News for hackers. Stuff that matters.

A comment on Posit floating point numbers: thin triangles and other tricks (2019). Parent comment

Re:Posit floating point numbers: thin triangles and other tr... by immibis on Thursday June 19, 2025 @05:52PM (#44322858)
Silly idea: what if we ran ML models on pure logarithms and used some completely terrible approximation for addition (maybe just the max function, or max plus a lookup table based on the difference between the two numbers when it's small)
[ Parent | on HN ]
Re:Posit floating point numbers: thin triangles and other tr... by adgjlsfhk1 on Thursday June 19, 2025 @07:33PM (#44323487)
once you're down to 8 bits or so you can totally do this with a lookup tables (for 8 bits, you need 512 bytes of table which isn't too bad.
[ Parent | on HN ]
Re:Posit floating point numbers: thin triangles and other tr... by fc417fc802 on Thursday June 19, 2025 @11:33PM (#44324489)
It's been done. [0][1] But the fact that BitNet [2] is possible calls into question the need for it. NNs are surprisingly robust to a wide array of distortions. [3]

[0] https://arxiv.org/abs/1603.01025

[1] https://par.nsf.gov/servlets/purl/10163070

[2] https://arxiv.org/abs/2310.11453

[3] https://arxiv.org/abs/1606.01981

[ Parent | on HN ]