International Journal of Machine Learning and Cybernetics

This study is concerned with the robust stability problem of uncertain genetic regulatory networks (GRNs) with random discrete time delays and distributed time delays which exist both in translation process and feedback regulation process. By utilizing a novel Lyapunov–Krasovskii functional which contains some triple integral terms and takes into account the ranges of delays, we derive sufficient delay-dependent conditions to ensure the asymptotically stability of GRNs with mixed time delays. Moreover, based on the idea of “delay decomposing”, “reciprocally convex combination approach”, less conservative conditions are obtained by using the lower bound lemma together with Jensen inequality. In addition, two corollaries are also been presented. Finally, numerical examples are presented to show the effectiveness of our proposed methods.

Aiming at existing symmetric regularized picture fuzzy clustering with weak robustness, and it is difficult to meet the need for image segmentation in the presence of high noise. Hence, a robust dynamic semi-supervised symmetric regularized picture fuzzy clustering with KL-divergence and spatial information constraints is presented in this paper. Firstly, a weighted squared Euclidean distance from current pixel value, its neighborhood mean and median to clustering center is firstly proposed, and it is embedded into the objective function of symmetric regularized picture fuzzy clustering to obtain spatial picture fuzzy clustering. Secondly, the idea of maximum entropy fuzzy clustering is introduced into picture fuzzy clustering, and an entropy-based picture fuzzy clustering with clear physical meaning is constructed to avoid the problem of selecting weighted factors. Subsequently, the prior information of the current pixel is obtained by means of weighted local membership of neighborhood pixels, and it is embedded into the objective function of maximum entropy picture fuzzy clustering with multiple complementary spatial information constraints through KL-divergence, a robust dynamic semi-supervised picture fuzzy clustering optimization model and its iterative algorithm are given. In the end, this proposed algorithm is strictly proved to be convergent by Zangwill theorem. The experiments on various images and standard datasets illustrate how our proposed algorithm works. This proposed algorithm has excellent segmentation performance and anti-noise robustness, and outperforms eight state-of-the-art fuzzy or picture fuzzy clustering-related algorithms in the presence of high noise.

Images captured in low-light environments suffer from severe degradation, which can be unfavorable for human observation and subsequent computer vision tasks. Although many enhancement methods based on deep learning have been proposed, the obtained enhancement images still suffer from drawbacks such as color distortion, noise, and blur. To solve these problems, we propose a progressive enhancement network based on detail supplementation (PEN-DS), which is implemented by building two modules: an image preprocessing module (IPM) and a progressive image enhancement module (PIEM). The IPM can obtain low-light images and low-detail maps at different scales by building an image pyramid structure. PIEM can enhance images at different scales progressively based on detail supplementation and luminance enhancement. In addition, to better train the network, the proposed method employs a multi-supervised joint loss function for the enhanced images of different scales. Experimental results show that the proposed method outperforms state-of-the-art approaches in terms of visual observation and objective evaluation.

Knowledge reduction is complicated with the dynamic change of the object set in applications. In this paper, we propose incremental approaches to computing the type-1 and type-2 characteristic matrices of coverings with respect to variation of objects. Also we present two incremental algorithms of calculating the second and sixth lower and upper approximations of sets when adding and deleting more objects in dynamic covering approximation spaces. Subsequently, we employ experiments to validate that the incremental approaches are more effective and efficient to construct approximations of sets in dynamic covering information systems. Finally, we preform knowledge reduction of dynamic covering decision information systems by using the incremental approaches.

Fuzzy c-means (FCM) algorithms have been shown effective for image segmentation. A series of enhanced FCM algorithms incorporating spatial information have been developed for reducing the effect of noises. This paper presents a robust FCM algorithm with non-local spatial information for image segmentation, termed as NLFCM. It incorporates two factors: one is the local similarity measure depending on the differences between the central pixel and its neighboring pixels in the image; the other is the non-local similarity measure depended on all pixels whose neighborhood configurations are similar to their neighborhood pixels. Furthermore, an adaptive weight is introduced to control the trade-off between local similarity measure and non-local similarity measure. The experimental results on synthetic images and real images under different types of noises show that the new algorithm is effective, and they are relatively independent to the types of noises.

