Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
An artificial neural network (ANN) is a machine-learning model that processes inputs through interconnected mathematical units and produces an output. During training, it adjusts internal parameters—especially weights and biases—so its outputs better match examples or targets. The name is inspired by the brain, but an artificial network is a computational model, not a miniature biological one.
What an artificial neural network does
An ANN learns a mapping from inputs to outputs. Its inputs might represent features such as measurements, words, or image values; its output might be a category, a number, or another sequence of values. The network’s learned parameters determine how information is transformed along the way.
In a common feed-forward arrangement, data moves from an input layer through one or more hidden layers to an output layer. Each layer passes a transformed representation to the next. This is one useful way to picture a network, not a rule that every ANN must have exactly these layers or connect every unit to every other unit.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →How a network unit computes
A simplified unit takes incoming values, multiplies each by a weight, adds the results and a bias, then applies an activation function. One compact expression is:
#1 Best Overall
a = σ(Σ wᵢxᵢ + b)
- Inputs (xᵢ): the values arriving at the unit.
- Weights (wᵢ): adjustable values that control how strongly each input contributes.
- Bias (b): an adjustable offset added to the weighted sum.
- Activation function (σ): a function applied to the sum. In common networks, nonlinear activations let layers represent more than a simple linear transformation.
The result, a, can become an input to units in a later layer. The equation is a teaching simplification: real architectures may include other operations and connection patterns.
How training changes the network
In supervised learning, training examples include inputs and target outputs. A typical training loop separates three jobs: the loss measures prediction error, backpropagation calculates how the parameters contributed to that error, and an optimizer uses those gradients to update the parameters.
Rank #2
- Make a prediction: pass an input through the network in a forward pass.
- Measure the error: compare the prediction with its target using a loss function.
- Calculate gradients: use backpropagation and the chain rule to estimate how changing weights and biases would affect the loss.
- Update parameters: an optimizer, often a form of gradient descent, adjusts the parameters using the gradient information.
- Repeat and evaluate: continue across training examples and iterations, then check performance on data that was not used to fit the model.
That final evaluation matters because a network can fit its training examples well yet perform poorly on new ones. This is called overfitting: the model has learned patterns too specific to the training data to generalize reliably.
Recommended Free Tools
How neural networks relate to AI and deep learning
A useful introductory hierarchy is that artificial intelligence (AI) is the broad field, machine learning is a group of approaches within AI, and neural networks are one family of machine-learning models. Deep learning commonly refers to learning with neural networks that have multiple hidden layers. The terms are sometimes used with variation, but the distinctions prevent a few common misunderstandings:
Rank #3
- Not all AI systems use neural networks.
- Not all machine learning is deep learning.
- Not every neural network is deep; “deep” conventionally signals multiple hidden layers.
Why neural networks are not all built alike
“Neural network” names a broad family of models, not one fixed layout. Feed-forward networks pass information forward through layers; other designs use different connection patterns or operations to handle different kinds of input. For example, convolutional and attention-based designs process information in different ways. No architecture is best for every task, and this definition alone is not enough to choose a model for a particular application.
For a textbook introduction, OpenStax’s neural-networks chapter presents the layered model and its learned weights and biases. IBM’s neural-network explainer also describes training and overfitting; the NCBI Bookshelf chapter discusses neural networks and deep learning.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
