Delta Rule

Definition of Delta Rule

The Delta Rule, also known as the Widrow-Hoff Learning Rule, is an algorithm used in training artificial neural networks through supervised learning techniques. It iteratively adjusts the weights of a network to minimize the difference between desired and actual output values. The rule helps optimize the performance of the network by employing gradient descent to find the minimum error or cost function.


The phonetics of the keyword “Delta Rule” would be:/ˈdɛltə ˈrul/It is pronounced as:- “delta” is pronounced as “dell-tuh”- “rule” is pronounced as “rool”

Key Takeaways

  1. The Delta Rule is a gradient descent learning algorithm used in neural networks to minimize the error between the predicted output and the actual output by adjusting the weights of the input nodes.
  2. It is a supervised learning method, meaning that the model learns through examples consisting of input-output pairs with known outcomes. This algorithm is mainly applied to single-layer neural networks, such as the perceptron and the Adaline.
  3. The main limitation of the Delta Rule is that it can find solutions only for linearly separable problems. Therefore, it is not suitable for complex, nonlinear problems, which require multiple hidden layers and the use of backpropagation algorithms.

Importance of Delta Rule

The Delta Rule, also known as the Widrow-Hoff learning rule or Least Mean Square algorithm (LMS), is important in the field of technology, specifically within artificial intelligence and neural networks, as it serves as a fundamental learning algorithm for adjusting the weights of artificial neurons.

Essentially, the Delta Rule is a form of supervised learning that enables neuron weights in a neural network to be updated to minimize the error between the predicted output and the actual output.

It relies on the concept of gradient descent to find the optimal weight values, which, in turn, help to improve the overall performance and accuracy of the model.

As such, the Delta Rule plays a crucial role in the development of efficient and powerful machine learning models, and ultimately contributes to a wide range of technological advancements in various industries and applications.


The Delta Rule, also known as the Widrow-Hoff learning rule or the Least Mean Square (LMS) algorithm, serves a crucial purpose in the realm of machine learning and artificial neural networks. It is employed to modify and optimize the synaptic weights of a neural network, which in turn enhances the learning accuracy of the model. The primary objective of the Delta Rule is to minimize the difference between the predicted output by the model and the actual output (or target) in a supervised learning environment.

This rule is particularly significant in training processes like regression and classification, where it facilitates the fine-tuning of neural networks by providing a concrete guideline for updating the weights through iteration. The Delta Rule operates by adjusting the weights proportional to the negative gradient of the mean squared error (MSE) in relation to the corresponding weights. Specifically, it employs a gradient descent algorithm that strives to reach the lowest point in the error surface, which symbolizes the minimal error between the estimated and the actual output values.

As the iterations progress, the algorithm successively converges to the optimum weight values that produce the smallest error. This continual process of optimization allows for an enhanced network performance and improved prediction accuracy. In summary, the Delta Rule substantially contributes to the development of effective and accurate machine learning models, fostering better understanding and decision-making processes in various real-world applications.

Examples of Delta Rule

The Delta Rule, also known as the Widrow-Hoff learning rule or the Least Mean Square (LMS) algorithm, is an essential concept in artificial neural networks and adaptive filter theory. It is primarily used for supervised learning and parameter adjustment in various fields and applications. Here are three real-world examples where the Delta Rule is applied:

Echo Cancellation in Telecommunications:In telecommunication systems, when a speaker’s voice is transmitted, it may cause an echo due to reflection from various network elements. This echo can lead to a decrease in voice quality for the listener. Algorithms based on the Delta Rule, such as the LMS algorithm, are widely used in adaptive echo cancelers to reduce the amplitude of the echo, thus improving the communication quality.

Noise Cancellation in Audio Systems:Active noise-canceling headphones and audio systems utilize adaptive filters based on the Delta Rule to effectively cancel the ambient noise. Microphones inside the headphone capture the external sounds, and the adaptive filter generates a counter sound wave that is “out-of-phase” with the external noise. By playing back this counter wave, the ambient noise is significantly reduced, resulting in a better listening experience.

Financial Market Prediction:Adaptive algorithms based on the Delta Rule can be used to predict stock prices or exchange rates by training neural networks on historical financial data. The LMS algorithm is utilized to minimize the prediction error and continuously adjust the model to account for new data points. This approach can potentially help investors make more informed decisions in a constantly changing financial landscape.

Delta Rule FAQ

1. What is the Delta Rule?

The Delta Rule, also known as the Widrow-Hoff learning rule or the least mean squares (LMS) algorithm, is a supervised learning algorithm in artificial neural networks used to optimize connection weights between nodes based on the error produced by the network’s output. It can be applied to both single-layer and multi-layer neural networks.

2. How does the Delta Rule work?

The Delta Rule works by iteratively updating the weights of the network to minimize the error between the desired output and the actual output. It achieves this by calculating the gradient of the error with respect to each weight and adjusting the weights based on the negative gradient multiplied by a learning rate. This process is repeated multiple times until the error converges to an acceptable minimum value.

3. What is the relation between the Delta Rule and the Perceptron Rule?

Both the Delta Rule and the Perceptron Rule are algorithms for adjusting the weights of a neural network, but they differ in the learning rule and the networks they can be applied to. The Perceptron Rule is designed for single-layer networks and relies on a simple threshold function, whereas the Delta Rule can be applied in both single-layer and multi-layer networks and uses a gradient-based approach to adjust the weights. The Delta Rule can be seen as the generalization of the Perceptron Rule.

4. Can the Delta Rule be used for unsupervised learning?

The Delta Rule is primarily designed for supervised learning algorithms, where a known set of input-output pairs are provided to the network for training. It may be possible to adapt the Delta Rule for unsupervised learning tasks by modifying the error calculation or using additional techniques, but it is not considered a standard method for unsupervised learning.

5. When should I use the Delta Rule in neural networks?

You should use the Delta Rule in situations where you have a supervised learning problem, a known set of input-output pairs for training, and the optimization and adaptation of connection weights are required. The Delta Rule can be particularly useful for training feedforward neural networks, including multi-layer ones, and its gradient-based approach makes it suitable for dealing with complex, non-linear problems.

Related Technology Terms

  • Gradient Descent
  • Artificial Neural Networks (ANNs)
  • Backpropagation
  • *

  • Error Minimization
  • *

  • Weight Adjustment

Sources for More Information


About The Authors

The DevX Technology Glossary is reviewed by technology experts and writers from our community. Terms and definitions continue to go under updates to stay relevant and up-to-date. These experts help us maintain the almost 10,000+ technology terms on DevX. 

See our full expert review panel.


About Our Editorial Process

At DevX, we’re dedicated to tech entrepreneurship. Our team closely follows industry shifts, new products, AI breakthroughs, technology trends, and funding announcements. Articles undergo thorough editing to ensure accuracy and clarity, reflecting DevX’s style and supporting entrepreneurs in the tech sphere.

See our full editorial policy.

Technology Glossary

Table of Contents

More Terms