I DISCOVERED SOMETHING RECENTLY
It takes an AUDACIOUS MAN to succeed.
A SHAMELESS MAN to refuse to quit.
A FOOLISH MAN to believe in an imagination and wanting to make it possible
Takes a STUPID MAN to keep going when stoping is the only thing that make sense.
And a CRAZY MAN to go all in and never look back
To be successful you have to be weird
So tell me which one are you?
@Hailekirkos However, compared to the total volume, he only read close to 2%(I'm generous) of it. Marking that it is too early to judge or flex about reading the book.
@Hailekirkos Explanation.
The image is taken as a selfie given that the image is reflected 180°.
The person in this picture has direct eye contact with the camera 📷 showing that he was not paying attention to the book.
The book pages are bent properly, showing he properly read these
🚀 I am thrilled to share the final Table of Contents for my upcoming book, Introduction to GIS Programming: A Practical Python Guide to Open Source Geospatial Tools! Currently wrapping up the last chapter on Distributed Computing with Apache Sedona. Aiming for a release before in two weeks. Stay tuned!
📚 Book website: https://t.co/csYxHzcg0v
📝 Table of Contents: https://t.co/nLtT08Uc0S
🔗 GitHub: https://t.co/TqQn0JtCsr
#Geospatial #OpenSource #Jupyter #Mapping
Conventional wisdom has assumed that aligning representations (eg. contrastive learning) from different views or modalities is critical for better performance.
In our work, we show a fundamental limit to how much we should align representations and scenarios where more alignment can actually *hurt* performance.
paper: https://t.co/9UByHDPggo
code: https://t.co/ppbVvLOBpR
Our findings unify and resolve several open questions posed by recent work, such as the Platonic Representation Hypothesis https://t.co/X3wD6GIq1K and Infomax/Infomin https://t.co/1tW1LkaI79
see this excellent thread by @mmtjandrasuwita for more details!
I took a brief look at the Harmonic Loss paper
tl;dr: instead of dot-product with softmax, do euclid dist with normalized 1/d**n.
I kinda want this to work. I've dabbled with preferring euclid many times throughout my career (eg triplet loss etc)
However...