An artificial neural network (ANN) is a computational system inspired by the interconnected structure of biological neurons. It consists of processing units called artificial neurons, which are arranged in layers and connected through weighted links.
A typical feedforward neural network contains an input layer, one or more hidden layers, and an output layer. The input layer receives features, hidden layers transform information, and the output layer produces predictions. Each connection has a weight that determines how strongly one neuron influences another. Neurons may also have biases that shift their activation values.
An artificial neuron calculates a weighted sum of its inputs, adds a bias, and applies an activation function. The activation function introduces nonlinearity, allowing the network to learn relationships more complex than a simple linear transformation.
During training, the network generates predictions through forward propagation. A loss function measures the difference between predictions and target values. Backpropagation calculates gradients, and an optimization algorithm updates the weights and biases to reduce the loss.
Different ANN architectures are designed for different tasks. Feedforward networks are used for general prediction, convolutional neural networks are especially useful for spatial data such as images, and recurrent or attention-based architectures can process sequential information. The architecture should be selected according to the structure of the data and the learning objective.
Artificial Neural Networks
Artificial neural networks are computational models made of interconnected artificial neurons that process information and learn relationships between inputs and outputs.
z = Σᵢ₌₁ⁿ wᵢxᵢ + b; a = f(z)
Explanation
Variables
| Symbol | Meaning |
|---|---|
xᵢ | Input value i |
wᵢ | Weight associated with input i |
b | Neuron bias |
z | Weighted input before activation |
f | Activation function |
a | Neuron output or activation |
n | Number of input connections |
Example or application
A neural network receives information about a house, including its area, number of bedrooms, and location-related features. Hidden layers process these inputs, and the output layer predicts the estimated house price. Training adjusts the network's weights and biases using examples with known prices.
Text
An artificial neural network contains interconnected artificial neurons organized into layers. Input layers receive data, hidden layers process information, and output layers generate predictions. Each neuron combines weighted inputs with a bias and applies an activation function. During training, the network adjusts its weights and biases to reduce prediction errors. Different architectures are designed for different kinds of data and tasks.
Key Points
Artificial neural networks contain interconnected computational neurons.
Weights determine the influence of connections.
Biases shift the weighted input.
Activation functions introduce nonlinear behavior.
Hidden layers learn intermediate representations.
Training adjusts weights and biases using optimization.
Network architecture depends on the task and data.
Applications
Classification and regression
Image recognition
Speech processing
Financial forecasting
Medical diagnosis support
Recommendation systems
Learning Outcomes
Identify the main components of an artificial neural network.
Explain how an artificial neuron calculates its output.
Describe the roles of weights, biases, and activation functions.
Explain how an ANN learns during training.
Practice Questions
What are the main layers of a feedforward neural network?
What is the role of a weight in an artificial neural network?
Why are activation functions necessary?
Related Concepts
Biological Neurons and Artificial Neurons
Perceptrons
Neural Network Architecture
Activation Functions
Backpropagation
Common Mistakes
Confusing the input layer with the hidden layers.
Assuming that weights and biases have the same role.
Thinking that a network without nonlinear activation can learn arbitrary nonlinear relationships.
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press.
- Haykin, S. (2009). Neural Networks and Learning Machines. Pearson.
Suggest a correction or source · More in Artificial Intelligence & Machine Learning