[en] Highlights: • A two-step selection strategy was proposed, using three selection algorithms to evaluate informative variables. • iPLS was introduced to efficiently eliminate uninformative variables without destroying synergistic effects among spectrum. • Combining variable combination population analysis and permutation analysis, a new selection strategy was proposed. • Experimental results showed the superiority of iPCPA and its wide application potential. As one of the most important preprocessing procedures in spectral detection, wavelength selection approaches play an irreplaceable role in reducing the model overfitting and prediction errors. In this paper, we propose a two-step wavelength selection method called interval permutation combination population analysis (iPCPA), which improves the selective performance by combining three different wavelength selection algorithms. First, interval partial least squares (iPLS) is used as the rough selection step to efficiently exclude the uninformative variables in the spectrum, which reduces the variable space and ensures that the following selection step can focus on selecting informative variables. Then, permutation combination population analysis (PCPA) is proposed, which introduces the core idea of permutation analysis into the variable combination population analysis (VCPA) and hence improves its ability in evaluating the importance of informative variables. Six state-of-the-art wavelength selection methods are used to compare with iPCPA and their performances are tested by using three real spectral datasets: corn, beer, and soil datasets. The final experimental results prove that iPCPA has the best predictive abilities, combined with a good selective performance.

[en] This paper proposes a novel scheme for analog circuit fault diagnosis utilizing features extracted from the time-frequency representations of signals and an improved vector-valued regularized kernel function approximation (VVRKFA). First, the cross-wavelet transform is employed to yield the energy-phase distribution of the fault signals over the time and frequency domain. Since the distribution is high-dimensional, a supervised dimensionality reduction technique—the bilateral 2D linear discriminant analysis—is applied to build a concise feature set from the distributions. Finally, VVRKFA is utilized to locate the fault. In order to improve the classification performance, the quantum-behaved particle swarm optimization technique is employed to gradually tune the learning parameter of the VVRKFA classifier. The experimental results for the analog circuit faults classification have demonstrated that the proposed diagnosis scheme has an advantage over other approaches. (paper)