Neural network training via quadratic optimization

A technique using quadratic optimization is proposed to find the weights of a single neuron, or a single-layer neural network, and is extended to the multilayer neural network. The weights for a neuron are found by minimizing a cost function that is quadratic with respect to the neutron's weight and are used as an answer for minimizing a cost function that is quadratic with respect to the neuron's outputs. By backpropagating, the output error through the neural network's layers, the method is extended to the multilayer neural network. The quadratic optimization algorithm for the multilayer neural network tends to work best for classification problems and to achieve successful results in a single iteration.<>