no, scaling is of help but its hard to quantify how. i have two other observations we can discuss then
2020-10-02 27632, 2020
ruaok
scaling, not full normalization?
2020-10-02 27656, 2020
_lucifer
ruaok: i only tried scaling till now, as we discussed the other day
2020-10-02 27613, 2020
ruaok
do you think we'll get an improvement going further?
2020-10-02 27645, 2020
_lucifer
1) i got scores like 0.4 and 0.2 so on on a scale of 0 to 1 when i used a smaller set of listens
2020-10-02 27646, 2020
_lucifer
2) these scores are somewhat in line with using the ones we got when we were using counts directly
2020-10-02 27614, 2020
_lucifer
hard to say ruaok, the question is how to tell if the recs became better
2020-10-02 27622, 2020
pristine___
_lucifer: what is mean is say I have two rec A and B, before scaling if the scores were 1, 2 and after scaling they are 10 and 20, does this mean any improvement?
2020-10-02 27658, 2020
_lucifer
pristine___: the scores are relative and not an indication of whether the user will like it, so a strong maybe
2020-10-02 27634, 2020
_lucifer
the scores make sense relatively but absolutely they are just a bunch of number
2020-10-02 27654, 2020
pristine___
I understand they are relative, just trying to understand if we really need to scale/normalize
2020-10-02 27601, 2020
pristine___
:)
2020-10-02 27602, 2020
_lucifer
pristine___: imo we probably should because that theoretically brings all users on a same scale but practically to decide that we need some way to see if tell whether the recs have become better or not
pristine___: what is the usual rmse for the model ?
2020-10-02 27641, 2020
pristine___
I don't remember :( You can generate model HTML file for that it had a table
2020-10-02 27650, 2020
pristine___
Has*
2020-10-02 27604, 2020
_lucifer
i did, its 0.12
2020-10-02 27614, 2020
_lucifer
i wanted to know the value without scaling
2020-10-02 27647, 2020
pristine___
I get that, but I don't remember:( I will search for the html in my pc once I reach hotel :)
2020-10-02 27611, 2020
_lucifer
np, thanks
2020-10-02 27637, 2020
shivam-kapila
_lucifer: any talks you preparing
2020-10-02 27656, 2020
_lucifer
no, shivam-kapila , you ?
2020-10-02 27657, 2020
shivam-kapila
Wanted to, but couldnt find anything interesting
2020-10-02 27652, 2020
pristine___
> imo we probably should because that theoretically brings all users on a same scale but practically to decide that we need some way to see if tell whether the recs have become better or not
2020-10-02 27602, 2020
pristine___
Yeah _lucifer. Even if they theoretically bring all users to same scale, I think we should only work on this further if recs are actually improved for that we need a way to determine if recs have improved or not.
2020-10-02 27622, 2020
shivam-kapila
Improved recs is rather a user dependent term
2020-10-02 27633, 2020
shivam-kapila
Feedback is the best here
2020-10-02 27620, 2020
pristine___
shivam-kapila: of course! Didn't pay attention to the best term here but I guess _lucifer understood 😉
shivam-kapila: Hi! About the folow button, I think it looks nice. If you want to try to hide text until the user hovers (you'll need to add icons there surely), check out the existing code: https://github.com/metabrainz/listenbrainz-server…
2020-10-02 27624, 2020
yvanzo
I filled Saturday’s timetable with “State of stuff” topics, Sunday can be for more sessions but there are only MB ones for now, more can be added to the wiki.
2020-10-02 27608, 2020
imdeni has quit
2020-10-02 27640, 2020
imdeni joined the channel
2020-10-02 27634, 2020
alastairp clones troi
2020-10-02 27641, 2020
shivam-kapila
Mr_Monkey: hi
2020-10-02 27653, 2020
yvanzo
updating beta.mb.o
2020-10-02 27659, 2020
shivam-kapila
I wanted to know what you prefer
2020-10-02 27612, 2020
shivam-kapila
Those icons ones or only text ones
2020-10-02 27615, 2020
CatQuest
I thikn riker would dsilike that
2020-10-02 27622, 2020
CatQuest
(or possibly *like* who knows :3)
2020-10-02 27602, 2020
shivam-kapila
I was confused as new ux concepts say to put less transitions
2020-10-02 27612, 2020
shivam-kapila
So that it doesnt steal focus
2020-10-02 27631, 2020
diru1100
is anyone into pictionary? or only sticking to among us
shivam-kapila: Ah, I see. Well, personally I like the transitions, and that the button is small when not hovered. But you could also just resize your follow button and have a similar result without the need for transitions.
2020-10-02 27602, 2020
shivam-kapila
Lets stick to what iliekcomputers suggested
2020-10-02 27606, 2020
ruaok
Mr_Monkey: was it you who brought the projector last year?
2020-10-02 27608, 2020
shivam-kapila
Thanks for feedback
2020-10-02 27619, 2020
Mr_Monkey
ruaok: Yis
2020-10-02 27628, 2020
ruaok
planning on bringing it again?
2020-10-02 27651, 2020
Mr_Monkey
Oh, shivam-kapila, one more thing: the button doesn't look vertically centered with the user name, as far as I can tell from a quick look
2020-10-02 27609, 2020
Mr_Monkey
ruaok: sure, if we need it no problem
2020-10-02 27627, 2020
ruaok
I think it might be nice to have, if you dont mind.
