Abstraction and LSH

2018-10-09 - Abstraction is dimensionality reduction - Dimensionality reduction is an abstraction

• To begin with, there is no way that the brain is “calculating the inner product in order and looking for the closest one” to find vectors with a high degree of similarity.

  • LSH?
    • https://www.slideshare.net/mobile/yaruki_nil/lsh
    • Boltzmann machine?
  • Asociatron?
    • http://www.sist.ac.jp/~kanakubo/research/neuro/associatron.html

• Rather, is a neural net (if realized on a device capable of parallel computation) a device that can efficiently compute probabilistic Near Neighbor

  • The cost is just too high in making it happen with computers.
    • http://people.csail.mit.edu/indyk/mmds.pdf

2023-12-01

• HNSW is a popular algorithm that can efficiently compute [approximate NearestNeighbors

---

This page is auto-translated from /nishio/抽象化とLSH using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.