Optimal choice of descent steps in gradient methods of solution of inverse heat-conduction problems

Modifications are proposed for the methods of steepest descent and conjugate gradients for the solution of multiparameter inverse problems in heat conduction.Notation A, B, L linear operators - u element of the solution space U - f exact initial data - f error in the initial data - value of the error in the initial data - A^–1 inverse operator - u^(k)() k-th derivative of the function u - _i() polynomials of degree i–1 - A^*, B^*, L^* operators conjugate to the operators A, B, L - J(g) discrepancy functional - J'g gradient of the discrepancy functional - _n ⁱ depth of descent with respect to the i-th component of the antigradient of the discrepancy in the n-th iteration - _m length of the observation interval Translated from Inzhenerno-Fizicheskii Zhurnal, Vol. 39, No. 2, pp. 264–269, August, 1980.