Generation processes of action potentials in a single neuron are described by Hodgkin-Huxley (HH)-type equations [1,2]. HH-type equations can describe electrical activities of various cells as well as neurons. It requires difficult physiological experiments to construct a model as HH-type equations. Doya et al. [3,4] proposed a method to estimate several parameters of HH-type equations using gradient-descent learning method in which many parameters can be estimated by the measurement of membrane potential only without further physiological experiments. In the present paper, we show the method of Doya et al. can be applied to the parameter estimation of pancreatic β-cell model which has a complicated bursting waveform.