Large residual norm detected. This is likely due to bad condition of the input matrix. {0,number,integer} of the {1,number,integer} requested eigenvalues converged. Eigenvalues that did not converge are NaN. {0,number,integer} of the {1,number,integer} requested eigenvalues converged. Specify convergence flag output to determine the values that did not converge. {0,number,integer} of the {1,number,integer} requested eigenvalues converged. Eigenvalues that did not converge are removed. Ambiguous syntax: First and second arguments are both scalars.\n Use EIGS(A,[],K,SIGMA) to specify standard eigenvalue problem, or EIGS(A,B,K,SIGMA) to specify generalized eigenvalue problem. Invalid data type. Input matrices must be double-precision. cholB field in the options structure specified, but B is not upper triangular. ''IsCholesky'' specified, but B is not upper triangular. Input matrices must be the same size for generalized problems. Unexpected output of function handle. The expected output is a column vector of length {0} and of type double. cholB field in the options structure must be a logical scalar. ''IsCholesky'' value must be a logical scalar. issym field in the options structure must be a logical scalar. isreal field in the options structure must be a logical scalar. ''IsFunctionSymmetric'' value must be a logical scalar. fail field in the options structure must be ''replacenan'', ''keep'', or ''drop''. ''FailureTreatment'' value must be ''replacenan'', ''keep'', or ''drop''. p field in the options structure must be a positive integer less than or equal to n. ''SubspaceDimension'' value must be a positive integer less than or equal to n. spdB field in the options structure must be a logical scalar. ''IsSymmetricDefinite'' value must be a logical scalar. permB field in the options structure must be a permutation of 1:n. ''CholeskyPermutation'' value must be a permutation of 1:n. tol field in the options structure must be a strictly positive real scalar. ''Tolerance'' value must be a strictly positive real scalar. Sigma can be a scalar double or one of these options: ''largestabs'', ''smallestabs'', ''largestreal'', ''smallestreal'', ''bothendsreal'', ''largestimag'', ''smallestimag'', or ''bothendsimag''. When the value of the cholB field in the options structure is true, the value of the spdB field must also be true. When ''IsCholesky'' value is true, ''IsSymmetricDefinite'' value must also be true. Incorrect number of input arguments. Each parameter name must be followed by a corresponding value. When the first argument is a function handle, the second argument must be the size of the matrix represented by the function handle. First argument must be a double matrix or a function. Number of eigenvalues requested must be a nonnegative integer. Number of eigenvalues requested must be a nonnegative integer. Instead, it was {0}. Number of eigenvalues requested must be a nonnegative integer. Instead, it was a struct. Size of problem, ''n'', must be a positive integer. First argument must be a square matrix or a function handle. Parameter name must be ''IsFunctionSymmetric'', ''Tolerance'', ''MaxIterations'', ''SubspaceDimension'', ''StartVector'', ''IsSymmetricDefinite'', ''IsCholesky'', ''CholeskyPermutation'', or ''FailureTreatment''. maxit field in the options structure must be a positive integer. ''MaxIterations'' value must be a positive integer. Matrix B is not symmetric. Value of spdB field in options structure must be false. Matrix B is not symmetric. ''IsSymmetricDefinite'' value must be false. Matrix B is not positive definite. ''IsSymmetricDefinite'' value must be false. p field in the options structure must be > k+1. ''SubspaceDimension'' value must be > k+1. disp field in the options structure must be 0 or 1. ''Display'' value must be 0 or 1. Too many input arguments. v0 field in the options structure must be a double column vector of length matching the problem size. ''StartVector'' value must be a double column vector of length n. Ignoring issym field in the options structure since the first input is not a function handle. Ignoring ''IsFunctionSymmetric'' since the first input is not a function handle. Ignoring permB field in the options structure since B is not its sparse Cholesky factor. Ignoring ''CholeskyPermutation'' since B is not its sparse Cholesky factor. The first input matrix, shifted by sigma, is singular, therefore sigma is an eigenvalue of the first input matrix. Consider specifying a perturbed numeric sigma value to improve the condition of the matrix. First input matrix is singular. Consider specifying a small nonzero numeric sigma value instead of ''smallestabs'' to improve the condition of the matrix. Singular second input matrix is only supported when sigma is ''SM'' or a scalar double. First input matrix is close to singular or badly scaled (RCOND = {0}) and results may be inaccurate. Consider specifying a small nonzero numeric sigma value instead of ''smallestabs'' to improve the condition of the matrix. Second input matrix B is close to singular or badly scaled. RCOND = {0}. Results may be inaccurate. The first input matrix, shifted by sigma, is close to singular or badly scaled (RCOND = {0}) and results may be inaccurate. Consider specifying a perturbed numeric sigma value to improve the condition of the matrix. The starting vector v0 could not be normalized. It may be zero or contain non-finite values. The starting vector v0 is in the nullspace of matrix B. Choose a different starting vector. Unable to build an orthogonal subspace because the problem is ill-conditioned. For generalized problems, B may have low rank. Input matrix is badly conditioned. It might contain Inf or NaN. Simple eigenvalue problem A*x = lambda*x Generalized eigenvalue problem A*x = lambda*B*x The eigenvalue problem is real symmetric. The eigenvalue problem is real non-symmetric. The eigenvalue problem is complex Hermitian. The eigenvalue problem is complex non-Hermitian. Matrix B is symmetric positive (semi-)definite. Matrix B is not symmetric positive (semi-)definite. Matrix B is Hermitian positive (semi-)definite. Matrix B is not Hermitian positive (semi-)definite. Computing {0} eigenvalues of type ''{1}''. Computing {0} eigenvalues closest to {1}. Function handle Afun defines the {0}-by-{1} matrix A. The function handle returns {2}. Parameters passed to Krylov-Schur method:\n Maximum number of iterations: {0}\n Tolerance: {1}\n Subspace Dimension: {2} Compute EIGS by calling EIG, because subspace dimension is equal to problem size. Find eigenvalues of A*x = lambda*x. Find eigenvalues of {0}\\x = mu*x.\nCompute decomposition of {1}... Find eigenvalues of R''\\(A*(R\\x)) = lambda*x, with B = R''*R. Find eigenvalues of R*{0}\\(R''*y) = mu*y, with y = R*x, B = R''*R.\nNo need to compute R, as it is only used implicitly.\nCompute decomposition of {1}... Find eigenvalues of B\\(A*x) = lambda*x.\nCompute decomposition of B... Find eigenvalues of {0}\\(B*x) = mu*x.\nCompute decomposition of {1}... Start of Krylov-Schur method Krylov-Schur failed to build a B-orthogonal subspace. Try a different strategy:\n Find eigenvalues of (A - sigma*B)\\(B*x) = mu*x. Iteration {0}: {1} of {2} eigenvalues converged. Smallest non-converged residual {3} (tolerance {4}). Iteration {0}: {1} of {2} eigenvalues converged. To check if eigenvalue multiplicities were missed, restart the method, looking for k+1 eigenvalues. No additional multiple eigenvalues found. Successful return. Some additional multiple eigenvalues were found. Restart again to check for more.