Databricks Machine Learning Associate Exam Questions

Databricks Certified Machine Learning Associate Exam Exam Objectives

1. 1

   1.0 Databricks Machine Learning (38%)

   • 1.1Identify the best practices of an MLOps strategy
   • 1.2Identify the advantages of using ML runtimes
   • 1.3Identify how AutoML facilitates model/feature selection.
   • 1.4Identify the advantages AutoML brings to the model development process
   • 1.5Identify the benefits of creating feature store tables at the account level in Unity Catalog in Databricks vs at the workspace level
   • 1.6Create a feature store table in Unity Catalog
   • 1.7Write data to a feature store table
   • 1.8Train a model with features from a feature store table.
   • 1.9Score a model using features from a feature store table.
   • 1.10Describe the differences between online and offline feature tables
   • 1.11Identify the best run using the MLflow Client API.
   • 1.12Manually log metrics, artifacts, and models in an MLflow Run.
   • 1.13Identify information available in the MLFlow UI
   • 1.14Register a model using the MLflow Client API in the Unity Catalog registry
   • 1.15Identify benefits of registering models in the Unity Catalog registry over the workspace registry
   • 1.16Identify scenarios where promoting code is preferred over promoting models and vice versa
   • 1.17Set or remove a tag for a model
   • 1.18Promote a challenger model to a champion model using aliases
2. 2

   2.0 ML Workflows (19%)

   • 2.1Compute summary statistics on a Spark DataFrame using .summary() or dbutils data summaries
   • 2.2Remove outliers from a Spark DataFrame based on standard deviation or IQR
   • 2.3Create visualizations for categorical or continuous features
   • 2.4Compare two categorical or two continuous features using the appropriate method
   • 2.5Compare and contrast imputing missing values with the mean or median or mode value
   • 2.6Impute missing values with the mode, mean, or median value
   • 2.7Use one-hot encoding for categorical features
   • 2.8Identify and explain the model types or data sets for which one-hot encoding is or is not appropriate.
   • 2.9Identify scenarios where log scale transformation is appropriate
3. 3

   3.0 Model Development (31%)

   • 3.1Use ML foundations to select the appropriate algorithm for a given model scenario
   • 3.2Identify methods to mitigate data imbalance in training data
   • 3.3Compare estimators and transformers
   • 3.4Develop a training pipeline
   • 3.5Use Hyperopt's fmin operation to tune a model's hyperparameters
   • 3.6Perform random or grid search or Bayesian search as a method for tuning hyperparameters.
   • 3.7Parallelize single node models for hyperparameter tuning
   • 3.8Describe the benefits and downsides of using cross-validation over a train-validation split.
   • 3.9Perform cross-validation as a part of model fitting.
   • 3.10Identify the number of models being trained in conjunction with a grid-search and cross-validation process.
   • 3.11Use common classification metrics: F1, Log Loss, ROC/AUC, etc
   • 3.12Use common regression metrics: RMSE, MAE, R-squared, etc.
   • 3.13Choose the most appropriate metric for a given scenario objective
   • 3.14Identify the need to exponentiate log-transformed variables before calculating evaluation metrics or interpreting predictions
   • 3.15Assess the impact of model complexity and the bias variance tradeoff on model performance
4. 4

   4.0 Model Deployment (12%)

   • 4.1Identify the differences and advantages of model serving approaches: batch, realtime, and streaming
   • 4.2Deploy a custom model to a model endpoint
   • 4.3Use pandas to perform batch inference
   • 4.4Identify how streaming inference is performed with Delta Live Tables
   • 4.5Deploy and query a model for realtime inference
   • 4.6Split data between endpoints for realtime interference