At the time of initial development of gCNV, Nd4j did not provide native implementation of BLAS level-3 functions. Now that we have them (as of nd4j 0.6.0), it is desirable to migrate all BLAS level-3 ops to native Nd4j (as opposed to converting NDArrays to Apache matrices back and forth). This includes LU decomposition and matrix inversion ops.
At the time of initial development of gCNV, Nd4j did not provide native implementation of BLAS level-3 functions. Now that we have them (as of nd4j 0.6.0), it is desirable to migrate all BLAS level-3 ops to native Nd4j (as opposed to converting NDArrays to Apache matrices back and forth). This includes LU decomposition and matrix inversion ops.