1: \begin{abstract}

2:   Power iteration is a fundamental algorithm in data analysis. It extracts the eigenvector corresponding to the largest eigenvalue of a given matrix. Applications include ranking algorithms, principal component analysis (PCA), among many others. Certain use cases may benefit from alternate, non-linear power methods with low complexity.\ In this paper, we introduce multiplication-avoiding power iteration (MAPI). MAPI replaces the standard $\ell_2$ inner products that appear at the regular power iteration (RPI) with multiplication-free vector products, which are Mercer-type kernels that induce the $\ell_1$ norm. For an $n\times n$ matrix, MAPI requires $n$ multiplications, while RPI needs $n^2$ multiplications per iteration. Therefore, MAPI provides a significant reduction of the number of multiplication operations, which are known to be costly in terms of energy consumption. We provide applications of MAPI to PCA-based image reconstruction as well as to graph-based ranking algorithms. When compared to RPI, MAPI not only typically converges much faster, but also provides superior performance.

3: % In this article, we propose a multiplication avoiding power iteration method based on a kernel-related with the $\ell_1$-norm. The kernel is computed using sign calculations and absolute value comparisons. It is energy efficient as the kernel computation does not require any multiplications. We use Kernel Power Iteration (KPI) method in a number of applications. We compare the KPI-based PCA with the regular PCA and $\ell_1$-PCA in an image reconstitution method. The KPI-PCA reaches 0.90 dB higher than the regular PCA with conventional power iteration on reconstructed images. Then, to show its convergence speed faster than the conventional power iteration, we test the algorithm on a synthetic dataset. Finally, we show that our min power iteration can be implemented into the Google PageRank algorithm to accelerate the convergence.

4: \end{abstract}

5: