Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Deep Neural Networks for YouTube Recommendations

research.google.com

Post reply on HN
Print view View original

81–90 of 102 posts

Re: Deep Neural Networks for YouTube Recommendations

#81

Post by hacker42 »

sevenless wrote: post #48
hacker42 wrote: post #37

Earlier quoted context omitted.

Do you study the phenomenon of information bubbles at Google? Let's say, a German user just happens to watch some right-wing populist video claiming that we need to stop Merkel's refugee politics. The next day the user might receive plenty of recommendations in their feed that confirm the message in the first video. They happen to stumble upon a video of some party convention by an uprising German populist party, and…

What about the reverse scenario, though? Should someone who watches videos about refugee suffering be given anti-refugee video recommendations, lest they be dragged into a 'left wing ideology'? I don't see how that would be acceptable. Would Holocaust documentaries be 'diversified' with Holocaust denial videos? 'Information bubbles' have existed as long as people have had a choice of newspapers to buy and TV channels…

I think personalized media might bring information bubbles to a whole new level because recommendation systems are more efficient and achieving millions of YouTube views has arguably fewer intellectual hurdles than achieving similar reading rates via print or TV. The information passes though fewer filters (none), e.g. proof readers, team discussions. If that causes an increase of misinformation, then it is on Google to fix it, i.e. to reduce the amount of misinformation at least to pre-YouTube levels.

Re: Deep Neural Networks for YouTube Recommendations

#82

Post by supergirl »

Like others here I am also disappointed by the youtube recommended videos. So I was investigating building a better recommender myself. I was actually searching for how the youtube recommender works yesterday but could only find the 2010 paper. Now I am starting to believe that it is not the recommender that is the problem. It is that youtube consists of 99% low quality videos.

Re: Deep Neural Networks for YouTube Recommendations

#83

Post by Eridrus »

pcovington wrote: post #17

Author here - happy to answer questions about the techniques in the paper. We're super excited to finally share this work externally. Feedback about YouTube recommendations in general also welcome.

Thank you for sharing this paper, it's the first attempt I've seen at using Neural Nets for the candidate generation portion, which was cool to see.

How do you decide what the N in Top N should be?

I see you guys scaled features yourselves, why not use BatchNorm?

Do you think you could have eliminated the manual feature engineering with some learnable feature engineering? I'm mostly thinking of some sort of parametric activation functions, but I'm curious if you've thought about it.

Any thoughts on the Wide & Deep paper, did you try incorporating similar ideas?

Did you experiment with LSTMs for turning watch/search histories into fixed vectors?

You guys trained a regression model, whereas the common wisdom is that neural nets aren't so hot at regression, did you try training this as a bucketized classification problem?

Again, thanks for the paper and taking time to answer questions :)

Re: Deep Neural Networks for YouTube Recommendations

#84

Post by vintermann »

seanwilson wrote: post #25
dsco wrote: post #22

One of the very best recommendation engines I've encountered is the "Discover Weekly" playlist from Spotify. It's helped me reconsider my relationship to music which I basically thought was dead since I had hit a rut on exploring new artists. There's an interesting presentation of how it's created on SlideShare http://www.slideshare.net/MrChrisJohnson/from-idea-to-execut...

> One of the very best recommendation engines I've encountered is the "Discover Weekly" playlist from Spotify. The addition of Discover Weekly really confused me. Shouldn't the features that create a radio station from an artist or a playlist fill this need already? Why is it only updated weekly? I haven't tried other services much but it feels like Spotify isn't doing as much as they can with recommendations.

I suspect that the reason Discover Weekly is so much better than radio, and is only available once per week, is that it's computationally costly.

Either way, radio-from-artist or radio-from-song can't utilize your listening history. It's quite possible you listen to an artist for different reasons than the reasons most people listen to that artist - in that case, you will get recommendations based on the majority's reasons.

You would expect personalized recommendations to have potential to do much better, and I'd argue that's exactly what we see with DW.

