Graph Signal Processing on Gain Graphs

• Here we perform signal denoising using Gain normalized laplacian and compare our results with those using Hermitian Laplacian and state-of-the-art feasible Method.
• Empirically, we show that gain normalized Laplacian achieves state-of-the-art performance in terms of signal denoising, as measured by denosing error, and more importantly, offers rich theory backing the experimental observations. In particular, it outperforms state-of-the-art feasible Method, and in weighted case, performs better than the recently defined Hermitian Laplacian.
• We have obtained the orthogonal matrix U.mat from running the Feasible Method and using the code provided in this paper. All other experimentals results with corresponding code has been provided in the notebook Graph Signal Processing on Gain Graphs.ipynb.
• In case of any bugs or queries, please feel to drop me an email at the address provided in the manuscript.