Medical Expert Systems is an active research area where data analysts and medical experts are continuously collaborating to make these systems more accurate and therefore, more useful in real life. Recent surveys by World Health Organization indicated a great increase in number of diabetic patients and the deaths that are attributed to diabetes each year. Therefore, early diagnosis of diabetes is a major concern among researchers and practitioners. The paper presents an application of automatic multilayer perceptron (AutoMLP) which is combined with an outlier detection method Enhanced Class Outlier Detection using distance based algorithm to create a novel prediction framework. AutoMLP is auto-tunable and performs parameter optimization automatically on the run during training process, which otherwise requires human intervention. Our framework performs outlier detection during pre-processing of data. A series of experiments are performed publicly available dataset: UCI (Prima Indian) and system achieved an accuracy of 88.7% which bests the highest reported results.
With advanced data analysis techniques, efforts for more accurate decision support systems for disease prediction are on the rise. According to the World Health Organization, diabetes-related illnesses and mortalities are on the rise. Hence, early diagnosis is particularly important. In this paper, we present a framework, Auto-MeDiSine, that comprises an automated version of enhanced class outlier detection using a distance-based algorithm (AutoECODB), combined with an ensemble of automatic multilayer perceptron (AutoMLP). AutoECODB is built upon ECODB by automating the tuning of parameters to optimize outlier detection process. AutoECODB cleanses the dataset by removing outliers. Preprocessed dataset is then used to train a prediction model using an ensemble of AutoMLPs. A set of experiments is performed on publicly available Pima Indian Diabetes Dataset as follows: (1) Auto-MeDiSine is compared with other state-of-the-art methods reported in the literature where Auto-MeDiSine realized an accuracy of 88.7%; (2) AutoMLP is compared with other learners including individual (focusing on neural networkbased learners) and ensemble learners; and (3) AutoECODB is compared with other preprocessing methods. Furthermore, in order to validate the generality of the framework, Auto-MeDiSine is tested on another publicly available BioStat Diabetes Dataset where it outperforms the existing reported results, reaching an accuracy of 97.1%.
Adversarial examples are used to evaluate the robustness of convolutional neural networks (CNNs) to input perturbations. Researchers have proposed different types of adversarial examples that attack CNNs to fool them. These attacks pose a serious threat to applications that use deep neural networks. Existing methods for adversarial image generation struggle in maintaining a balance between attack success rate and imperceptibility (measured in terms of l 2 -norm) of the generated adversarial examples. Recent sparse methods for this problem focus on limiting the number of pixels in an image but do not cater to the overall imperceptibility of the adversarial images. To address these problems, we introduce adversarial attacks based on K-singular value decomposition sparse dictionary learning. The dictionary is learned using feature maps of the targeted images from the first layer of CNN. The proposed method is evaluated in terms of attack success rate and l 2 -norm. The extensive experimentation shows our attack achieves a high success rate while maintaining a low imperceptibility score compared to state-ofthe-art methods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.