Today among the three top players -those in the money- at the Higgs challenge we see the appearance of Lubos Motl, whom I had signalled as a participant in an earlier posting. We all know that Lubos is a smart guy, but I doubted whether he would take this very seriously. However, it seems he is. As we speak he has submitted almost 100 solutions (you can submit up to 5 solutions per day, so that means having worked at this at least 20 days in a row).
In the clip below you see the top standers from the challenge site's leaderboard:
So Lubos is currently third, and at this point he looks like one of the likely winners - but let's not forget that there are still three months to go, and some of us are keeping our solutions in the pocket rather than posting them on the site; this is in fact a bit like a poker game, where you don't want to show your hand too quickly.
In my spare time I am also working at a classification algorithm of my own - something entirely written by myself, unlike the tools that most participants are declaredly using. I am getting good results, but I have so far only ran tests on a fraction of the total training dataset - which decreases the machine's ability to classify at its best signal and background events.
Despite the fact that I am running on less than a tenth of the training sample, here is what I have been getting as of late: the screen capture below shows the output of my program this afternoon (below the black screen with my program's output you can read the leaderboard page posted above). To give you a hint, the Score in the challenge's page is called "AMS".
So as you see, the last three months of the challenge may reserve surprises yet!