<p>The existence of the outlier increases the error variances and results in the poor identification performance. This paper presents a heterogeneous mixed kernel correntropy-based robust iterative identification algorithm for the nonlinear systems with outliers. Instead of using a single kernel or the mixture of multiple homogeneous kernels in the correntropy criterion, two types of different kernels (i.e., the Gaussian kernel and the Cauchy kernel) are combined as the kernel function to increase the flexibility and improve the estimation accuracy. Moreover, a heterogeneous mixed kernel correntropy-based hierarchical iterative algorithm is developed to lower the computational burden. The proposed algorithms do not require the prior distribution knowledge of the outliers. The theoretical analysis shows that when the bandwidth of the Gaussian kernel is proportional to the square of the bandwidth of the Cauchy kernel, the convergence of the proposed algorithms can be guaranteed based on the contraction mapping theorem. The simulation examples exhibit the effectiveness of the proposed algorithms.</p>