Skip to content

Latest commit

 

History

History
84 lines (43 loc) · 5.23 KB

File metadata and controls

84 lines (43 loc) · 5.23 KB

Lab Overview

Azure Machine Learning designer (preview) gives you a cloud-based interactive, visual workspace that you can use to easily and quickly prep data, train and deploy machine learning models. It supports Azure Machine Learning compute, GPU or CPU. Machine Learning designer also supports publishing models as web services on Azure Kubernetes Service that can easily be consumed by other applications.

In this lab, we will be compare the performance of two different multiclass classification approaches: Two-Class Support Vector Machine used with One-vs-All Multiclass module vs Multiclass Decision Forest. We will apply the two approaches for the letter recognition problem and compare their performance. We will do all of this from the Azure Machine Learning designer without writing a single line of code.

Exercise 1: Create Training Pipeline

Task 1: Open Sample 12: Multiclass Classification - Letter Recognition

  1. In Azure portal, open the available machine learning workspace.

  2. Select Launch now under the Try the new Azure Machine Learning studio message.

    Launch Azure Machine Learning studio.

  3. When you first launch the studio, you may need to set the directory and subscription. If so, you will see this screen:

    Launch Azure Machine Learning studio.

    For the directory, select Udacity and for the subscription, select Azure Sponsorship. For the machine learning workspace, you may see multiple options listed. Select any of these (it doesn't matter which) and then click Get started.

  4. From the studio, select Designer, Show more samples.

    Image highlights the steps to open the sample pipeline, Sample 12: Multiclass Classification - Letter Recognition.

  5. Select Sample 12: Multiclass Classification - Letter Recognition.

    Image highlights the steps to open the sample pipeline, Sample 12: Multiclass Classification - Letter Recognition.

Task 2: Setup Compute Target

  1. In the settings panel on the right, select Select compute target.

    Image highlights the link to select to open the setup compute target editor.

  2. In the Set up compute target editor, select the available compute, and then select Save.

Note: If you are facing difficulties in accessing pop-up windows or buttons in the user interface, please refer to the Help section in the lab environment.

Image shows how to select the existing compute target named qs-compute.

Exercise 2: Submit Training Pipeline

Task 1: Create Experiment and Submit Pipeline

  1. Select Submit to open the Setup pipeline run editor.

    Image shows where to select the submit button to open the setup pipeline run editor.

    Please note that the button name in the UI is changed from Run to Submit.

  2. In the Setup pipeline run editor, select Experiment, Create new and provide New experiment name: letter-recognition, and then select Submit.

    Image shows how to provide the experiment name in the setup pipeline run editor and start the pipeline run.

  3. Wait for pipeline run to complete. It will take around 10 minutes to complete the run.

  4. While you wait for the model training to complete, you can learn more about the One-vs-All Multiclass module used in this lab by selecting One-vs-All Multiclass.

Exercise 3: Compare Model Performance

Task 1: Open Evaluation Results

  1. Select Evaluate Model, Outputs, Visualize to open the Evaluate Model result visualization dialog.

    Image shows how to open the evaluate model result visualization dialog.

Task 2: Compare Performance Metrics

  1. Select the regression performance metric Overall_Accuracy and compare performance of the two algorithms: Two-Class Support Vector Machine and Multiclass Decision Forest.

    Image shows the evaluate model result visualization dialog.

Task 3: Conclusion

  1. The Two-Class Support Vector Machine algorithm is extended for multiclass classification problem by using the One-vs-All Multiclass module.

  2. As you can observe that the native multiclass algorithm Multiclass Decision Forest outperforms the Two-Class Support Vector Machine across all key performance metrics.

  3. One recommendation for next steps is to increase the Number of iterations parameter for the Two-Class Support Vector Machine module to an higher value like 100 and observe its impact on the performance metrics.

Next Steps

Congratulations! You have trained and compared performance of two different multiclass classification machine learning models. You can continue to experiment in the environment but are free to close the lab environment tab and return to the Udacity portal to continue with the lesson.