“…Numerous machine learning and signal processing methods have been investigated to solve the NILM problem, including stochastic finite state machines (Hidden Markov Model and its variants) [19,20], support vector machine (SVM) [21], decision tree [21,22], dynamic time warping [22], k-nearest neighbours (K-NN) [16,21,[23][24][25], sparse coding [26,27], neural networks [9,[28][29][30], graph signal processing (GSP) [18,31], and optimisation via learning of appliance models and occupancy information [32][33][34][35][36]. Furthermore, advanced hybrid approaches have been proposed for improving core NILM performance, e.g., k-means clustering based training followed by disaggregation using SVM [37], GSP with result refinement using simulated annealing [17], deep neural network utilised to learn deeper and multiple layers of sparse signal representation [38].…”