Retraining strategies in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the main goal of retraining a machine learning model?

The main goal is to update the model with new data so it stays accurate and adapts to changes over time.

intermediate

What is incremental retraining?

Incremental retraining means updating the model by training it on new data without starting from scratch, saving time and resources.

beginner

Why is it important to monitor model performance before retraining?

Monitoring helps detect when the model's accuracy drops, signaling that retraining is needed to maintain good predictions.

intermediate

What is the difference between full retraining and incremental retraining?

Full retraining trains the model from scratch using all data, while incremental retraining updates the existing model using only new data.

intermediate

Name one challenge of retraining machine learning models.

One challenge is avoiding overfitting to new data, which can reduce the model's ability to generalize to unseen data.

What triggers the need for retraining a machine learning model?

A When model performance decreases

B When the model is first created

C When the training data is perfect

D When the model is deployed

Retraining is needed when the model's performance drops, indicating it no longer fits the data well.

Which retraining strategy updates the model using only new data?

A Batch training

B Full retraining

C Transfer learning

D Incremental retraining

Incremental retraining updates the existing model with new data without retraining from scratch.

What is a risk of retraining a model too often?

A Model becomes outdated

B Model size decreases

C Model overfits to recent data

D Model stops learning

Frequent retraining can cause the model to overfit recent data and lose generalization.

Which of these is NOT a common retraining trigger?

A Model accuracy improves

B Data distribution changes

C New data becomes available

D Performance metrics degrade

Improved accuracy does not trigger retraining; retraining is triggered by performance drops or data changes.

What does full retraining involve?

A Training only on new data

B Training from scratch on all data

C Updating model weights manually

D Using a smaller dataset

Full retraining means training the model from scratch using the entire dataset.

Explain why and when a machine learning model should be retrained.

Think about what happens when the model sees new or different data.

Describe the difference between full retraining and incremental retraining.

Consider how much data is used and how long retraining takes.