unconventional-computing

Concept paper for using living mycelium as a self-repairing substrate for AI, plus a Python prototype: scikit-learn models that predict mycelium growth from environmental readings, served with FastAPI. Bring your own data; no bio-interface yet.

Nuclear compute (radiative compute) as an alternative computing substrate to electronic and quantum machines: one radioactive medium supplies power, logic, memory, and interconnect, enabling sealed portable computers that compute without external power.

Code for "Evolution of Acoustic Logic Gates in Granular Metamaterials", Parsa et al., EvoApplications 2022

Code for "Universal Mechanical Polycomputation in Granular Matter", Parsa et al., GECCO 2023

Code for "Evolving Programmable Computational Metamaterials", Parsa et al., GECCO 2022

Code for "Gradient-based Design of Computational Granular Crystals", Parsa et al.

TopoCore is an experimental spatial execution architecture where programs are represented as traversable 2D layouts rather than linear instruction streams. Inspired by Befunge, semiotics, and semasiographic systems, it explores topology-driven control flow, direction-aware computation and non-linear execution 🕸️.

A PyTorch framework for training neural stochastic differential equations (SDEs) to model dynamical systems with both deterministic and stochastic components. This framework enables the learning of dynamical systems through a combination of neural networks representing drift and diffusion terms, with adversarial training. Docs ⬇️

Code and movie files of myxobacteria movement modelled using P-System methodologies.

Finite-element study of quasicrystal-perforated MEMS resonators and what their mode structure can (and cannot) compute as a physical reservoir. Includes the paper.

A nonvolatile memory cell made of gravity: a bit stored in which Lagrange island (L4/L5) a body librates in — written by orbit insertion, rewritten by a real flyby, and validated against Jupiter's real 148-year Trojan libration.

A package of simple reservoir computing tools

Python and JAX-based simulation framework for investigating non-equilibrium attractor dynamics, finite-size scaling, phase synchronisation, and adaptive polarity modulation in harmonically confined density-feedback Yukawa plasmas. Developed as part of an independent research framework exploring physical precursors for active-matter phase logic.

Lab report: a symmetry selection rule for computation in phononic / quasicrystal MEMS reservoirs.

This repository contains a collection of computational models inspired by calcium signaling dynamics in biological systems. These models demonstrate how principles from cellular calcium dynamics can inform novel computing paradigms with unique capabilities beyond traditional computing approaches.

Logic gates and a 4-gate priority-encoder pipeline built from nothing but Newtonian point-mass gravity