Analyses of coronary graft patency after aprotinin use: results from the International Multicenter Aprotinin Graft Patency Experience (IMAGE) trial

OBJECTIVE: We examined the effects of aprotinin on graft patency, prevalence of myocardial infarction, and blood loss in patients undergoing primary coronary surgery with cardiopulmonary bypass. METHODS: Patients from 13 international sites were randomized to receive intraoperative aprotinin (n = 436) or placebo (n = 434). Graft angiography was obtained a mean of 10.8 days after the operation. Electrocardiograms, cardiac enzymes, and blood loss and replacement were evaluated. RESULTS: In 796 assessable patients, aprotinin reduced thoracic drainage volume by 43% (P < .0001) and requirement for red blood cell administration by 49% (P < .0001). Among 703 patients with assessable saphenous vein grafts, occlusions occurred in 15.4% of aprotinin-treated patients and 10.9% of patients receiving placebo (P = .03). After we had adjusted for risk factors associated with vein graft occlusion, the aprotinin versus placebo risk ratio decreased from 1.7 to 1.05 (90% confidence interval, 0.6 to 1.8). These factors included female gender, lack of prior aspirin therapy, small and poor distal vessel quality, and possibly use of aprotinin-treated blood as excised vein perfusate. At United States sites, patients had characteristics more favorable for graft patency, and occlusions occurred in 9.4% of the aprotinin group and 9.5% of the placebo group (P = .72). At Danish and Israeli sites, where patients had more adverse characteristics, occlusions occurred in 23.0% of aprotinin- and 12.4% of placebo-treated patients (P = .01). Aprotinin did not affect the occurrence of myocardial infarction (aprotinin: 2.9%; placebo: 3.8%) or mortality (aprotinin: 1.4%; placebo: 1.6%). CONCLUSIONS: In this study, the probability of early vein graft occlusion was increased by aprotinin, but this outcome was promoted by multiple risk factors for graft occlusion.     

Online learning of task-driven object-based visual attention control

We propose a biologically-motivated computational model for learning task-driven and object-based visual attention control in interactive environments. In this model, top-down attention is learned interactively and is used to search for a desired object in the scene through biasing the bottom-up attention in order to form a need-based and object-driven state representation of the environment. Our model consists of three layers. First, in the early visual processing layer, most salient location of a scene is derived using the biased saliency-based bottom-up model of visual attention. Then a cognitive component in the higher visual processing layer performs an application specific operation like object recognition at the focus of attention. From this information, a state is derived in the decision making and learning layer. Top-down attention is learned by the U-TREE algorithm which successively grows an object-based binary tree. Internal nodes in this tree check the existence of a specific object in the scene by biasing the early vision and the object recognition parts. Its leaves point to states in the action value table. Motor actions are associated with the leaves. After performing a motor action, the agent receives a reinforcement signal from the critic. This signal is alternately used for modifying the tree or updating the action selection policy. The proposed model is evaluated on visual navigation tasks, where obtained results lend support to the applicability and usefulness of the developed method for robotics.     

Attention control with reinforcement learning for face recognition under partial occlusion

In this paper a new method for handling occlusion in face recognition is presented. In this method the faces are partitioned into blocks and a sequential recognition structure is developed. Then, a spatial attention control strategy over the blocks is learned using reinforcement learning. The outcome of this learning is a sorted list of blocks according to their average importance in the face recognition task. In the recall mode, the sorted blocks are employed sequentially until a confident decision is made. Obtained results of various experiments on the AR face database demonstrate the superior performance of proposed method as compared with that of the holistic approach in the recognition of occluded faces.     

Optimal Local Basis: A Reinforcement Learning Approach for Face Recognition

This paper presents a novel learning approach for Face Recognition by introducing Optimal Local Basis. Optimal local bases are a set of basis derived by reinforcement learning to represent the face space locally. The reinforcement signal is designed to be correlated to the recognition accuracy. The optimal local bases are derived then by finding the most discriminant features for different parts of the face space, which represents either different individuals or different expressions, orientations, poses, illuminations, and other variants of the same individual. Therefore, unlike most of the existing approaches that solve the recognition problem by using a single basis for all individuals, our proposed method benefits from local information by incorporating different bases for its decision. We also introduce a novel classification scheme that uses reinforcement signal to build a similarity measure in a non-metric space. Experiments on AR, PIE, ORL and YALE databases indicate that the proposed method facilitates robust face recognition under pose, illumination and expression variations. The performance of our method is compared with that of Eigenface, Fisherface, Subclass Discriminant Analysis, and Random Subspace LDA methods as well.     

大型锻件热处理过程的数值模拟研究

在对大型锻件传热分析的基础上,建立了大型锻件的物理模型,对其温度场的变化规律进行了模拟计算,同时将模拟结果和实验结果进行了对比分析,并通过实例说明了热处理过程数值模拟方法的准确性和优越性,为进一步分析此类工件的组织转变和应力应变等奠定了基础.     

SIMULTANEOUS STATE ESTIMATION AND LEARNING IN REPEATED COURNOT GAMES

The aim of this article is to propose that an intelligent agent can be able to decide properly in an incomplete information repeated Cournot game. The market model and the competitors’ decision models are not known to the players. The proposed agent employs a combination of the k-nearest neighbor (KNN) method and the Bayes classifier to predict the next action of its rivals, using the market decision history. The agent takes the predicted actions as an estimate of its next state and learns the expected payoff of its state-action pairs interactively using the reinforcement learning (RL) algorithm. The results of the proposed agent's competition with two benchmark competitors in different simulated Cournot games are presented. The simulation results show that the proposed agent can significantly earn more payoffs in comparison with the two benchmark agents.     

Comparative study of 1-D and 2-D metamaterial lens for microwave nondestructive evaluation of dielectric materials

A novel microwave nondestructive evaluation (NDE) sensor was developed in an attempt to increase the sensitivity of the microwave NDE method for detection of material defects small relative to a wavelength. The sensor was designed on the basis of a negative index material (NIM) lens. Transmission at the resonant frequency through the 1-D lens was determined to be about 10 times higher than that with the 2-D lens. However, the focusing ability of the 1-D lens was found to be slightly lower to the 2-D lens (focus spot size for the 1-D lens was determined to be 0.7λ vs. 0.48λ for the 2-D lens). A fiberglass material sample with a 3 mm (0.037λ) diameter through hole (perpendicular to the propagation direction of the wave) was tested with both lenses. The hole was successfully detected with an 8.2 cm wavelength electromagnetic wave with both lenses, but the image obtained with a 2-D lens was much sharper. Therefore, the choice of the lens to be used in a sensor is prescribed by the specific requirements of the testing system. For example, a 1-D lens should be considered when the simplicity of the testing system is deemed more important than the quality of the image obtained from a defect. A 1-D lens also allows for a longer sample standoff distance and higher transmission.