TopAceEssays

Part I is split into two parts as Part I.B is a short composition.
Part I.A (40 points each)

1. A child’s game has a spinner with three outcomes: red, blue, and
   yellow. Suppose that we vary the percentage of the area under
   the blue spinner, ranging from 1/6 to 5/6, with the other colors
   sharing whatever probability is left over equally. Compute the
   entropy for each configuration and create a plot of entropy as a function of the
   probability that the outcome is blue. If you write a program to do this, show your
   source code (your program may not call library functions that compute entropy or
   information). If you do it by hand, show your work. Where is the entropy
   maximized and why?
2. A toy machine learning algorithm has 10 examples (0, 1, 2, …, 9). Create a table
   of train and test splits for a 5-fold cross validation.
3. A 3×3 convolutional kernel has weights:
   [
   1 2 2
   0 0 2
   0 0 1
   ]. Assuming a stride of 2 in either
   direction, compute the top-left entry of the
   convolutional output and the subsequent
   convolution to the right and below. Assume
   the intensity image of the 8 shown here.
   The target area for the top-left convolution
   is shown in red.
   CS 450 ARTIFICIAL INTELLIGENCE
   PROFESSOR YU, COURTESY OF PROFESSOR ROCH
   2
4. Given the two-feature data set below where two classes are distinguished by
   color,
   would a single neuron neural net be more likely to have high bias or high variance?
   Justify your answer.
5. In the slides, we updated the weight for w1 in a toy neural network (slides 14-18).
   Showing your work, use gradient descent to update the weight w2 with a learning
   rate ε=.01 for the same network and show how the loss is reduced.
   Part I.B
   This short writing assignment (60 points) is to provide you with an opportunity to
   practice written communication for different audiences, something that people in industry
   and academia are frequently expected to do. In it, you will write three short singlespaced paragraphs. In each of them, you will explain a decision tree learner to a different
   audience. The paragraphs are limited to one-half page each (standard 8.5 x 11” page size
   and margins with 12 point type). Do not turn in more than one and a half pages, as no
   credit will be given if you exceed the allotted space. Part of good communication is
   learning to be concise.
6. The first paragraph is to be directed to a professional colleague who knows
   very little about computing in general and AI in particular. Describe to her
   what a decision tree learner is and what it is intended to do. Make sure to use
   the language she can understand or define the required terms for her.
7. The second paragraph is directed to a colleague who is a physician. He knows
   a bit more about computing than your first colleague. Explain to him what a
   f2
   f1
   f2
   CS 450 ARTIFICIAL INTELLIGENCE
   PROFESSOR YU, COURTESY OF PROFESSOR ROCH
   3
   decision tree is and how it can be used in the context of medicine. Go into as
   much detail as you can in the allotted space and use relevant terms from
   science and computer science.
8. The third paragraph is directed to a colleague who is a computer scientist and
   has a Ph.D. in AI. Explain to her the issues that you’re having implementing
   your decision tree system. Get into the technical details of a made-up scenario
   about your system.

Get Homework Help Now