In recent years, high-level image representation is gaining popularity in image classification and retrieval tasks. This paper proposes an efficient scheme known as semantic context model to derive high-level image descriptors well suited for the retrieval operation. Semantic context model uses an undirected graphical model based formulation which jointly exploits low-level visual features and contextual information for classifying local image blocks into some predefined concept classes. Contextual information involves concept co-occurrences and their spatial correlation statistics. More expressive potential functions are introduced to capture the structural dependencies among various semantic concepts. The proposed framework proceeds in three steps. Initially, optimal values of model parameters that impose spatial consistency of concept labels among local image blocks are learned from the training data. Then, the semantics associated with the constituent blocks of an unseen image are inferred using an improved message-passing algorithm. Finally, a compact but discriminative image signature is derived by integrating the frequency of occurrence of various regional semantics. Experimental results on various benchmark datasets show that semantic context model can effectively resolve local ambiguities and consequently improve concept recognition performance in complex images. Moreover, the retrieval efficiency of the new semantics based image feature is found to be much better than state-of-the-art approaches.

Health data collection of children with autism spectrum disorder (ASD) is challenging, time-consuming, and expensive; thus, working with small datasets is inevitable in this area. The diagnosis rate in ASD is low, leading to several challenges, including imbalance classes, potential overfitting, and sampling bias, making it difficult to show its potential in real-life situations. This paper presents a data analytics pilot-case study using a small dataset leveraging domain-specific knowledge to uncover differences between the gestural patterns of children with ASD and neurotypicals. We collected data from 59 children using an elastic display we developed during a sensing campaign and 9 children using the elastic display as part of a therapeutic program. We extracted strength-related features and selected the most relevant ones based on how the motor atypicality of children with ASD influences their interactions: children with ASD make smaller and narrower gestures and experience variations in the use of strength. The proposed machine learning models can correctly classify children with ASD with 97.3% precision and recall even if the classes are unbalanced. Increasing the size of the dataset via synthetic data improved the model precision to 99%. We finish discussing the importance of leveraging domain-specific knowledge in the learning process to successfully cope with some of the challenges faced when working with small datasets in a concrete, real-life scenario.

The privileged information that is available only for the training examples and not available for test examples, is a new concept proposed by Vapnik and Vashist (Neural Netw 22(5–6):544–557, 2009). With the help of the privileged information, learning using privileged information (LUPI) (Neural Netw 22(5–6):544–557, 2009) can significantly accelerate the speed of learning. However, LUPI is a standard supervised learning method. In fact, in many real-world problems, there are also a lot of unlabeled data. This drives us to solve problems under a semi-supervised learning framework. In this paper, we propose a semi-supervised learning using privileged information (called Semi-LUPI), which can exploit both the distribution information in unlabeled data and privileged information to improve the efficiency of the learning. Furthermore, we also compare the relative importance of both types of information for the learning model. All experiments verify the effectiveness of the proposed method, and simultaneously show that Semi-LUPI can obtain superior performances over traditional supervised and semi-supervised methods.

This paper presents an imperceptible, robust, secure and efficient image watermarking scheme in lifting wavelet domain using combination of genetic algorithm (GA) and Lagrangian support vector regression (LSVR). First, four subbands low–low (LL), low–high (LH), high–low (HL) and high–high (HH) are obtained by decomposing the host image from spatial domain to frequency domain using one level lifting wavelet transform. Second, the approximate image (LL subband) is divided into non overlapping blocks and the selected blocks based on the fuzzy entropy are used to embed the binary watermark. Third, based on the correlation property of each transformed selected block, significant lifting wavelet coefficient act as target to LSVR and its neighboring coefficients (called feature vector) are set as input to LSVR to find optimal regression function. This optimal regression function is used to embed and extract the scrambled watermark. In the proposed scheme, GA is used to solve the problem of optimal watermark embedding strength, based on the noise sensitivity of each selected block, in order to increase the imperceptibility of the watermark. Due to the good learning capability and high generalization property of LSVR against noisy datasets, high degree of robustness is achieved and is well suited for copyright protection applications. Experimental results on standard and real world images show that proposed scheme not only efficient in terms of computational cost and memory requirement but also achieve good imperceptibility and robustness against geometric and non geometric attacks like JPEG compression, median filtering, average filtering, addition of noise, sharpening, scaling, cropping and rotation compared with the state-of-art techniques.