This preview shows page 1. Sign up to view the full content.
Unformatted text preview: (Computer Exercise)
Faculty of Computers and Information Faculty Department of Computer Science Pattern Recognition A complete reallife problem on linear regression: A reproduction to the prostate cancer example.
• Download the prostate cancer dataset from the book website. • Standardize the data using two diﬀerent methods (centering followed by sample variance, and minmax to scale it to [−1, 1]) and do a scatter plot for each. • Devide the data, randomly, to two and onethirds (for training and testing). Build three full linear models, one on the original data and two on the two versions of standardization above. Make sure that your prediction vector on the testing set is identical for the three models. What is the estimated error errtr ? • Remodel the problem using all the variables but with regularization using rdige regression (use the centered model). Use the testing set to select the best λ by plotting errtr (λ) vs. λ. Hint: for just one arbitrary value of λ, use the noncentered model with regularization (check the lecture notes) and make sure that the prediction is the same as the centered model. • Use the centered model, the twothirds of the dataset (as a training set), and 10K CV to choose the best λ (and accordingly the best β ). Train youTest your selected model on the testing set (the onethird of the dataset) to ﬁnd a ﬁnal estimate of err ...
View
Full
Document
This note was uploaded on 01/06/2011 for the course IT 342 taught by Professor Waleeda.yousef during the Spring '10 term at Helwan University, Helwan.
 Spring '10
 WaleedA.Yousef
 Computer Science

Click to edit the document details