CS1674: Homework 11 - Written

Due: 12/5/2016, 11:59pm

This assignment is worth 15 points.

  1. Say a vision system predicts the following scores for a cat image: 10 for the category "cat," 5 for the category "dog," and 3 for the category "cow." Another system predicts scores 8, 6 and 1 for the same categories. Is the softmax loss the same or different for these two systems? (You do not need to show calculations.)
  2. Briefly, how does gradient descent work?
  3. What is mini-batch?
  4. How can we prevent overfitting in a neural network?
  5. What do lower vs higher convolutional layers in a convolutional neural network capture?
  6. Why is it not necessary to have a massive amount of data to use convolutional neural networks?