ERRATA for High Performance Discovery in Time Series

Ordered by appearance in the text. Last updated 4/1/06.
Send additional errors and comments to: shasha@cs.nyu.edu .
Page 26, second to the last paragraph.
Change d(x,x_tilde) = ||x - x_tilde||^2 to d(x,x_tilde) = ||x - x_tilde|| .
Reported 3/30/06 by Lee Rhodes.
Page 42, Definition 2.37.
The last line should be the series (x(2),x(4),...,x(n)).
Reported 3/30/06 by Lee Rhodes.
Page 44, equation (2.111).
The DB2 filters should have + sign before \sqrt 3 in the third number of d_n^L,r_n^H and the second number of d_n^H and r_n^L.
Reported 2/2/06 by Lee Rhodes.
Page 92, line 13.
Change 2n to 2\sqrt n.
Reported 3/28/06 by Lee Rhodes.
Page 92, Definition 4.6.
Note: The Pearson Correlation Coefficient only holds if either a) n is large and the sample standard deviation (divided by n-1) is used or b) if the population standard deviation (divided by n) is used.
Reported 3/28/06 by Lee Rhodes.
Page 93, 2nd paragraph.
Last sentence should read: "A normalized Euclidean distance of \sqrt {2n} corresponds to 0 correlation. Also normalized Euclidean distances larger than \sqrt {2n} give negative correlation, with normalized (squared) Euclidean distances of 2\sqrt n corresponding to the perfect anti-correlation -1."
Reported 3/28/06 by Lee Rhodes.
Page 155, Figure 7.2.
Change For j=1 TO size(b)-1 STEP 2 to For j=1 TO size(b)-1.
Reported 10/13/06 by Henrique Andrade.