to split the data in Python, you need to add code. The code is still simple, but again, for a person who doesn't have coding experience, you can basically go ahead and use the machine learning designer over here. We haven't done much yet in terms of building the machine learning model, but I'd like to go over all of the important aspects of the designer. Now, we can just go ahead and click on "Submit" to build this machine learning pipeline. This machine learning pipeline is going to go ahead and actually split our data. So now over here, if you want to go ahead and run anything in the pipeline, you have to go ahead and create something known as a compute target. So over here I can click on "select compute target." So over here, we don't have any compute targets in place. We can actually go ahead and click on "Create New." So what is the point of this compute target? So all of the operations for this machine learning process need to run somewhere. So what this compute target is actually going to do is create an Azure virtual machine. So this will be a virtual machine on the cloud. This virtual machine will actually be used to run that underlying machine learning process. So, on that virtual machine, probably Python is going to be used to go ahead and split the data. So I said that instead of you writing code to split data, this machine learning design of this machine learning service is going to go ahead and split the data for you. But in the end, it still needs to go ahead and have an environment for splitting the data. So it's going to run code on this as your Virtual Machine in the background, and it's going to go ahead and split the data for you. So over here, we have to go ahead and set up something known as a compute target. Just for now, I'm going to pause this video. I'm going to go on Azure. So I'll go on to the next chapter. Let me just go ahead and create an Azure Virtual Machine. Now you're free to go ahead and skip this video, in which I'm going to go ahead and show you how to create an Azure virtual machine, so you can go on to the next video when we are expanding our pipeline in Azure machine learning. But for those who are not familiar with Azure, I want to go out and show you what an Azure Virtual Machine is.