Training history and visualization in TensorFlow - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is the purpose of the training history object in TensorFlow?

The training history object stores the values of metrics like loss and accuracy for each epoch during model training. It helps track how the model improves over time.

beginner

How can you access the training loss values after training a model in TensorFlow?

You can access the training loss values using history.history['loss'], where history is the object returned by model.fit().

intermediate

Why is it useful to plot training and validation accuracy during training?

Plotting training and validation accuracy helps you see if the model is learning well and if it is overfitting or underfitting by comparing performance on training and unseen data.

beginner

What Python library is commonly used to visualize training history in TensorFlow?

Matplotlib is commonly used to create plots of training and validation metrics over epochs.

intermediate

What does it mean if validation loss starts increasing while training loss keeps decreasing?

It usually means the model is overfitting: it learns the training data too well but performs worse on new data.

What does the 'history' object returned by model.fit() contain?

A The dataset used for training

B Only the final model weights

C Training and validation metrics for each epoch

D The model architecture

The history object contains the recorded metrics like loss and accuracy for each epoch during training.

Which method is used to train a model in TensorFlow and get the training history?

A model.fit()

B model.evaluate()

C model.predict()

D model.compile()

model.fit() trains the model and returns a history object with training metrics.

If training accuracy improves but validation accuracy stays the same, what might be happening?

A The model is underfitting

B The model is overfitting

C The model is perfectly trained

D The data is corrupted

This suggests overfitting: the model fits training data well but does not generalize.

Which Python library is best for plotting training history graphs?

A Matplotlib

B Pandas

C NumPy

D Scikit-learn

Matplotlib is widely used for creating plots and visualizations.

What does a decreasing training loss and increasing validation loss indicate?

A Data leakage

B Underfitting

C Good model generalization

D Overfitting

This pattern indicates overfitting, where the model fits training data too closely.

Explain how to use the training history object to visualize model performance over epochs.

Think about what data is stored in history and how plotting helps understand training.

Describe what it means when validation accuracy is lower than training accuracy and how visualization helps identify this.

Consider how training and validation metrics relate to model generalization.