Microsoft DP-100 Exam Dumps

Pearson's correlation statistic, or Pearson's correlation coefficient, is also known in statistical models as the r value. For any two variables, it returns a value that indicates the strength of the correlation

Pearson's correlation coefficient is the test statistics that measures the statistical relationship, or association, between two continuous variables. It is known as the best method of measuring the association between variables of interest because it is based on the method of covariance. It gives information about the magnitude of the association, or correlation, as well as the direction of the relationship.

https://docs.microsoft.com/en-us/azure/machine-learning/studio-module-reference/filter-based-feature-selection

https://www.statisticssolutions.com/pearsons-correlation-coefficient/

When you need to create a custom model flavor for a model that:

<ul><li>Is not natively supported by MLflow</li><li>Cannot be serialized in Pickle format</li><li>Has complex source code</li><li>Requires its Python library to be packaged with the model</li></ul>

<strong>MLflow.pyfunc (Python Function flavor)</strong> is the correct choice. It allows you to define a custom <code>PythonModel</code> class with <code>load_context()</code> and <code>predict()</code> methods, enabling you to:

<ul><li>Package custom dependencies</li><li>Implement custom serialization logic</li><li>Handle complex model requirements</li><li>Bundle the model source code and libraries together</li></ul>

This provides the flexibility needed for non-standard models while maintaining MLflow compatibility.

For image classification in automated machine learning, <strong>accuracy</strong> is the primary metric to optimize. Accuracy measures the proportion of correct predictions (both true positives and true negatives) among all predictions, making it the standard choice for multi-class classification tasks like image classification.

For a training job in Azure Machine Learning that requires scaling to handle larger datasets with multi-node cluster support:

<ul><li><strong>Azure Machine Learning Compute Cluster:</strong> This is the correct choice as it supports automatic scaling across multiple nodes. It can dynamically adjust the number of compute nodes based on job requirements.</li></ul>

<strong>Why not other options:</strong>

<ul><li>Compute Instance: Single-node only, cannot scale horizontally.</li><li>Kubernetes cluster: Requires manual setup and configuration.</li><li>Serverless compute: While available, it doesn't provide the same multi-node cluster control as Compute Cluster.</li></ul>

Compute Cluster is purpose-built for distributed training jobs that need automatic scaling.

For minimal cost compute targets for model inference/testing in Azure Machine Learning:

<ul><li><strong>Azure Container Instances (ACI)</strong> - Provides serverless containerized compute, ideal for low-traffic inference workloads with minimal management overhead and pay-per-use pricing</li><li><strong>Azure Functions</strong> - Enables serverless compute execution with automatic scaling and consumption-based pricing, suitable for lightweight inference tasks</li></ul>

Both options minimize operational costs compared to persistent compute clusters. AKS would be overkill for testing, and Compute Clusters are for training rather than inference cost minimization.