DSA Studio

Optimization Checkpoint

Reconstruct a new trace and diagnose implementation mistakes. Select one answer per multiple-choice question.

Get every question correct to pass this checkpoint. You can review feedback, revise your answers, and submit again.

1. For f(x)=x^2, start at x=3 and use lr=0.25. What is x after one gradient-descent update? Enter a decimal.
2. Three optimizer steps produce how many parameter-history entries in this exercise?

Select one answer.

3. For f(x)=x^2 and lr=1, what happens from x=2 under plain gradient descent?

Select one answer.

4. Momentum uses v_next=0.5*v+g. If v=4, g=2, x=1, and lr=0.1, what is x_next? Enter a decimal.
5. Linear regression trains for one epoch and returns a single loss. Which parameters produced that recorded loss?

Select one answer.

6. A learner differentiates mean squared error but omits its factor of two. Which independent check most directly detects the issue?

Select one answer.

Join Discord