What is a Random Matrix?
A random matrix is any matrix square (nxn) or rectangular matrix (mxn) that has the property of completely randomized data, selected from some specified distribution, inside of the matrix. This is often used in the initialization of weights in machine learning. It is often the case that there is no prior information to encode before learning and no basis to justify a specific weight matrix. A random matrix is used in these cases, often with very small values so as not to prejudice the models. Below is an example of a random matrix with the specified distribution of integers from one to five:
All properties are completely random from the number of occurrences of a specific datum to the placement inside of the matrix. However, given a large enough matrix, the datum inside the matrix will have a normal distribution.