Problem Sheet 9#
Question 1#
a. Prove the following theorem on the convergence of gradient descent for convex and smooth functions:
Let
converges to the global minimum of
b. Explain the statement and relevance of the above theorem.
c. Outline in your own words the proof of the above theorem.
d. Outline possible strengths, weakness and pitfalls with the implementation of gradient descent for finding a minimum of a function.
Question 2#
a. In your own words explain the statement and briefly outline the proof of the Universal Approximation Theorem for a sigmoid function given by:
Let
Let
are dense in
for all
b. Explain the relevance of the Universal Approximation Theorem to Artificial Neural Networks.