IDS575_PS1_Q4

pdf

School

University of Illinois, Chicago *

*We aren’t endorsed by this school

Course

575

Subject

Mathematics

Date

Apr 3, 2024

Type

pdf

Pages

5

Uploaded by maabba

Report
4 Points -NN performs efficiently when there are only a few prediction queries. Q4 k-Nearest Neighborhood 36 Points Consider a binary classification problem with two real valued features and . Figures 1 and 2 illustrate two different training sets and . White circles indicate the positively-labeled examples, whereas black squares indicate the negatively-labeled examples. To classify a new instance point, we will use (unweighted) -Nearest Neighbors with Euclidean distance and different values. Thus the label for a new point will be predicted by the majority class (i.e., positive or negative) among the k closest examples around the query point. Q4.1 8 Points Draw the decision boundaries of and when . Q4.1.pdf Download k True False x 1 x 2 S 1 S 2 k k S 1 S 2 k = 1
1 of 1 Q4.2 5 Points Which label would you predict for the query points (3, 2) and (4, 2) given the decision boundary of in Q4.1? Ties must be broken toward predicting the positive class. (Auto: Your answer must be look like "negative, negative") negative, positive S 1
Q4.3 5 Points Which label would you predict for the query points (4.5, 4) and (4, 2) given the decision boundary of in Q4.1? Ties must be broken toward predicting the positive class. (Auto: Your answer must be look like "negative, negative") positive, positive Q4.4 10 Points When , a partition of spaces like the above is called the -th order Voronoi Diagram or Voronoi Tesselation. Try to draw the decision boundaries of when . (Hint: Try to draw every bisector between all pairs of positive and negative examples) Q4.4.pdf Download S 2 k > 1 k S 1 k = 3
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
  • Access to all documents
  • Unlimited textbook solutions
  • 24/7 expert homework help
1 of 1 Q4.5 8 Points If the -coordinate of four example points in are multiplied by 5, what would happen to its decision boundary when ? Could this effect cause problems when working with real data? Describe your idea: how to alleviate it. (Free Response) Multiplying only x2 coordinates with any scalar value (>1), significance of x2 coordinates will have larger impact as compare x 2 S 1 k = 1
to x1 coordinates. And so distribution of nearest point will change accordingly which will result in change in decision boundary. To eliminate it, we can standardize or bring all attributes on same scale which makes more sense in terms of end result on real data. GRADED Problem Set (PS) #01 STUDENT Urvashiben Patel TOTAL POINTS 93 / 100 pts QUESTION 1 Instance-based Learning 30 / 30 pts 1.1 (no title) 5 / 5 pts 1.2 (no title) 5 / 5 pts 1.3 (no title) 5 / 5 pts 1.4 (no title) 5 / 5 pts 1.5 (no title) 5 / 5 pts 1.6 (no title) 5 / 5 pts QUESTION 2 Mathematical Function 10 / 10 pts 2.1 (no title) 5 / 5 pts 2.2 (no title) 5 / 5 pts QUESTION 3 k-Nearest Neighborhood Basic 24 / 24 pts 3.1 Basic 5 / 5 pts