Don’t forget to measure time…
I just did a batch of test runs of my NMF code, and I noticed that I forgot to capture the elapsed execution time in the code for the projected gradient approach. I also forgot to use the time command in my bash test scripts. Which means that the 3 hours of testing produced exactly zero time measurements for me on one of the test machines. And this si essentially all I really cared about at the moment. Sigh…
Oh well, at least I got some data for convergence graphs out of this. Remember kids - if your researching new ways to optimize old algorithms, it is usually a good idea to try to capture the execution time of your test runs. 
I can’t wait to be done with this.
Related Posts:
