Abstract: A novel adaptive moment estimation algorithm based on maximum correntropy criterion (AdamMCC) was proposed to improve the prediction accuracy of chaotic sequence in the non-Gaussian noises． The maximum correntropy criterion was chosen as the cost function of the proposed AdamMCC owing to its robustness against non- Gaussian noises． The first and second moments of gradients in the cost function were used to adjust the weight of the parameters in the algorithm，which provides a better search direction for the optimal weight，thus improved the prediction performance of the proposed AdamMCC． Simulations on the prediction of the Mackey-Glass chaotic time sequence and Lorenz chaotic time sequence illustrate that the proposed AdamMCC can achieve better prediction performance with affordable computational complexity and maintain robustness，compared with the least mean square algorithm (LMS)，the maximum correntropy criterion algorithm (MCC)，and the fractional-order maxi- mum correntropy criterion algorithm (FMCC) in the presence of non-Gaussian noises．

Key words: non-Gaussian noise, chaotic time sequence, prediction accuracy, adaptive moment estimation, maximum correntropy criterion, robustness