Neural Network Components

Understanding the components of a neural network's design

• Tensor - The key data structure which flows through the network

• Nodes - Individual units, connected using weights and grouped within layers

• Weights - Weighted connections between nodes

• Bias term - An offset parameter which is added to the weighted sum of the inputs

• Gradients - Represent the partial derivatives of the loss function with respect weights and biases

• Loss Function - Measures the neural network’s predictions against the target values

• Activation Function - Applied to the output of each node, introduces non-linearity into the network

• Regularisation - Used to prevent overfitting by adding a penalty to the loss function

• Optimisation - Used to update the model parameters based on the computed gradients