Dataflow matrix machines

[dmm]

Dataflow matrix machines: a class of expressive self-modifiable neural machines

• which can fluently modify their own weights, connectivity patterns, and size;
• and can serve as a programming platform, resulting in a programming formalism with continuously deformable programs.

---

Reference paper: Dataflow Matrix Machines and V-values: a Bridge between Programs and Neural Nets

arxiv.org/abs/1712.07447

More details:
anhinga-anhinga.dreamwidth.org/83104.html
anhinga-anhinga.dreamwidth.org/82703.html

*** Dataflow Matrix Machines internet resources: anhinga.github.io

Interdisciplinary list of open research directions (I am looking for collaborators): dmm-collaborative-research-agenda.pdf

*** My blogs:

Main Dreamwidth blog: dmm.dreamwidth.org (partial mirror: anhinga-travel.livejournal.com )
Main LiveJournal: anhinga-anhinga.livejournal.com (mirror: anhinga-anhinga.dreamwidth.org )
Auxiliary LiveJournal: anhinga-drafts.livejournal.com (mirror: anhinga-drafts.dreamwidth.org )


🇺🇦 🇺🇦 🇺🇦 🇺🇦 🇺🇦 🇺🇦 🇺🇦

Comments

[dmm]

So... the second time in a couple of hours I see a mention of John Hughes... amazing... (The other one was his "arrows" as an inspiration for combinators in the TensorFlow's Fold library...)

Time to sleep and meditate on this...