I am sorry, in the first paraghraph I mean I call addPreference with -1, not removePreference (I just kick out whatever searches happened before the window). ... View more

I am sorry, I think I was unclear. For decaying old data I just call removePreference myself so there is no need for that to be done automatically by Oryx; that part is settled.    But in some cases that I was not able to reproduce consistently, I found that subsequent generations do not contain data from a generation earlier than model.generations.keep, so the generations are not really cummulative (I did read the thread you had with Jason Chen about generations from some while back, but I posted because I noticed this different behavior). It only happens once in a while and I don't know how to trigger this. I only got it a few times in the past few days (e.g. today I was not able to reproduce this at all). I will come back if I manage to produce some reliable steps that lead to this.   I do have a question on addPreference with negative values, what happens when we decreased enough to reach 0? I've tried that and the known items still remember these "nullified" items. recommend still returns results but they all have a strength of 0. Which I think is expected, as at this point we're supposed to know nothing about this user? For my particular use case, would be best if the items reaching 0 would move out of known items (and I can do that myself by calling removePreference) and, when all the items reached 0, the user would become unknown. Is there a method for this last part?    Thank you.       ... View more

Hi Sean   Just to let you know the outcome of this, all of my tests yesterday with Hadoop, with various parameters, on the one month of searches dataset, went on fine.   I will not continue testing this further on the whole big dataset, as for the moment it looks like Hadoop is out of the picture, since I managed to get hold of a machine with 512GB of RAM which prooved up to the challange of running Oryx in memory. The dataset is 421MB, with roughly 20 million records, and it took just a few minutes to go through 29 iterations, so well done! Seemed like a big portion of time was spent writing the model (this is an SSD machine). (I will continue further by looking at recommendations response times, how's that affected when I ingest users etc etc)   Thank you for the help with the bugs and all the explanations along the way.         ... View more

I understand now, thank you. I will be running tests all day today on the Hadoop version and come back if there are any issues.     I am looking forward to see how this will scale, as my whole dataset has around 20 milion records; at this time I can't try it out as I do not have a Hadoop cluster (the company I work for will be able to give me some VMs only towards the end of January). (... I might be able to run this sooner on a Google compute API, but it's not sure). ... View more

Great news, thank you. I will restart testing everything tomorrow.    In sort of related news - and if you've got time to explain - I've very briefly looked at the code changes but I still do not understand why this happens on Hadoop only? ... View more

You're welcome, and I should be thankful you're looking into this and that the project exists in the first place, and it's open source, and I don't have to write my own code for real time model updates.   If it's of any help, the dataset is weirdly shaped, pointed, if I can say that, because there is aggressive marketing around certain products and the majority of searches are centered around those. Users don't stay on the website too long and plenty of them have clicked on just one thing then left.   Another thing I have noticed, last week when trying to generate identical data with in memory and Hadoop, to see where things go different, by fixing the random generators seeds, was that I couldn't :). I presumed it was because of the in-proc vs. multiple processes execution - the Hadoop jobs (and presumed the random generators get 'reset' when launching a new process, if things like give me next random are used). Didn't dwell too much into it as you fixed the rank problems, but again, if it's of any help.    ... View more

(...wish there was an edit post button)   Conf settings for both:   model.test-set-fraction=0.25 model.features=6 model.lambda=1 model.alpha=30 model.iterations.max=60   Latest version of Oryx: I definitely have it because I wiped out the folder, cloned again and built. Also I am not getting the insufficient rank behavior anymore.    ... View more

So to summarize, earlier, when I checked the bug was fixed, I wanted to do it fast so ran the in memory and Hadoop computations on the "11" dataset, Hadoop converged faster but MAPs were similar so I though it's alright.   Then I started checking the lenghtier but closer to reality dataset, and that's when the difference became clear.    Just some example of runs with in memory: Converged at MAP score (at 10) 11 0.1058467372 27 0.1177788843 32 0.1187595734 18 0.1202960727 31 0.1206682346 26 0.1208719179 20 0.1209679965 21 0.1224116387   Hadoop: tried 3 so far and they all converged at 2 with 0.00x MAPE.  ... View more