Check your understanding of this module before moving on.
?
0/3
Checkpoint quiz 3 questions
A quick check on this module's key ideas before you move on
Question 01
If the training loss is stuck at a high value and barely decreasing from the very first epoch, what is a common first thing to check?
A is correct. A loss that won't move from the start usually points to a basic setup issue: an inappropriate learning rate, a label/loss mismatch, or broken gradient flow (e.g. due to a bug or numerical instability), rather than a deep architectural problem.
Question 02
What does it typically mean if training loss decreases smoothly but validation loss is erratic or much higher throughout training?
A is correct. A persistent gap between training and validation performance points toward overfitting, or sometimes toward a mismatch between how the training and validation data are prepared (e.g. different normalization, leakage, or distribution differences).
Question 03
Why is it useful to first try overfitting a tiny model on a tiny subset of data (e.g. 10 examples) as a debugging step?
A is correct. A model with enough capacity should always be able to memorize a tiny dataset if everything (data, loss, gradients) is wired correctly. Failure to do so is a fast, cheap signal that something in the pipeline is broken, before spending time debugging at full scale.