A neural network is a machine learning model or programme that makes decisions in a similar manner to the human brain. It does this by using processes that mimic the way biological neurons work together to identify events, weigh options and arrive at conclusions.
Neural networks consist of layers of interconnected nodes, or artificial "neurons," that process data. These neurons are organised into three types of layers:
Input Layer: Receives the raw data and passes it to the network.
Hidden Layer(s): Process the data by adjusting weights through training.
Output Layer: Produces the result.
Each "neuron" in a neural network can be thought of like a decision-maker. It looks at the information it receives and based on certain rules (its "weight" and "threshold (or bias)"), it decides whether to pass that information along or stop it. If the information is strong enough (exceeds the threshold), the neuron sends it to the next step. If not, it doesn't pass anything forward.
A Convolutional Neural Network; a CNN, is a is a type of deep learning model that is commonly used for tasks involving images and videos. They are sophisticated models that focus on hierarchical features and ignore insufficient details of images.
Neuron: An artificial neural network consists of artificial neurons that work together to solve problems, much like organic neurons in the human brain, which form interconnected networks to process information
Threshold: A threshold is a limit that determines whether the neuron will fire or activate. After the weighted inputs are summed up, this sum is compared to the threshold. If the sum is greater than or equal to the threshold, the neuron activates and passes data forward. If it's below the threshold, the neuron stays inactive, and nothing is passed on.
Example: A threshold is like a gate. Only if the input is strong enough (above the threshold) will the gate open and allow the information to pass through.
Weight: A weight is a number that represents the importance of the connection between two neurons. When a neuron receives input, each input is multiplied by its corresponding weight. The higher the weight, the more influence that input has on the neuron's output. Essentially, weights control how much one neuron contributes to the next.
Example: Think of weight like the volume control on a speaker. It adjusts how loud or quiet a certain input is before it moves forward in the network.
Bayliss, E (2024) Neural Network [Digital Artwork]
Bayliss, E (2025) Convolutional Neural Network - Exploratory [Digital Artwork]
IBM (2023). What Are Neural Networks? [online] www.ibm.com. Available at: https://www.ibm.com/topics/neural-networks.