2020-10-02 27613, 2020
shivam-kapila
Mr_Monkey: old one or proposed one
2020-10-02 27640, 2020
Mr_Monkey
The screen you shared yesterday (was it yesterday?)
iliekcomputers suggested not to change the button near username
2020-10-02 27644, 2020
shivam-kapila
But only in feed list
2020-10-02 27617, 2020
Mr_Monkey
Ah, I see
2020-10-02 27620, 2020
alastairp
ruaok: do you want review feedback on troi too? I feel the answer to most of my questions of the form "why did you do it like this?" will be "because that's how it organically grew"
2020-10-02 27629, 2020
alastairp
I can contribute changes in, but I'm not sure how much of it is in flux
2020-10-02 27626, 2020
ruaok
yes, I'd like that feedback. there are a lot of things still in flux, but I think things are starting to stabilize.
2020-10-02 27610, 2020
ruaok
I forgot that I had already added troi.acousticbrainz.annoy -- so that part is already done.
2020-10-02 27630, 2020
alastairp
yep, I just have a few parameters to change there
2020-10-02 27636, 2020
ruaok
I'm writing a manage.py-like entry script as we speak.
2020-10-02 27653, 2020
alastairp
and I'm just putting in a new distance metric now
2020-10-02 27655, 2020
alastairp
I remember that we had this discussion about BPM. for reasons, we currently treat integer multiples of a bpm as the same (e.g. I just tried with a 71bpm track and got back a 142bpm track as very similar)
2020-10-02 27634, 2020
ruaok
peak finding in frequency space?
2020-10-02 27642, 2020
alastairp
yeah, basically
2020-10-02 27611, 2020
alastairp
people aren't really good at distinguishing them in many situations either
2020-10-02 27615, 2020
kieto joined the channel
2020-10-02 27640, 2020
ruaok
I wonder how many real world scenarios that will go wrong in.
2020-10-02 27649, 2020
ruaok
I look forward to discovering that bug.
2020-10-02 27642, 2020
alastairp
I'm trying to remember where I encountered this, I think someone I know was building a dataset of known tempos in order to evaluate an algorithm
2020-10-02 27611, 2020
alastairp
get many people to annotate each song, etc, in order to be sure that the annotation is good
2020-10-02 27632, 2020
alastairp
and some songs there was a 50% split between people who thought it was 60 and who thought it was 120
2020-10-02 27636, 2020
alastairp
I'm trying to look it up
2020-10-02 27638, 2020
ruaok
and these weird mutliples started showing up...
2020-10-02 27654, 2020
alastairp
anyway, I think that when asking for songs of bpm 120, people would rather have a song _not_ show up (because the algorithm incorrectly chose 60 instead of 120) instead of having a bunch of obviously 60bpm tracks show up as "close"
2020-10-02 27652, 2020
ruaok
agreed.
2020-10-02 27657, 2020
alastairp
ah, some interesting feedback from MTG
2020-10-02 27614, 2020
alastairp
> you should take other variables into account then. tempo is highly correlated with genre
2020-10-02 27619, 2020
alastairp
> don’t recommend techno 60BPM to people who listen to ambient with 60BPM
2020-10-02 27611, 2020
alastairp
sounds like our hybrid similarities (that include more than 1 feature) are going to be necessary soon
2020-10-02 27641, 2020
_lucifer
like building a content profile ?
2020-10-02 27659, 2020
ruaok
alastairp: that's totally duh in hindsight. :)
2020-10-02 27648, 2020
alastairp
yeah, but you can see how starting with an observation (tempo octave errors happen), and going down a particular path gets you in trouble
I would stay that there are many many paths in recommendation work... most of them get you into trouble.
2020-10-02 27636, 2020
alastairp
(it's 200 pages, you don't need to read it all)
2020-10-02 27617, 2020
alastairp
in section 6.6.3 he proposes a new metric for comparing tempo estimation error by using log2 scale, so that the scale of the error is the same in different octaves
2020-10-02 27652, 2020
alastairp
most people evaluate algorithms using "if it's within 4%, it's good". but 4% of 60 is not the same as 4% of 120!
2020-10-02 27603, 2020
ruaok
ding
2020-10-02 27659, 2020
alastairp
I'm getting some good suggestions in another chat, which I'll note down
2020-10-02 27610, 2020
alastairp
it also depends on if we want to do "similarity", or just plain search
2020-10-02 27618, 2020
alastairp
I wonder if it makes sense for these features with fewer dimensions to use annoy, or if we actually just need solr
2020-10-02 27634, 2020
ruaok
BPM and key?
2020-10-02 27652, 2020
alastairp
yeah. "get me things with feature x" should just be search
2020-10-02 27657, 2020
ruaok
I think they are a poor fit for annoy. the results end up in too few buckets.
2020-10-02 27602, 2020
ruaok
+1
2020-10-02 27610, 2020
alastairp
when you need to introduce fuzzyness into it, that's where it excels
2020-10-02 27626, 2020
ruaok
yep, with range data.
2020-10-02 27653, 2020
CatQuest
diru1100: I'd love some other game than fightinggame tbh
2020-10-02 27617, 2020
CatQuest
"why did you do it like this?" will be "because that's how it organically grew"
2020-10-02 27617, 2020
CatQuest
this needs to be quoted for truth. mb 20 yr history, that is the answer to a lot of newer users question :D
2020-10-02 27639, 2020
CatQuest
diru1100: https://skribbl.io/#collapseHow does look liek fun! reosarevok wanna play this isntead/inaddition to this among us thing?
2020-10-02 27629, 2020
diru1100
It's pictionary
2020-10-02 27646, 2020
diru1100
You draw the word that you got and others will guess
