Jun 9, 2022
Bennett’s answer to Igor’s Supersymmetric Artificial Neural Network Question:
There are a number of ways I proposed to construct a supersymmetric deep learning model, which can be reasonably distilled as variants to your inquiries/question parts.
____________
A. Re: your 1st Question about Grasmann relation:
My 2016 paper proposed a C∞ Supermanifold model, involving charts, related to Grasmann and Stiefel manifolds.
We know reasonably, that anticommutation involves Grasmannian numbers.
**Source**: Seen in the naive Supersymmetric architecture in my paper, attached to Url[a].
____________
B. Re: your 2nd Question about Exact Susy solutions:
In 2017, I proposed Supersymmetrizing a Boltzmann machine, aka creating a transverse field Ising spin compatible super Hamiltonian, which reasonably concerns N=1 SQCD, where exact solutions may be computed with finite length.
Side note: As seen in Url[d], a somewhat informal colleague had become interested in my work in 2017, then wrote about the this proposal in 2019, in addition to a number of proposals or pathways of mine for my so called supersymmetric neural network. (I am computer scientist, while the colleague originates from a Physics background!)
*Source**: Seen in my Github repository’s question collection, attached to Url[b] and Url[c].
____________
Urls/Appendix
[a] My supersymmetric neural network [2016]:
- https://openreview.net/pdf?id=SJewsu6qOV
- https://github.com/JordanMicahBennett/Supersymmetric-artificial-neural-network
[b] My github repository collection concerning my 2017 proposal for transverse field ising spin compatible super hamiltonian (last edited 2017):
[c] A particular [2017] forum post containing a sample question from collection [b], concerning my 2017 proposal for Transverse Field Ising Spin compatible Super Hamiltonian:
https://www.physicsoverflow.org/39603/possible-create-transverse-ising-compatible-hamiltonian
[d] Mitchel Porter’s reasonably robust review [2019] citing my work in section 5.1, page 5, where Michel correctly identifies the ways I proposed a Susy Deep Learning model:
