<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Manthan Thakker on Medium]]></title>
        <description><![CDATA[Stories by Manthan Thakker on Medium]]></description>
        <link>https://medium.com/@manthanthakkar?source=rss-5d55f80febb8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*iRzEcnAVRmD05Ui8.</url>
            <title>Stories by Manthan Thakker on Medium</title>
            <link>https://medium.com/@manthanthakkar?source=rss-5d55f80febb8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 17:47:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@manthanthakkar/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Azure-Automated Machine Learning Service — Review]]></title>
            <link>https://medium.com/@manthanthakkar/azure-automated-machine-learning-service-review-a38c7df156a2?source=rss-5d55f80febb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/a38c7df156a2</guid>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Manthan Thakker]]></dc:creator>
            <pubDate>Wed, 12 Jun 2019 00:23:50 GMT</pubDate>
            <atom:updated>2019-06-12T00:23:50.131Z</atom:updated>
            <content:encoded><![CDATA[<h3><em>Azure-Automated Machine Learning Service — Review</em></h3><p><em>Machine Learning models as web services in minutes</em></p><p>Note: This blog is not affiliated to any organization and is solely my personal review.</p><blockquote>Have you ever wanted to build a simple machine learning model quickly to perform any basic task such as forecasting or classification?</blockquote><blockquote>Have you ever struggled in selecting which machine learning model to choose?</blockquote><blockquote>Have you ever struggled in how to tune your hyper-parameters for a model and compare performance?</blockquote><blockquote><em>Have you ever struggled in deploying a model as a web service?</em></blockquote><p>If your answer to any of the above questions is yes, then you should read this post.</p><h4>A typical workflow in building a Machine Learning Model</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*gKHdKCOjrlx3XSlDRkMQIg.png" /></figure><h3><strong>What is Automated Machine Leaning?</strong></h3><p><strong><em>Automated Machine Learning strives to target automating:</em></strong></p><ul><li><strong><em>Feature engineering</em></strong></li><li><strong><em>Model selection and hyperparameter tuning</em></strong></li><li><strong><em>automated deployment</em></strong></li></ul><h3><strong>Why use Automated Machine Learning?</strong></h3><ol><li><strong>I am New to Machine Learning and Programming:</strong> This is a great tool for you as you need to know little about coding machine learning models to leverage Machine Learning. It will be a great starting point for anyone who wants to understand Machine Learning basics.</li><li><strong>I am a Software Engineer:</strong> You are an expert in Python and you might want to check out the Python SDK available which is very flexible to accommodate most of the use cases.</li><li><strong>I am Data Scientist:</strong> You can use this tool to fine-tune the hyperparameters or even run initial analysis to compare the performance of a wide range of algorithms with little effort</li></ol><p>While there are many tools that allow you to automate Machine learning today, I will be going over Azure’s automated machine learning service that was announced recently. It took me less than 60 minutes to train, test and deploy the model as a web service. The sheer speed from a machine learning idea to a full-fledged consumable API is the inspiration to write this blog.</p><p><strong>Pre-requisites:</strong> You need to have an <a href="https://docs.microsoft.com/en-us/learn/modules/create-an-azure-account/">Azure account</a> and <a href="https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-manage-workspace">create a machine learning workspace</a>.</p><h3>1. Data Preparation &amp; Experiment Configuration</h3><p><strong>1.1:</strong> Download the famous <a href="https://archive.ics.uci.edu/ml/machine-learning-databases/housing/">Boston housing dataset</a> from the UCI repository and upload it to the blob attached to the machine learning workspace. The Boston housing dataset has 13 attributes to predict the price of the house depending based on attributes such as crime rate, tax rate, the average number of rooms per dwelling, etc. It is a typical regression problem in Machine Learning. We will use Automated Machine Learning service by azure to predict the house prices. Check out this link for more details about the <a href="https://archive.ics.uci.edu/ml/machine-learning-databases/housing/">dataset</a>.</p><p><strong>1.2:</strong> Go ahead and create a new experiment and create a compute if not created.