Published March 15, 2019 © CC BY-NC-ND

Neuronium

Light sculpture that represents a training artificial neural network in its simplest architecture.

IntermediateShowcase (no instructions)4,465

Things used in this project

Hardware components

Arduino UNO

Story

Neuronium (Prototype)

Intro

Neuronium is a light sculpture that represents a training Artificial Neural Network in its simplest architecture.

It consists of two inputs, one hidden layer with three nodes and one output. The represented network learns to solve the “Exclusive Or” (XOR) problem.

Light intensity shows the calculated strength of the artificial neurons and synapses at every point of time during the training. The dynamic light flow makes the "hidden" layers in a neural network visible.

Inputs: 0 | 1

Concept & Motivation

Neural Networks are fed with large datasets running through an even bigger amount of hidden layers to train the system to give a certain output. This process is inscrutable for us.

Neuronium is planned as a sculpture in human scale. Referencing the Atomium (1958) the embodied abstract concept of an artificial neural network is placed in eyes height to symbolise the severity of Machine Learning on our future lifes and to promote a peaceful use of the technology.

Neuronium Model

A prototype of the Neuronium in small scale was realised, that shows the light flow and a projection of the weights calculation.

Realisation

1 / 2

The XOR training is executed in python code. The mapped results for the different inputs, outputs and weights are send as values for the luminosity of the LED's to the Arduino via Serial.