tag:blogger.com,1999:blog-193178369124442285.post9096057236744370989..comments2022-11-12T23:40:40.957-08:00Comments on Jessica Kirkpatrick Research: Comparing Correlation Functionberkeleyjesshttp://www.blogger.com/profile/00255252868558666987noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-193178369124442285.post-83920344334713057472010-04-06T20:22:18.504-07:002010-04-06T20:22:18.504-07:00So the default depth of the tree is depth=10. It ...So the default depth of the tree is depth=10. It seems that a lower depth, and thus lower resolution, is better for these very large 10-degree search angles. Here are some timing numbers that show depth=6 is best. You would do this to set your depth:<br /><br /> h=esutil.htm.HTM(depth)<br /><br />Still, there is not much speedup. When we were talking at the meeting, you had<br />said it took "days" to compute these, so maybe we were just talking about different regimes. Also<br />I didn't realize you were working at 10 degree angles. Your best speedup might be to work<br />with physical scales so you optimize your search<br />radius. This is easy to do with the bincount()<br />code.<br /><br />depth seconds<br />3 137.153673172<br />4 114.97971487<br />5 104.600896835<br />6 99.5818519592<br />7 102.73042202<br />8 119.931453943<br />9 185.077499866<br />10 460.641659021<br />11 1502.35429096Erin Sheldonhttps://www.blogger.com/profile/06086920650824088529noreply@blogger.com