Knowledge check

Quiz

In Progress

Knowledge check

Evaluate your understanding of this unit by completing the knowledge check below.

What is one of the main problems caused by high-dimensional data?

It speeds up model training significantly.It reduces the chance of model overfitting.It increases computational complexity and the risk of overfitting.It guarantees better model generalisation.

Which method directly reduces dimensionality by projecting data into a smaller number of principal components?

Filter methodsPrincipal Component Analysis (PCA)Wrapper methodsGenetic algorithms

Featuretools and tsfresh are examples of tools primarily used for:

Visualising large datasetsReal-time feature servingAutomated feature generationFeature drift monitoring

In health care data science, comorbidity indices like Charlson are used to:

Generate synthetic patient dataEncode multiple chronic conditions into a single risk measureNormalise feature scales across hospitalsCorrect missing values in vital signs

Which technique allows a machine learning model to learn both features and predictions together?

AutoencodersPCAEnd-to-End Feature LearningWrapper Methods

Which advantage does using Domain-Specific Embeddings offer when engineering features?

Forces all entities into a one-hot encoding.Compresses entity information into dense vectors, improving scalability.Makes models slower but more accurate.Reduces the need for feature drift monitoring.

Why is monitoring feature drift critical in production ML systems?

It helps models predict future values more accurately.It identifies when feature distributions shift, risking model degradation.It guarantees better model generalisation by itself.It completely eliminates the need for retraining.

Which strategy improves the automation and efficiency of a feature engineering pipeline?

Using manually curated features to maintain full control over the pipeline.Avoiding feature selection to retain all raw data.Implementing automated feature generation tools like Featuretools and tsfresh.Recomputing all features from scratch before every model training cycle.

Which factor is critical when designing scalable feature pipelines for machine learning applications?

Using modular components that allow independent updates without disrupting the pipeline.Relying exclusively on batch processing for feature updates.Avoiding automation to maintain full customisation in feature engineering.Ensuring feature extraction happens manually for better control.

t-SNE and UMAP are primarily used for:

Reducing computational complexity of training datasets.Creating end-to-end learnt features in deep learning pipelines.Visualising high-dimensional data in lower dimensions.Encoding categorical variables for tabular ML.

In progress