Видео

The key concept for me was "the closer a double is to 0, the more exact its representation"

very nice and detailed lecture

It was confusing to understand what happened in 12:47 when the indexes [0,2,3,3,4] appeared. So this is what's going on, the filter removes 2 entries, and only remained indexes 0,2,3 , and those are exactly the first 3 elements in the vector [ *0,2,3* , 3,4], so the last 2 elements 3,4 are ignored. And this is specified by the first vector [ [0,2] , [2,1] ] that tells the first 2 elements is the first row, and the third element is the second row. (there's no mention on what to do with the rest of indexes, so they are ignored)

Thanks for the amazing presentation.

Great presentation and interesting idea! It would be interesting to see what the compression ratio would be for 32 bit floats, bit packing the integers would be SIMD-friendly but I suspect probably reduce the ratio by a significant amount for "normal" values.

👇 'Promo sm'

정말 미친 놈들이 많구나

👏

good

thanks

Data really powers everything that we do

“The world is now awash in data and we can see consumers in a lot clearer ways.”....👍👍👍