Search

Highlights • Enhanced nonlinearities may improve expert systems performance. • Proposal of the activation function DReLU. • DReLU presents the best training speed in all cases. • DReLU enhances the ReLU performance in all scenarios. • DReLU provides the best test accuracy in almost all experiments. Abstract A substantial number of expert and intelligent systems rely on deep learning methods to solve problems in areas such as economics, physics, and medicine. Improving the accuracy of the activation functions used by such methods can directly and positively impact the overall performance and quality of the mentioned systems at no cost whatsoever. In this sense, enhancing the design of such theoretical fundamental blocks is of great significance as it immediately impacts a broad range of current and future real-world deep learning based applications. Therefore, in this paper, we turn our attention to the interworking between the activation functions and the batch normalization, which is practically a mandatory technique to train deep networks currently. We propose the activation function Displaced Rectifier Linear Unit (DReLU) by conjecturing that extending the identity function of ReLU to the third quadrant enhances compatibility with batch normalization. Moreover, we used statistical tests to compare the impact of using distinct activation functions (ReLU, LReLU, PReLU, ELU, and DReLU) on the learning speed and test accuracy performance of standardized VGG and Residual Networks state-of-the-art models. These Convolutional Neural Networks were trained on CIFAR-100 and CIFAR-10, the most commonly used deep learning computer vision datasets. The results showed DReLU speeded up learning in all models and datasets. Besides, statistical significant performance assessments (p < 0.05) showed DReLU enhanced the test accuracy presented by ReLU in all scenarios. Furthermore, DReLU showed better test accuracy than any other tested activation function in all experiments with one exception, in which case it presented the second best performance. Therefore, this work demonstrates that it is possible to increase performance replacing ReLU by an enhanced activation function. [ABSTRACT FROM AUTHOR]

Expert and intelligent systems try to simulate intelligent human experts in solving complex real-world problems. The domain of problems varies from engineering and industry to medicine and education. In most situations, the system is required to take decisions based on multiple inputs, but the search space is usually very huge so that it will be very hard to use the traditional algorithms to take a decision; at this point, the metaheuristic algorithms can be used as an alternative tool to find near-optimal solutions. Thus, inventing new metaheuristic techniques and enhancing the current algorithms is necessary. In this paper, we introduced an enhanced variant of the Flower Pollination Algorithm (FPA). We hybridized the standard FPA with the Clonal Selection Algorithm (CSA) and tested the new algorithm by applying it to 23 optimization benchmark problems. The proposed algorithm is compared with five famous optimization algorithms, namely, Simulated Annealing, Genetic Algorithm, Flower Pollination Algorithm, Bat Algorithm, and Firefly Algorithm. The results show that the proposed algorithm is able to find more accurate solutions than the standard FPA and the other four techniques. The superiority of the proposed algorithm nominates it for being a part of intelligent and expert systems. [ABSTRACT FROM AUTHOR]

Mixture of experts (ME) as an ensemble method consists of several experts and a gating network to decompose the input space into some subspaces regarding to the experts specialties. To increase the diversity between experts in ME, this paper incorporates a correlation penalty function into the error function of ME. The significant of this modification is providing an occasion to encourage experts to specialize on different parts of the input space and to create decorrelated experts. The experimental results of this approach reveals that the impacts of this penalty function is extremely improved the diversity of experts and the tradeoff between the accuracy and the diversity in ME. Moreover in the implementation of this method, the experts are trained simultaneously and they can communicate by the aid of the correlation penalty function. The performance of the proposed method on ten classification benchmark datasets shows that the average of accuracy of this method improves 1.94%, 3.7%, and 3.74% compared with the mixture of negatively correlated experts, ME and the negative correlation learning, respectively. Thus the proposed method can be considered as a better classifier for healthy and medical problems and also when the great non-stationary data should be classified. [ABSTRACT FROM AUTHOR]

• Expert system to analysis and diagnosis of human stress. • Temperature behavior in facial and fingertips was investigated to stress. • Infrared thermography and image processing to automatic recognition of the ROI. • Statistical analysis and decision tree to diagnose human stress. • Improvement in the accuracy of the classification and diagnosis in human stress. Human stress is a physical or emotional feeling. It can come from any situation or thought that makes one feel frustrated or nervous. Different biological manifestations take place in the presence of stress, such as tension, headache, and insomnia. Recent studies have reported that human stress can be related to facial expressions and fingertips due to temperature changes in the skin. Infrared thermography is a non-invasive technology that allows for the monitoring and analysis of human skin temperature. However, many works reported in the literature perform a manual analysis, depending on expert personnel, resulting in considerable human and economic efforts. In addition, the analyses reported to date with thermography are only based on the study of body parts. To reduce the limitations of these methodologies, expert systems have been proposed that simulate the thought process of a human expert to solve decision problems, which may be helpful for people focused on the health and psychology area who do not have expertise in the study of stress. Therefore, this paper proposes a novel expert system based on infrared thermography and thermal analysis of facial skin and fingertips, a rule-based method, and heuristic knowledge to classify and diagnose human stress in undergraduate university students. The system had the support of experts in the study of human stress and was validated in a stress study. The novel expert system was implemented in a local database that consisted of a group of 100 participants, undergraduate university students, of which 70 were stimulated by the Trier Social Stress Test (TSST) protocol and 30 were not induced to human stress, obtaining an accuracy in the expert system stress classification of 91.0%. [ABSTRACT FROM AUTHOR]