MTH 448/563 Data-Oriented Computing

Fall 2018

Day 21 = Day -9, Election Day Tuesday, Nov 6

Don't forget to vote!


In one or two sentences (and a diagram if you wish), describe the feature vector used in LeafSnap (which I told you about in class last Thursday).

Don't do this on reports

Do not say something that contradicts the results you obtained.

Neural networks for object classification

Supplementary reference: CS231n, a superb course from Stanford. Example lecture

a fully connected layer

class scores

the goal


a good starting point?

a "loss" function


minimizing loss

gradient descent

Exercise: Use Inkscape or other program to draw a path of gradient descent from the tip of the arrow on this topographic map. Upload to UBlearns when you are done.