GitHub

Internal Documentation

Documentation for the internals of MNN.jl. These are subject to change.

Contents

Index

Public Interface

MNN.Behaviour — Type
abstract type Behaviour

Abstract supertype of all behaviours the network can be optimized for. Implemented behaviours:

TODO: Add implemented behaviours here.

source
MNN.House — Type
mutable struct House <: Network

Rough model of a house with a square layout and gravity. The purpose is to model earthquakes and analyse the effects on the "house" and effectiveness of training MNNs for earthquake / vibration protection.

source
MNN.Network — Type
abstract type Network

Abstract supertype of all mechanical neural networks. Implemented networks:

  • MNN.TestNetwork: Network described by Lee et al. (2022) with springs arrangd in equilateral triangles.
  • MNN.House: Rough model of a house with a square layout and gravity.
source
MNN.Neuron — Type
struct Neuron

Represents a node between springs. Position and velocity are stored in the MNN.Network.

Fields

  • movable::Bool: Is the neuron movable (not fixed to wall)?
source
MNN.Spring — Type
struct Spring

Represents an edge between neurons.

Fields

  • spring_constant::Float64: The spring constant.
  • length::Float64: The length of the spring at rest; this is constant and can't be changed after creation of the object.
source
MNN.calc_ydist — Method
calc_ydist(xdist)

Calculate necessary distance on y axis to create an equilateral triangle with given distance between columns on x axis.

source