Generalized Delta Rule

Understanding the Generalized Delta Rule in Neural Networks

The Generalized Delta Rule, also known as backpropagation, is a fundamental algorithm in the field of neural networks and deep learning. It is the cornerstone of training for many types of neural networks, including feedforward networks used for tasks such as image and speech recognition.

What is the Generalized Delta Rule?

The Generalized Delta Rule is an algorithm used to update the weights in an artificial neural network during training. It is a generalization of the Perceptron's learning rule and is designed to minimize the error in the output of a neural network through gradient descent.

The rule is based on the concept of calculating the gradient of the loss function with respect to each weight in the network and adjusting the weights in the opposite direction of the gradient, hence "backpropagating" the error from the output layer back to the input layer.

How the Generalized Delta Rule Works

The process of applying the Generalized Delta Rule involves several key steps:

  1. Forward Pass: An input is fed through the neural network to compute the output.
  2. Error Calculation: The output of the network is compared to the desired output, and the error is calculated using a loss function.
  3. Backward Pass:

    The gradient of the loss function with respect to each weight is computed by applying the chain rule of calculus, starting from the output layer and moving towards the input layer.

  4. Weight Update: The weights are updated by subtracting a fraction of the gradient, where the fraction is determined by the learning rate.

The algorithm is iterative, with the forward and backward passes being repeated for many epochs or until the network's performance reaches a satisfactory level.

The Importance of the Learning Rate

The learning rate is a hyperparameter that controls the size of the weight updates during training. A learning rate that is too high can cause the network to overshoot the minimum of the loss function, while a learning rate that is too low can result in slow convergence or getting stuck in local minima.

Mathematical Formulation of the Generalized Delta Rule

The mathematical expression for the weight update in the Generalized Delta Rule is as follows:


  • Δwij is the change in weight from node i to node j,
  • η is the learning rate,
  • δj is the error term for node j, and
  • xi is the input from node i.

The error term δj is computed differently for output nodes and hidden nodes, reflecting the role of each node in the network's prediction.

Applications of the Generalized Delta Rule

The Generalized Delta Rule is widely used in various applications, including:

  • Training deep neural networks for image and speech recognition,
  • Optimizing neural networks for natural language processing tasks,

  • Improving the performance of reinforcement learning agents, and

  • Any other domain where neural networks are employed for prediction or classification.


The Generalized Delta Rule is a powerful tool for training neural networks, enabling them to learn complex patterns and make accurate predictions. Its ability to efficiently propagate errors backward through the network and adjust weights accordingly is what makes it a staple in the deep learning community.


  • Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323(6088), 533-536.
  • Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.

Please sign up or login with your details

Forgot password? Click here to reset