Matrix Addition and Scalar Multiplication
The simplest way to manipulate a matrix is through addition and scalar scaling. These operations are 'entry-wise', meaning they act on every individual cell independently.
Matrix Addition
If two matrices have the same dimension, they can be added. This is used in AI to combine 'layers' of information or update parameters.
The Rule of Alignment
Matrix $A$ + Matrix $B$ is only possible if both are $m \times n$. It creates matrix $C$, where $c_{ij} = a_{ij} + b_{ij}$.
Scalar Multiplication
Multiplying a matrix by a single number (scalar) scales every element in the grid. This is how we 'dampen' or 'amplify' a signal in a neural network.
Learning Rate Application
When we multiply an update matrix by the Learning Rate, we are performing scalar multiplication to control the speed of the model's growth.