BACKGROUND: The aim was to develop a personalized survival prediction deep learning model for cervical adenocarcinoma patients and process personalized survival prediction. METHODS: A total of 2501 cervical adenocarcinoma patients from the surveillance, epidemiology and end results database and 220 patients from Qilu hospital were enrolled in this study. We created our deep learning (DL) model to manipulate the data and evaluated its performance against four other competitive models. We tried to demonstrate a new grouping system oriented by survival outcomes and process personalized survival prediction by using our DL model. RESULTS: The DL model reached 0.878 c-index and 0.09 Brier score in the test set, which was better than the other four models. In the external test set, our model achieved a 0.80 c-index and 0.13 Brier score. Thus, we developed prognosis-oriented risk grouping for patients according to risk scores computed by our DL model. Notable differences among groupings were observed. In addition, a personalized survival prediction system based on our risk-scoring grouping was developed. CONCLUSIONS: We developed a deep neural network model for cervical adenocarcinoma patients. The performance of this model proved to be superior to other models. The results of external validation supported the possibility that the model can be used in clinical work. Finally, our survival grouping and personalized prediction system provided more accurate prognostic information for patients than traditional FIGO stages.

This paper addresses the problem of inferring link loss rates based on network performance tomography in noisy network systems. Since network tomography emerged, all existing tomography-based methods are limited to the fulfillment of a basic condition: both network topologies and end-to-end routes must be absolutely accurate, which in most cases is impractical, especially for large-scale heterogeneous networks. To overcome the impracticability of tomography-based methods, we propose a robust tomography-based loss inference method capable of accurately inferring all link loss rates even when the given knowledge about the system is unreliable. Rather than computing the link loss rates directly from end-to-end loss rates, it calculates an upper bound for each link loss rate. It then infers all the link loss rates that most closely conform to the measurement results within their upper bounds. For a scenario where noisy measurements are caused by link (or router port) failures, we propose a hidden fault diagnosis approach that utilizes the inferred link loss rates to pinpoint the insidious faults that are hardly detected. It first determines the possible fake routes based on inferred link loss rates. Subsequently, it finds the maximum probable faults that can best explain the fake routes. Through intensive experiments, the results strongly confirm the promising performance of our proposed approaches.

Tag services have recently become one of the most popular Internet services on the World Wide Web. Due to the fact that a Web page can be associate with multiple tags, previous research on tag recommendation mainly focuses on improving its accuracy or efficiency through multi-label learning algorithms. However, as a Web page can also be split into multiple sections and be represented as a bag of instances, multi-instance multi-label learning framework should fit this problem better. In this paper, we improve the performance of tag suggestion by using multi-instance multi-label learning. Each Web page is divided into a bag of instances. The experiments on real-word data from delicious suggest that our framework has better performance than traditional multi-label learning methods on the task of tag recommendation.

Due to the dynamic and anonymous nature of open environments, it is critically important for agents to identify trustful cooperators which work consistently as they claim. In the e-services and e-commerce communities, trust and reputation systems are applied broadly as one kind of decision support systems, and aim to cope with the consistency problems caused by uncertain trust relationships. However, challenges still exist: on the one hand, we require more flexible trust computation models to satisfy various personal requirements since agents in these communities are heterogeneous; on the other hand, trust and reputation systems calculate the trustworthiness of agents based on the agents' past behavior. The open environments are dynamic, agents are anonymous and the records about agents' past behavior are distributed in the environments, so agents have to search the required records through the environments due to their lack of valid information. Thus, efficient, scalable and effective information collection strategies are required to address these issues. In this paper we present a distributed trust and reputation system to cope with the challenges. We propose a novel and flexible trust computation model based on artificial neural networks. With the advantages of ANN, our trust model tunes the parameters automatically to adapt to various personal requirements. We propose a broker-assisting information collection strategy based on clustering method. With the support of brokers, subcommunities are managed by reputation mechanism in an efficient and scalable way and help their members collect information with high quality. We show the performance of our trust and reputation system by simulation.

Aiming at Automated Guided Vehicle (AGV) dynamic model characteristics, a Variable Structure Control based on genetic algorithm (GA) and least square-support vector machine (LS-SVM) was designed. Parameters, predetermined by conventional reaching law, were regulated by LS-SVM online. It was shown that system shattering is eliminated. Simulation results indicated that this method possesses the advantages of higher precision, greater adaptability and robustness, as compared to the conventional Variable Structure Control methods.