Mark Hammond wrote: >> That thread was interesting, but still runs under the assumption that >> balanced training is the ideal. >> > > I read that thread as *demonstrating* why unbalanced training will skew your > results. It makes no assumptions at all, but simply considers the facts > We are suffering a semantic disagreement here. That thread explains how an imbalance influences token scores, but doesn't explain one way or another if that harms classification in the wild. Presumably at some point it will cause harm, but just when is unclear. > about how spambayes works and the math behind it. The assumptions you refer > to are a direct result of the facts presented there. > > Do you disagree with the analysis of the math in that thread? If you don't > disagree, then I completely miss your point. > > >>> Google also shows anecdotal reports of poor results after an imbalance as >>> low as 2:1, so I don't think it would be responsible to re-evaluate that >>> Anecdotal, as you also point out is not the same as proof. >> "responsible"? I'm not sure what you mean. >> > > responsible: worthy of or requiring trust. > > In my opinion, it would be irresponsible to our users, who generally trust > I was assuming that you didn't mean that, since in no way was I suggesting changing anything at all. At least not until the facts behind the statement had been reevaluated and proven to be incorrect. > the spambayes developers, for us to give out information that current wisdom > says to be incorrect, especially when backed up by a solid theoretical > understanding of why that wisdom exists. It would be irresponsible for us > to change our current wisdom based on anecdotes of a single individual, > especially when opposite anecdotes can be easily found. > I really don't know what this was in response to. I ask a question and you are replying as if I am demanding change. That has never been the spambayes way. > >> I assumed that running a test to evaluate the effects of imbalance would >> be the way to generate or refute such evidence? >> > > One person running a test is unlikely to cut it. If you design a test, you > may have luck getting others to run it against their email, in which case > the results will start to get interesting as the number of people increase. > Again, I never claimed that one person would "cut anything". However, if one person doesn't start carry the effort forward, it will never get reevaluated. If my efforts at building a test lead to a result confirming the status quo, then nothing more need be said. At least until some other wise guy comes along and challenges it :) Since the nature of Spam changes over time, the results of these challenges may also change? > >> When I get back from Hawaii, I think I'll dust off the old test corpus and >> > try > >> some tests. If anybody else has some test results, I'd be very interested >> > in > >> seeing them. >> > > Google is your friend here - you can find many discussions about the effects > of imbalances, and plenty of discussions about why a single test from a > single user isn't a useful indicator of anything. Searching for anything > Tim Peters has to say would be the most productive thing to do :) > That is a good idea, since gathering info from some of the core folks would seem to save time in developing an effective test. >> My current thought is that getting a (very) large mount of spam with >> very few clues results in each email results in the imbalance. I've >> just checked some of todays spam and some had as few as 31 clues. With >> so few clues, it is relatively easy for a spam message to end up with an >> unsure or even ham classification while the most ham is being correctly >> classified. The alternative to an imbalanced training set is to find an >> easy way to train on extra ham, but only the ham that still has some >> classification value to add. >> > > I'm glad that spambayes appears to work well for you with a significant > imbalance, but I think we've already pointed out that there is solid > reasoning behind our position. > As a long time spambayes user, promoter and sometimes contributer, I've yet to find a user for whom spambayes does not work very well! I understand the theory but would like to understand the real world implications, especially given the anecdotal evidence that imbalance doesn't always harm performance. Even Tim Peters has a 3:1 ratio :)
Anyway, I'm not trying to rock the boat, start a fight or anything else. I've now got to go and pack, Brendon. _______________________________________________ SpamBayes@python.org http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
