STAT 503X Assignment 2 Recognizing Letters (10pts) Due: in class Feb 28 The objective is to identify each of a large number of black-and-white rectangular pixel displays as one of the 26 capital letters in the English alphabet. The character images were based on 20 dierent fonts and each letter within these 20 fonts was randomly distorted to produce a le of 20,000 unique stimuli. Each stimulus was converted into 16 primitive numerical attributes (statistical moments and edge counts) which were then scaled to t into a range of integer values from 0 through 15. We typically train on the rst 16000 items and then use the resulting model to predict the letter category for the remaining 4000. The number of instances (cases) is 20000, and there are 17 attributes (variables): 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. lettr x-box y-box width high onpix x-bar y-bar x2bar y2bar xybar x2ybr xy2br x-ege xegvy y-ege yegvx capital letter (26 values from horizontal position of box vertical position of box width of box height of box total # on pixels mean x of on pixels in box mean y of on pixels in box mean x variance mean y variance mean x y correlation mean of x * x * y mean of x * y * y mean edge count left to right correlation of x-ege with y mean edge count bottom to top correlation of y-ege with x 1 to 26) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) (integer) The numbers in each class are as follows: 789 A 734 H 766 B 755 I 736 C 747 J 1 805 D 739 K 768 E 761 L 775 F 792 M 773 G 783 N 753 O 764 V 803 P 752 W 783 Q 787 X 758 R 786 Y 748 S 734 Z 796 T 813 U This data has been broken into 3 groups, A-I, J-R, S-Z, each about 6500 long. Each working group will be given one of these data sets to classify. The XGobi-ready les on Vincent are (in /home/dicook/data/): letter-atoi.dat, .col, .row letter-jtor.dat, .col, .row letter-stoz.dat, .col, .row This data is from the UCI machine learning database: . ftp://ftp.ics.uci.edu/pub/machine-learning-databases/letter-recognition/ -- Creator: David J. Slate -- Odesta Corporation; 1890 Maple Ave; Suite 115; Evanston, IL 60201 -- Donor: David J. Slate (dave@math.nwu.edu) (708) 491-3867 -- Date: January, 1991 ||||||||{ Some hints on working with this data. First step is to break it into training and test data sets. Use graphics to get a rough break out of important variables for dierent letters, then use various numerical classiers to nd good classication schemes. The grade for the assignment will depend on neatness of the report, how comprehensive is the analysis (are all possible questions answered), clarity of results presentation, and preciseness of nal conclusions. 2