Barlow Twins: a new super-simple self-supervised method to train joint-embedding architectures (aka Siamese nets) non

Basic idea: maximize the normalized correlation between a variable in the left branch and the same var in the right branch, while making the normalized cross-correlation between one var in the left branch and all other vars in the right branch as close to zero as possible.
2/N
In short: the loss tries to make the normalized cross-correlation between the embedding vectors coming out of the left branch and the right branch as close to the identity matrix as possible.
3/N
The 2 branches are always fed with differently-distorted version of the same image, and there is no need for dissimilar training pairs.

The objective makes the embedding vectors of the two branches as similar as possible, while maximizing their information content.
4/N
No contrastive samples, no huge batch size (optimal is 128), nor predictor, no moving-average weights, no vector quantization, nor cut gradients in one of the branches.
5/N
Competitive results on ImageNet with a linear classifier head.
Great results on semi-supervised ImageNet in the low labeled-data regime and on transfer tasks.
6/N
Results on ImageNet with linear classifier head
7/N
Results with 1% and 10% of ImagNet labeled images
8/N
Results on transfer tasks.
9/N
Arch is standard ResNet50 with 2048-D feature vec.
But contrary to others, the embedding size (projector output) is larger. The perf keeps going up as the embedding dim grows (we stopped at 16384).
Probably cause the feature vars are made independent, not just decorrelated.
10/N
Why Barlow? Horace Barlow was a pioneer of visual neuroscience who proposed the idea that the brain tries to minimize redundancy in representations.

By Jure Zbontar, Li Jing, Ishan Misra, yours truly, and Stéphane Deny.
All from FAIR.
To appear at ICML 2021
11/N
Don't you just hate slicing what would be a decent-size post into threaded thin tweets?
12/N
No, really. Don't you hate reading those long thread slices?
If you do, you could just read my Facebook post:
https://t.co/dQii7BEPQ5
13/N
N=13
Typo: optimal batch size is 1024, not 128.
14/13 (haha).

More from All

You May Also Like

🌿𝑻𝒉𝒆 𝒔𝒕𝒐𝒓𝒚 𝒐𝒇 𝒂 𝑺𝒕𝒂𝒓 : 𝑫𝒉𝒓𝒖𝒗𝒂 & 𝑽𝒊𝒔𝒉𝒏𝒖

Once upon a time there was a Raja named Uttānapāda born of Svayambhuva Manu,1st man on earth.He had 2 beautiful wives - Suniti & Suruchi & two sons were born of them Dhruva & Uttama respectively.
#talesofkrishna https://t.co/E85MTPkF9W


Now Suniti was the daughter of a tribal chief while Suruchi was the daughter of a rich king. Hence Suruchi was always favored the most by Raja while Suniti was ignored. But while Suniti was gentle & kind hearted by nature Suruchi was venomous inside.
#KrishnaLeela


The story is of a time when ideally the eldest son of the king becomes the heir to the throne. Hence the sinhasan of the Raja belonged to Dhruva.This is why Suruchi who was the 2nd wife nourished poison in her heart for Dhruva as she knew her son will never get the throne.


One day when Dhruva was just 5 years old he went on to sit on his father's lap. Suruchi, the jealous queen, got enraged and shoved him away from Raja as she never wanted Raja to shower Dhruva with his fatherly affection.


Dhruva protested questioning his step mother "why can't i sit on my own father's lap?" A furious Suruchi berated him saying "only God can allow him that privilege. Go ask him"