System enhancement methodology via matrix multiplication efficiency speedup using sparse basis approach

1. A computer-implement method of machine learning executed by one or more processors for feature extraction and classification, the method comprising acts of:

identifying a first matrix and an intermediate matrix to be multiplied for producing a sparse basis matrix, wherein the first matrix is a matrix of input data, and the intermediate matrix is a filter used in the machine learning;

applying the Gauss-Jordan Elimination (GJE) to the intermediate matrix to obtain a second matrix in reduced row echelon form (RREF), wherein the second matrix includes:

the first non-zero term in the non-zero row being equal to 1, and being defined as a pivot;

if a column contains a pivot, then the other entries in the column being zero;

the location of the pivot in each non-zero row being from left to right and from the first row to last row; and

zero rows being at the bottom of the second matrix;

removing all zero rows from the second matrix for producing the sparse basis matrix; and

determining an output matrix by multiplying a third matrix based on the sparse basis matrix, wherein the third matrix is a matrix denoting the coefficients required for a linear combination for the sparse basis matrix.

2. The method as claimed in claim 1 , wherein the act of determining an output matrix by multiplying the third matrix based on the sparse basis matrix further comprising acts of

creating a pseudoinverse matrix according to the sparse basis matrix, wherein the sparse basis matrix has linearly independent rows; and

the third matrix is determined by executing the matrix multiplication of the pseudoinverse matrix and the intermediate matrix.

3. The method as claimed in claim 1 , wherein the act of applying the Gauss-Jordan Elimination (GJE) to the intermediate matrix further comprises acts of

swapping rows in the intermediate matrix to the leftmost element in the first row being non-zero, wherein the value of the leftmost element is defined as the pivot, and the row contains the pivot is defined as the pivot row;

all elements of the pivot row being divided by the value of the pivot that results the value of the pivot being equal to 1;

multiple pivot rows being added to other rows to eliminate all elements in the columns containing pivots, wherein except for the pivots, all elements in the columns containing pivots become zeros; and

repeating the previous act for each row starting from the second row until rows being processed or no non-zero row being located below the pivot row.

4. The method as claimed in claim 1 , further comprising act of

applying a low rank approximation to the intermediate matrix.

5. A system comprising:

a memory for storing matrix data; and

one or more processors being configured to perform matrix multiplication instruction to the matrix data stored in the memory using sparse basis approach in a convolutional neural network (CNN), wherein the operation of the processor comprising acts of

identifying a first matrix and an intermediate matrix to be multiplied for producing a sparse basis matrix, wherein the first matrix is a matrix of input data, and the intermediate matrix is a filter of the CNN;

applying the Gauss-Jordan Elimination (GJE) to the intermediate matrix to obtain a second matrix in reduced row echelon form (RREF); and

determining an output matrix by multiplying a third matrix based on the sparse basis matrix, wherein the third matrix is a matrix denoting the coefficients required for a linear combination for the sparse basis matrix.