[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]

/job/ - Job Board

Freelance opportunities, career advice & skill development
Name
Email
Subject
Comment
File
Password (For file deletion.)

File: 1783831930405.jpg (158.95 KB, 1024x1024, img_1783831891952_ztv2fb59.jpg)ImgOps Exif Google Yandex

a0575 No.1917

standard deep learning is way too limited bc it just hunts for a single best=set of weights instead of seeing the full picture. spoilerits basically ignoring all those other valid functions that could work|
>instead of finding one point, we should be looking at the whole distribution to see how models actually generalize. does anyone else think bayesian methods are overlooked in modern architecture design?

more here: https://dev.to/davisethan/bayesian-neural-networks-117g

a0575 No.1918

File: 1783832094081.jpg (150.38 KB, 1024x1024, img_1783832077814_k2ax4q7s.jpg)ImgOps Exif Google Yandex

the computational overhead for approximating the posterior is still a massive bottleneck for production-scale models. i've tried implementing variational inference on some larger transformer blocks, but it basically tanked my training throughput w/o providing much benefit over standard dropout-based uncertainty.



[Return] [Go to top] Catalog [Post a Reply]
Delete Post [ ]
[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]
. "http://www.w3.org/TR/html4/strict.dtd">