I am focusing on Dataflow Matrix Machines which perform neuromorphic computations with arbitrary linear streams.
Dataflow Matrix Machines allow us to obtain general-purpose continuously deformable programs and neural machines with convenient and flexible self-modification facilities for compositional metalearning:
I presented a virtual poster at JuliaCon 2021 on July 30:
"Multiplying monochrome images as matrices: A*B and softmax":
https://github.com/anhinga/JuliaCon2021-poster
Materials for my JuliaCon 2023 talk:
"Exploring synthesis of flexible neural machines with Zygote.jl":
https://github.com/anhinga/DMM-synthesis-lab-journal/tree/main/JuliaCon2023-talk
โ
I am looking for collaborators (ping me by opening an issue).
๐บ๐ฆ ๐บ๐ฆ ๐บ๐ฆ ๐บ๐ฆ ๐บ๐ฆ ๐บ๐ฆ ๐บ๐ฆ