The Bionic Man
88.5K posts
The Bionic Man
@TheRealMasud
I won’t stop dancing, until the curtain falls, by then my story will be a compelling movie.
Katılım Mayıs 2010
4.6K Takip Edilen3K Takipçiler
'It's only when Argentina plays that opposing coaches and players complain about the ref'
Fabrizio Romano@FabrizioRomano
🚨🇫🇷 Didier Deschamps: “I will ask a question… does the referee possess the required level to manage a World Cup semi-final?”.
English
@TheRealMasud Shey na this one be small price?
God bless you sir
English
Guys I won the prize.
Thank you very much sir @TheRealMasud
God will continue to bless you.
I am happy for real
The Bionic Man@TheRealMasud
Is there a data science newbie that can explain what just happened on twitter Small price dey
English
@Zlatandiary It’s the most important are on a field
Plus the only wing they really lack is CF, they are good accross board and close to perfect in midfield
English
Of course. As long as the midfield is good, everything else can be optimised.
pretty Stella@prettys50452851
@Zlatandiary this is what you have always said.
English
Is that Sami Khedira?
Fabrizio Romano@FabrizioRomano
José Mourinho and his staff following the semi final. 😁
हिन्दी
The Bionic Man retweetledi
😂😂
I won't be shock if twitter decide to make likes public next and make it private only for paid users.
Even this feature of seeing your followers tweet.
I won't be shocked if overtime it becomes only available for paid users.
English
Is there a data science newbie that can explain what just happened on twitter
Small price dey
English
@aykhalid_1 Since you are the one that addressed the issue topic, even if you didn’t address the issue , so send your account in my dm
English
@TheRealMasud Thanks for the detailed explanation. I wish I could explain it better than I did.
I will be glad if you can get more of this coming once in a while.
You won't believe my sleep disappeared doing research just to get the right answer.
English
You are the only one that touched it
Twitter wasn’t initially created for vitality, like TikTok, Elon switching to “for you” feed like TikTok means he banked a lot on all major forms of recommendation in one, but weight of allocation was distorted and your followers content was relegated to the background, that’s the actual design of a “for you” page
Technically they shouldn’t have changed it , if you want your followers go to “following” but since user experience is almost dead they had to yield to our calls
English
@aykhalid_1 Not exactly that, this is a continuation of my initial response . What’s described in the video is kinda like graph network analysis , kinda like estimating span of influence , LinkedIn show it more with the level of connection 1st, 2nd , 3rd ish
x.com/TheRealMasud/s…
The Bionic Man@TheRealMasud
So twitter isn’t a platform with detail profile and metadata beyond location, so they focus more on collaborative filtering really Now the actual issue is, twitter feed used to be like LinkedIn, what you see is based on what the people you follow do, no recommendation engine was used Then Elon switch that , start two feeds “For You” and “Following” for you is exploratory , you are getting recommendations based on interaction , not based on who you follow , they combine content filtering and collaborative filtering , so metadata is key (this is what made the feed disruptive ) So even if you’re not following me , if you interact with me, you’ll see more of my tweets, not just that, you’ll see people like me or that tweet like me, or If you like my tweet, you’ll see tweets that others who like my tweet, also like And the metadata filter brings in search relating content filtering All combined made it a mad feed. They would have made it better by applying a reasonable scale for the feed , say 50 percent from your followers, 20 for content based collaborative filtering , 20 content based collaborative or 10 content based filtering Instead they placed priority on the recommendations, eliminating your followers posts (this is the actual problem and what I was looking for, the explanation was just to differentiate it from what you explained) They’ve now decided to prioritise your followers content over the random ones brought from recommendation Now remember recommendations isn’t saying you’ll definitely like what’s recommended, it’s saying you probably will (mostly around 10-20percent chance you’ll like it) Imagine flooding your feed with something that you have 10-20 percent chance of liking, it’s huge amount of unbearable noise Lasting solution is , weighted distribution of feed based on the example above, and also allowing users to give explicit feedback (like and unlike) and also taking implicit feedback from users (if you give me content from a user like thrice and I don’t interact, it mean I don’t like it, don’t give me anymore)
English
I remember this Mutiu header and Olise shot like yesterday
🇳🇬 Super Eagles@NGSuperEagles
Beating Spain isn’t for everyone
English
So twitter isn’t a platform with detail profile and metadata beyond location, so they focus more on collaborative filtering really
Now the actual issue is, twitter feed used to be like LinkedIn, what you see is based on what the people you follow do, no recommendation engine was used
Then Elon switch that , start two feeds “For You” and “Following” for you is exploratory , you are getting recommendations based on interaction , not based on who you follow , they combine content filtering and collaborative filtering , so metadata is key (this is what made the feed disruptive )
So even if you’re not following me , if you interact with me, you’ll see more of my tweets, not just that, you’ll see people like me or that tweet like me, or If you like my tweet, you’ll see tweets that others who like my tweet, also like
And the metadata filter brings in search relating content filtering
All combined made it a mad feed. They would have made it better by applying a reasonable scale for the feed , say 50 percent from your followers, 20 for content based collaborative filtering , 20 content based collaborative or 10 content based filtering
Instead they placed priority on the recommendations, eliminating your followers posts (this is the actual problem and what I was looking for, the explanation was just to differentiate it from what you explained)
They’ve now decided to prioritise your followers content over the random ones brought from recommendation
Now remember recommendations isn’t saying you’ll definitely like what’s recommended, it’s saying you probably will (mostly around 10-20percent chance you’ll like it)
Imagine flooding your feed with something that you have 10-20 percent chance of liking, it’s huge amount of unbearable noise
Lasting solution is , weighted distribution of feed based on the example above, and also allowing users to give explicit feedback (like and unlike) and also taking implicit feedback from users (if you give me content from a user like thrice and I don’t interact, it mean I don’t like it, don’t give me anymore)
English
Collaborative filtering can either be Content based or User based
Content based collaborative filtering is “people who like this, also like that” , so since everyone that like A also like B, so if you like A , you are expected to like B
User based collaborative filtering is if Person A likes item U V W X Y Z, and person be likes U V W X , it’s expected that they’ll like Y Z too
Content filtering is using item and user metadata to predict what a person will like, it doesn’t rely on interaction data like the other one
Content filtering is to use your profile and preference to recommend, user based collaborative filtering is for exploration or preference , item based collaborative is to widen your need
English
I’ll be more confident of them facing France than Spain tho
Subomi@ADEKOYASUBOMI
England is winning this World Cup. I would have been more concerned if they were facing France in the final, but against Spain, I believe they can nick it. They need to get past Argentina tomorrow. One game at a time.
English
@aykhalid_1 Not correct, but first person that is addressing a tiny bit of the issue
It’s related to recommendations, but not what you explained
English
@TheRealMasud Should I say it is a type of recommendation system. Specifically hybrid recommendation system, which is a combination of content based filtering and collaborative filtering.
Unlike what it used to be where context was prioritized.
English
@TheRealMasud Both Kounde and Digne were already average, they then decided to raise the bar even lower tonight
English
We all know France’s weakness is their fullback, and Spain exploited it big time
English
@SirJarus Even France didn’t seem like they respected Spain enough
Same way Spain won euros , they look basic on paper , but very good
English
