An artificial neural network is a computational model inspired by the structure of the brain's biological neurons. It consists of connected nodes organized into layers: the input layer receives the data, the hidden layers process it, and the output layer produces the result.
When there are many hidden layers, the network is called deep — this is exactly what deep learning is based on, which today is applied to estimate biological age.
How it is built
Each artificial neuron (node) receives weighted signals from the nodes connected to it, sums them and passes them through an activation function. The strength of the connections is set by weights — numbers that change during training. The connections between nodes model synapses, and the nodes themselves model simplified neurons.
The layers build a hierarchy: from simple features to complex ones. This is how the network learns to transform "raw" data into a useful answer — for example, an age estimate or the probability of a disease.
How it learns
Learning is the adjustment of the weights so that the prediction error becomes minimal. The main method is backpropagation: the network computes the gradient of the error and corrects the weights, moving from the output to the input. The process is repeated until accuracy stops improving.
Why this is used in aging research
Neural networks are trained to predict a person's age from DNA methylation, from blood values and from medical images. The difference between predicted and chronological age is interpreted as age acceleration and is regarded as a biomarker of aging. This is an actively developing field: the models show high accuracy in estimating age, but their clinical value as predictors of disease is still being refined.
Calculate using this marker
Calculators where Neural network (AI model) is used directly:
Frequently asked questions
Does an artificial neural network work like a real brain?
Only in the general idea. It is inspired by neurons, but it is a mathematical model of nodes and weights, not biological tissue.
What are weights and network training?
Weights are the strength of the connections between nodes. Training adjusts them (by backpropagation) so that the network's predictions become more accurate.