So there are two ways we can feed that value according to users preferences
2020-03-28 08847, 2020
_lucifer
gr0uch0mars: done. submitted the final proposal
2020-03-28 08850, 2020
jwf
iliekcomputers / ruaok: I just tried to delete a lot of my listens all at once, but looks like LB timed out. Not sure if it's something you want to look into from logs
2020-03-28 08802, 2020
jwf
Not sure if I should try again or if there is a bigger bug to report
2020-03-28 08810, 2020
iliekcomputers
jwf: ouch
2020-03-28 08817, 2020
iliekcomputers
can you open a ticket?
2020-03-28 08823, 2020
iliekcomputers
shivam-kapila: ^
2020-03-28 08823, 2020
jwf
504 Gateway Timeout
2020-03-28 08825, 2020
jwf
Yep, sure
2020-03-28 08835, 2020
iliekcomputers
thanks jwf !
2020-03-28 08801, 2020
iliekcomputers
shivam-kapila: can you take a look at jwf's problem first?
2020-03-28 08812, 2020
shivam-kapila
Sure
2020-03-28 08814, 2020
iliekcomputers
the pagination thing can be done later anyways
2020-03-28 08843, 2020
iliekcomputers
shivam-kapila: i think the delete-listens function might be timing out for big measurements
That way we can change it as we come across problems. Keep it an year in the beginning and we'll see if it doesn't work, change the time interval to something shorter
2020-03-28 08836, 2020
shivam-kapila
Okay
2020-03-28 08857, 2020
iliekcomputers
Thx!
2020-03-28 08811, 2020
iliekcomputers
I'm off for the day. See ya tomorrow, everyone! 👋🏼
iliekcomputers: When you are back please look at this. I guess we shouldn't iterate upto 50 years. How about deleting say 10000 listens ib batches. l will look if its possible in influx. Its possible in SQL though.
shivam-kapila: just do it until 2002 (the year when last.fm started and our oldest source of listens) and then drop the measurement
2020-03-28 08838, 2020
shivam-kapila
Ok ok. Thanks for replying even after day off :)
2020-03-28 08804, 2020
scientist1642 joined the channel
2020-03-28 08810, 2020
scientist1642
Hey Hey, is there a separate channel for discussing GSOC ideas?
2020-03-28 08842, 2020
yvanzo
hi scientist1642, not really, this is the channel.
2020-03-28 08817, 2020
apetresc joined the channel
2020-03-28 08853, 2020
thomasross joined the channel
2020-03-28 08841, 2020
scientist1642
yvanzo Good! I am a bit of late to the party and sprinting through organizations. :] I saw this talk a few weeks ago, and it was really fascinating https://www.youtube.com/watch?v=FD5ZKi-moMU
2020-03-28 08832, 2020
scientist1642
Don't know how much work is done in ML community for tagging emotions on music tracks, but I see Moods field in acusticbrainz with probabilities. Could anyone point out how acousticbrainz currently does this?
2020-03-28 08810, 2020
heyarne[m] joined the channel
2020-03-28 08822, 2020
heyarne[m]
hi! i have a collection of a bit more than 10000 songs. i'd like to set up acousticbrainz so i can query information on all of them. how do i actually generate the dataset? the acousticbrainz site only has info on how to import one and use the api to query it, as far as i can see
2020-03-28 08859, 2020
alastairp
scientist1642: hi, the moods in AcousticBrainz use SVM models, trained on a handful of audio examples
If you're interested in emotion and music in general, you could also take a look at our classification challenge that we ran last year (and will run again this year): http://www.multimediaeval.org/mediaeval2019/music/
2020-03-28 08817, 2020
alastairp
heyarne[m]: you can use the submission tool and tell it to submit to your local server instead of acousticbrainz.org
what's your final goal? if you just want to process the audio and do something with it, it _may_ be faster to just use the underlying tools instead of setting up acousticbrainz
2020-03-28 08851, 2020
heyarne[m]
yes, i want to process the audio and do something with it. that something involves querying the info via api, at least i thought it would, because it's going to be a web app since that's where i feel at home
2020-03-28 08847, 2020
heyarne[m]
but i'm not even sure what "the underlying tools" are exactly. the page mentions the essentia toolkit, but that seems vast
if you're just doing your own personal project, I recommend compiling the extractor tool yourself, because it's got lots of improvements since the AB one was released. If you want compatibility with AB, use the one from the download page
2020-03-28 08805, 2020
scientist1642
alastairp woah thanks.
2020-03-28 08807, 2020
scientist1642
In general, is it too late to work on a gsoc proposal now? I have background in machine learning and react/react-native. Seems like there are some suitable projects.
2020-03-28 08820, 2020
alastairp
scientist1642: you're welcome to submit a proposal if you want, but we typically prefer that potential students spend some time with us in the weeks before proposals are due, and iterate over a few drafts of the proposal first
2020-03-28 08809, 2020
alastairp
it's unlikely that anyone will be able to give you feedback on a proposal until Monday evening which means that you'll probably be cutting it quite fine, but it's up to you
2020-03-28 08807, 2020
scientist1642
alastairp ah I see. Makes sense.
2020-03-28 08823, 2020
alastairp
scientist1642: what's your ML experience in?
2020-03-28 08803, 2020
scientist1642
I did an MSc in data mining and went through regular ML courses and assignments but I was mostly interested in deep reinforcement learning.
2020-03-28 08800, 2020
heyarne[m]
I didn't know it's so easy to actually submit data; I'm currently subbmiting info for the first release. No use in keeping the analyses private :)
2020-03-28 08816, 2020
alastairp
scientist1642: ah neat, we do a bit of that at https://www.upf.edu/web/mtg/, though I'm not involved with the deep learning groups
2020-03-28 08825, 2020
alastairp
heyarne[m]: perfect! thanks for the contribution
2020-03-28 08806, 2020
scientist1642
Cool group! I think there is so much to discover in neuroscience regarding how we perceive music. tbh right now I am more inclined to study some basic neurobiology and explore unconventional approaches in AI like http://numenta.com/ is pushing through.
