This paper presents an unsupervised learning approach for simultaneous sample and feature selection, which is in contrast to existing works which mainly tackle these two problems separately. In fact the two tasks are often interleaved with each other: noisy and high-dimensional features will bring adverse effect on sample selection, while informative or representative samples will be beneficial to feature selection. Specifically, we propose a framework to jointly conduct active learning and feature selection based on the CUR matrix decomposition. From the data reconstruction perspective, both the selected samples and features can best approximate the original dataset respectively, such that the selected samples characterized by the features are highly representative. In particular, our method runs in one-shot without the procedure of iterative sample selection for progressive labeling. Thus, our model is especially suitable when there are few labeled samples or even in the absence of supervision, which is a particular challenge for existing methods. As the joint learning problem is NP-hard, the proposed formulation involves a convex but non-smooth optimization problem. We solve it efficiently by an iterative algorithm, and prove its global convergence. Experimental results on publicly available datasets corroborate the efficacy of our method compared with the state-of-the-art.
Topological insulators, with metallic boundary states protected against time-reversal-invariant perturbations, are a promising avenue for realizing exotic quantum states of matter including various excitations of collective modes predicted in particle physics, such as Majorana fermions and axions. According to theoretical predictions, a topological insulating state can emerge from not only a weakly interacting system with strong spin-orbit coupling, but also in insulators driven by strong electron correlations. The Kondo insulator compound SmB6 is an ideal candidate for realizing this exotic state of matter, with hybridization between itinerant conduction electrons and localized $f$-electrons driving an insulating gap and metallic surface states at low temperatures. Here we exploit the existence of surface ferromagnetism in SmB6 to investigate the topological nature of metallic surface states by studying magnetotransport properties at very low temperatures. We find evidence of one-dimensional surface transport with a quantized conductance value of $e^2/h$ originating from the chiral edge channels of ferromagnetic domain walls, providing strong evidence that topologically non-trivial surface states exist in SmB6.
