Rubric

Keep in mind that 25 students have already been assessed using this rubric. Changing it will affect their evaluations.
A3
A3
Criteria Ratings Pts
Exercise 1
Metropolis Hastings
threshold: pts
20 to >0.0 pts Full Proof Correct explanation of transition matrix Proper use of detailed balance to show reversibility Derivation of equations and logical flow Clear final conclusion connecting to stationary distribution
0 pts No Evidence
pts
20 pts
--
Exercise 2
MCMC Implementation
threshold: pts
20 pts Exceeds Proper implementation using chi-squared proposal (4) Correct results for Rayleigh (5) Correct results for Gamma (5) Histograms match expected distribution (6)
0 pts No Evidence
pts
20 pts
--
Exercise 3
LP Formulation of Norm Optimization
threshold: pts
20 to >0.0 pts Exceeds Correct transformation of L1 norm using auxiliary variables (6) Correct treatment of L∞ norm with max constraints (5) Matrix formulation of P, q, r (5) Explanation of equivalence and variable reduction (4)
0 pts No Evidence
pts
20 pts
--
Exercise 4
KKT Conditions
threshold: pts
20 to >0.0 pts Exceeds Optimization with Circle Constraint (10) Projection onto Unit Ball (10)
0 pts No Evidence
pts
20 pts
--
Exercise 5
EM Algorithm for GMM
threshold: pts
20 to >0.0 pts Exceeds Derivation E-step expression with posterior p̂i (4) M-step updates for all parameters (6) Implementation Generates synthetic data with seed (2) Converges to reasonable parameter estimates (4) Compares to ground truth, shows iterations (4)
0 pts No Evidence
pts
20 pts
--
Exercise 6 (Optional)
Alternating Minimization Robust Recovery (up to +20 pts)
threshold: pts
20 to >0.0 pts Exceeds Correct implementation of AM_RR (5) Working IHT and sparsity enforcement (5 pts) Demonstrates convergence and recovery accuracy (5 pts) Justified discussion on performance and limitations (5 pts)
0 pts No Evidence
pts
20 pts
--