Anneliese Brei

This artwork visualizes how artificial neural networks make decisions in machine learning. It interactively steps through each stage of a perceptron, the smallest complete building block of a neural network. In this way, an observer engages with the artwork and learns how this abstract model (based off of biological neurons) works.

A neural network is a complicated mathematical expression that answers a single, focused question. Here, it asks: Should I participate in a particular study abroad program? To answer this query, the model processes five inputs through a neuron and receives a single output: yes or no.

The first component of this SciArt piece is a work of electronic literature. The audience chooses input values and weights, which are computed as they would be in an actual neural network to produce an appropriate result. The second component is an animated graphic visualizing the same neural network. It shows the calculations performed at each stage of the perceptron. In this way, the observer sees how the previous interactive queries fit into the big picture.*

This SciArt piece demonstrates how coding can make abstract Computer Science concepts engaging and more understandable to a general audience.

*Please note the model is simplified for demonstration purposes. It does not include the bias term added to the summation, and it does not explain backpropagation, a learning algorithm.