Re: Deep Neural Networks for YouTube Recommendations

#85

Post by vintermann »

pcovington wrote: post #17

Author here - happy to answer questions about the techniques in the paper. We're super excited to finally share this work externally. Feedback about YouTube recommendations in general also welcome.

First question: Are your innovations being used on Youtube today? At what date did you "pull the switch" and go over to the DNN-based recommender systems?

I'd say recommendations have gotten somewhat better, if a little too clickbaity still (you saw one video with a squirrel? here are ten squirrel video compilations!).

Second question: Did humans at some point assign names to DNN-established clusters or vector elements or what they're called? Sometimes I get OK recommendations, but with a really bad label (for instance a 100% minecraft LPer recommended as an example of a "shooter").

Re: Deep Neural Networks for YouTube Recommendations

#86

Post by Firadeoclus »

pcovington wrote: post #17

Author here - happy to answer questions about the techniques in the paper. We're super excited to finally share this work externally. Feedback about YouTube recommendations in general also welcome.

Hi Paul,

Thanks for making your hard work available. It is very interesting from a technical point of view. I'm struck by just how huge a challenge this is given the enormous corpus size.

Are you familiar with Joe Edelman's work?[0] He specifically uses YouTube recommendations as an example of many algorithms designed to use the wrong metrics which leads to undesirable outcomes for users.

Have you ever looked into attributing reasons to users' visits? It seems likely that many users aren't looking for general recommendations that blend their entire use of YouTube together but want specialized recommendations linked to why they visited YouTube this time.

[0] https://medium.com/@edelwax/is-anything-worth-maximizing-d11...

Re: Deep Neural Networks for YouTube Recommendations

#87

Post by Ntrails »

genofon wrote: post #74
thr0waway1239 wrote: post #73

Earlier quoted context omitted.

I can confirm this. I saw a lot of region-based, really stereotyping videos when I got a new computer and visited YouTube. While I understand they need to fill up the home page with something, I just prefer they can be a little more creative about the whole thing. After glancing through the top few, I quickly went for the search box. To their credit, once I do a few searches, the recommendations drastically improved…

if they don't have any good information about you other than your location what else would you recommend rather than the most popular in that location?

When you lack the information to make a good recommendation, "the best we can guess from really generic or sparse information" tends to be annoying. That's really bad.

Yes, I watched a bunch of Dota replays during a recent tournament. No, I don't normally go on youtube. So I watched a daily show video clip that was linked. All my "watch next" and "recommended" are Dota. That's not smart, that's aggravating. I would have been ok watching a couple more ds clips, but instead exclusively bad recommendations were made based on poor data.

I dislike the idea that my world gets filtered by algorithms, but I really hate when they're obviously bad at it. Although I suppose I should be grateful that it's easily spotted when it's bad?

Re: Deep Neural Networks for YouTube Recommendations

#88

Post by cscurmudgeon »

tnones wrote: post #66
cscurmudgeon wrote: post #54

Earlier quoted context omitted.

Ah good old moral relativism.

As opposed to planting your flag in the ground that your camp is always right and the outgroup is evil?

As opposed to at least trying to think a bit.

If you think relativism is fine. Then "As opposed to planting your flag in the ground that your camp is always right and the outgroup is evil?" is also fine.

See how silly and immediately self-contradictory relativism is?

Re: Deep Neural Networks for YouTube Recommendations

#89

Post by cscurmudgeon »

force_reboot wrote: post #55
cscurmudgeon wrote: post #54

Earlier quoted context omitted.

Ah good old moral relativism.

Isn't the concept of an "information bubble" already inherently relativistic? sevenless was simply pointing out the relativism cuts both ways.

I was just pointing out the obvious.

Re: Deep Neural Networks for YouTube Recommendations

#90

Post by bryanrasmussen »

Is this the source of all those Recommendations that I look at Ben and Holly cartoons in the middle of the day when my daughter is at school? Or more Italian daytime television when I've just watched a video my wife never would? In short, is this the reason why there's never anything interesting for me when I go to the frontpage of youtube?!?!
Post reply on HN