15-883 Sample Final Exam Questions

These questions will give you an idea of the level of detail you can expect on the final exam. The actual questions will be more involved (have more parts to them), but they won't be more difficult than these samples. Only material covered since the midterm will be included on the final.

  1. Write down the equation for the Hebbian covariance learning rule. What is the relationship between Hebbian learning and principal components analysis? Explain Oja's contribution in this area.

  2. Describe four conditioning phenomena handled by the Rescorla-Wagner learning rule.

  3. Consider the temporal difference learning model. Write down the formula for V(t), the total discounted future reward, as a function of r(t) and γ. Then write down the formula for the the error signal δ(t).

  4. Draw a diagram of the connection patterns in the basal ganglia, including cortical projections, striatum, globus pallidus (two parts), subthalamic nucleus, and substantia nigra (two parts). Label each projection as excitatory, inhibitory, or modulatory. Where are the dopamine cells?

  5. Give a simple mathematical model of the retinal ganglion cell receptive field (1 sentence). Compare this with the receptive fields of "simple" cells in V1.

Dave Touretzky
Last modified: Tue Dec 6 02:00:58 EST 2011