Artificial intelligence / machine learning model · BEGINNER LESSON
Neural network
A neural network is a computer model made of connected layers of artificial neurons, also called nodes, that process inputs and produce an output. It uses weights and thresholds to decide how signals move through the network, and it can be trained from data.
What it is and what is measured
A neural network is a model in artificial intelligence (AI), which means a computer-based system designed to perform tasks linked to learning or prediction. In a neural network, the main measurable parts are the inputs entering the model, the weights on connections between nodes, the thresholds that control activation, and the output the model produces. The network is built from layers: an input layer, one or more hidden layers, and an output layer. The layers are arranged in order so data passes from one layer to the next.
[1] Reference Guide on Artificial Intelligence - NCBI - NIH [2] Table 5.
How it works
Each node, or artificial neuron, receives data from earlier nodes. The node combines the incoming signals using programmed rules, then compares the result with a threshold. If the result is above that threshold, the node activates and sends data onward; if not, the signal stops there. Training is the process of adjusting the network so its outputs better match examples in the training data. Because the model learns from examples, the quality and amount of training data matter a lot.
[1] Reference Guide on Artificial Intelligence - NCBI - NIH [2] Table 5.
Why layers matter
Layers let the model build from simple to more complex patterns. The input layer accepts the raw data, hidden layers transform it, and the output layer gives the final result. A network with three or more hidden layers is often called a deep learning model. This layered design can help the model represent complex, nonlinear relationships, where the connection between input and output is not a straight-line pattern.
[1] Reference Guide on Artificial Intelligence - NCBI - NIH [2] Table 5.
Limits and weaknesses
Neural networks are often described as opaque or a black box because it may be hard to see exactly how the model reached a decision. They can also need large amounts of data and computation, and their design may involve trial and error. If the training data are limited, shifted, noisy, or biased, the model’s output may become less reliable. A network can also overfit, meaning it becomes too tuned to its training examples and handles new cases poorly.
[1] Reference Guide on Artificial Intelligence - NCBI - NIH [2] Table 5.
SEE THE IDEA
How a neural network turns data into an output
Receives the starting data points.
Apply weights, thresholds, and activations to transform the data.
Produces the final classification, score, or prediction.
Data moves forward through connected layers; each layer changes the information before passing it on.
A neural network processes data step by step through layers, with each node deciding whether to pass a signal onward.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.