US Patent 11677449 Eigensystem optimization in artificial neural networks

Systems, methods, computer program products, and devices provide for computing an eigensystem from a first data set; computing updated eigenvalues that approximate an eigensystem of at least a second data set based on the eigensystem of the first data set; and evaluating a plurality of features in each of the first and at least second data sets using a cost function. The cost function uses fewer than the total number of eigenvalues and can include a condition number. The cost function can perform a coarse approximation of the eigenvalues to de-select at least one of the data sets. This can be useful for learning and/or online processing in an artificial neural network.