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6fGBTpwZLQ-2XiThDco1hw.png" /><figcaption>1.2 Create an new experiment &amp; compute</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IxeNq-W7QyZ-GJDoT2bbAg.jpeg" /><figcaption>1.3 Advanced settings — Primary metric, exit criteria and number of cores</figcaption></figure><p><strong>1.3:</strong> Advanced Settings</p><p>You can configure the experiment to block certain algorithms</p><blockquote>Primary metric: Main metric used for scoring your model.</blockquote><blockquote>Exit conditions: <em>Training job time,</em> <em>Max number of iterations etc</em></blockquote><blockquote>Max cores per iteration: Select the multi-core limits you would like to use when using multi-core compute</blockquote><p>You can see the full list <a href="https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-create-portal-experiments"><em>here</em></a></p><h3>2. Model Selection &amp; Analysis</h3><p><strong>2.1</strong>: Start the experiment. Once the experiment completes, you will be able to see all the algorithms and preprocessing strategies tried by autoML with their corresponding metrics. It took 10 minutes to train and validate 24 combinations of models.</p><p>A collection of several models working together on a single dataset is called an Ensemble. For our case, the Voting ensemble yielded out really good results. Let’s deep dive into seeing precise results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xR978wa5k3Xet-mtx5jigw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8sJQvUQcZvvIMqbYMq9c7w.jpeg" /></figure><p>Predicted vs True value plots the value predicted by the model once trained vs actual value in the dataset. The mean absolute error is 2.21 and median around 1.6 with R2 score of 0.87 which is decent.</p><h3><strong>3. Deploy model</strong></h3><p>3.1 After analysis, we select the voting ensemble method and decide to deploy the model. Before deploying we need to first register the model with Azure. Once registered, your model will be saved in the azure workspace and you will be able to create a docker image from it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jhmuzKD8Wie3me0bJ5hgYA.jpeg" /></figure><p>3.2 To view all the registered models navigate to the model&#39;s section. Create an image by simply selecting the model, upload the scoring and environment files. The scoring file is required so as it can create a web service to generate predictions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jvjRfmRiBz9ch-vBwoXshA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vsH2lAnBxCreAJCLwi5H-w.jpeg" /></figure><p>3.3 Once the image is created we are ready to deploy the image to either one: Azure container instance(ACI) or Kubernetes services(AKS)</p><p>Head to the images section and deploy the image. I am testing and don&#39;t need to scale so I am going to select ACI.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OUw9Ifk7udA0jTtzhvNKkQ.jpeg" /></figure><p>Grab the scoring URL /endpoint from the deployment, and it&#39;s ready to do predictions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9xcpWZypgBSSXlMwON11rQ.png" /></figure><h3><strong>4. Consume REST Endpoint</strong></h3><p>Any application (React/Mobile application/Data Pipeline) will now be able to consume the predictions by simply consuming the REST endpoint.</p><p>Here is a sample:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1003/1*1BuvZ__-JYd1V9pyQFt_Ug.jpeg" /></figure><h3><strong>Final Thoughts</strong></h3><p>In a nutshell, I see Azure Automated Machine learning service as:</p><ol><li>A great tool for automating the entire Machine learning workflow which takes days</li><li>Helps people with different skillsets to leverage machine learning and use them not only professionally in products but personally too.</li><li>Provides a platform for data scientists to leverage this for initial analysis</li><li>Reduces the time for experimentation to delivery for typical machine learning tasks.</li></ol><p>Note: To all my fellow Software Developers, check out this comprehensive Python<a href="https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-configure-auto-train"> SDK</a> to leverage this in your software products and create cool applications.</p><h3><strong>References</strong></h3><p>Azure Documentation.</p><p><a href="https://docs.microsoft.com/en-us/azure/machine-learning/service/concept-automated-ml">https://docs.microsoft.com/en-us/azure/active-directory/managed-identities-azure-resources/qs-configure-portal-windows-vm</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a38c7df156a2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>