With serious advertising budget constraints, advertisers have to adjust their daily budget according to the performance of advertisements in real time. Thus we can leave precious budgets to better opportunities in the...
详细信息
With serious advertising budget constraints, advertisers have to adjust their daily budget according to the performance of advertisements in real time. Thus we can leave precious budgets to better opportunities in the future, and avoid the surge of ineffective clicks for unnecessary costs. However, advertisers usually have no sufficient knowledge and time for real-time advertising operations in search auctions. We formulate the budget adjustment problem as a state-action decision process in the reinforcement learning (RL) framework. Considering dynamics of marketing environments and some distinctive features of search auctions, we extend continuous reinforcement learning to fit the budget decision scenarios. The market utility is defined as discounted total clicks to get during the remaining period of an advertising schedule. We conduct experiments to validate and evaluate our strategy of budget adjustment with real world data from search advertising campaigns. Experimental results showed that our strategy outperforms the two other baseline strategies.
Peculiarity oriented mining (POM), aimed at discovering peculiarity rules hidden in a dataset, is a data mining method. Peculiarity factor (PF) is one of the most important concepts in POM. In this paper, it is proved...
详细信息
Peculiarity oriented mining (POM), aimed at discovering peculiarity rules hidden in a dataset, is a data mining method. Peculiarity factor (PF) is one of the most important concepts in POM. In this paper, it is proved that PF can accurately characterize the peculiarity of data sampled from a normal distribution. However, for a general one-dimensional distribution, it does not have the property. A local version of PF, called LPF, is proposed to solve the difficulty. LPF can effectively describe the peculiarity of data sampled from a continuous one-dimensional distribution. Based on LPF, a framework of local peculiarity oriented mining is presented, which consists of two steps, namely, peculiar data identification and peculiar data analysis. Two algorithms for peculiar data identification and a case study of peculiar data analysis are given to make the framework practical. Experiments on several benchmark datasets show their good performance.
This paper presents a novel closed-loop method for a multilink robotic fish to mimic the C-start maneuver, in which the turning speed and precision are emphasized. The turning speed is maximized by carefully designed ...
详细信息
Modern power grid is a typical multi-level complex giant system. The conventional analytical methods based on reductionism can't provide sufficient guidance for its operation and management. complex system theory,...
详细信息
Due to FPGA's flexibility and parallelism, it is popular for accelerating image processing. In this paper, a double-parallel architecture based on FPGA has been exploited to speed up median filter and edge detecti...
详细信息
Image-To-Class distance is first proposed in Naive- Bayes Nearest-Neighbor. NBNN is a feature-based image classifier, and can achieve impressive classification accuracy. However, the performance of NBNN relies heavily...
详细信息
ISBN:
(纸本)9781457701221
Image-To-Class distance is first proposed in Naive- Bayes Nearest-Neighbor. NBNN is a feature-based image classifier, and can achieve impressive classification accuracy. However, the performance of NBNN relies heavily on the large number of training samples. If using small number of training samples, the performance will degrade. The goal of this paper is to address this issue. The main contribution of this paper is that we propose a robust Image-to-Class distance by local learning. We define the patch-to-class distance as the distance between the input patch to its nearest neighbor in one class, which is reconstructed in the local manifold space;and then our image-toclass distance is the sum of patch-to-class distance. Furthermore, we take advantage of large-margin metric learning framework to obtain a proper Mahalanobis metric for each class. We evaluate the proposed method on four benchmark datasets: Caltech, Corel, Scene13, and Graz. The results show that our defined Image-To-Class Distance is more robust than NBNN and Optimal-NBNN, and by combining with the learned metric for each class, our method can achieve significant improvement over previous reported results on these datasets.
Histogram features, such as SIFT, HOG, LBP et al, are widely used in modern computer vision algorithms. According to [18], chi-square distance is an effective measure for comparing histogram features. In this paper, w...
详细信息
ISBN:
(纸本)9781457701221
Histogram features, such as SIFT, HOG, LBP et al, are widely used in modern computer vision algorithms. According to [18], chi-square distance is an effective measure for comparing histogram features. In this paper, we propose a new method, named the Quadric-chi similarity metric learning (QCSML) for histogram features. The main contribution of this paper is that we propose a new metric learning method based on chi-square distance, in contrast with traditional Mahalanobis distance metric learning methods. The use of quadric-chi similarity in our method leads to an effective learning algorithm. Our method is tested on SIFT features for face identification, and compared with the state-of-art metric learning method (LDML) on the benchmark dataset, the labeled Faces in the Wild (LFW). Experimental results show that our method can achieve clear performance gains over LDML.
Background modeling is a fundamental yet challenging issue in video surveillance. Traditional methods usually adopt single feature type to solve the problem, while the performance is usually unsatisfactory when handli...
详细信息
ISBN:
(纸本)9781457701221
Background modeling is a fundamental yet challenging issue in video surveillance. Traditional methods usually adopt single feature type to solve the problem, while the performance is usually unsatisfactory when handling complex scenes. In this paper, we propose a multi-scale framework, which combines both texture and intensity feature, to achieve a robust and accurate solution. Our contributions are three folds: first, we provide a multi-scale analysis for the issue;second, for texture feature we propose a novel texture operator named Scale-invariant Centersymmetric Local Ternary Pattern, and a corresponding Pattern Adaptive Kernel Density Estimation technique for its probability estimation;third, we design a Simplified Gaussian Mixture Models for intensity feature. Our method is tested on several complex real world videos with illumination variation, soft shadows and dynamic backgrounds. The experimental results clearly demonstrate that our method is superior to the previous methods.
Recently, Independent Component Analysis based foreground detection has been proposed for indoor surveillance applications where the foreground tends to move slowly or remain still. Yet such a method often causes disc...
详细信息
ISBN:
(纸本)9781457701221
Recently, Independent Component Analysis based foreground detection has been proposed for indoor surveillance applications where the foreground tends to move slowly or remain still. Yet such a method often causes discrete segmented foreground objects. In this paper, we propose a novel foreground detection method named Contextual Constrained Independent Component Analysis (CCICA) to tackle this problem. In our method, the contextual constraints are explicitly added to the optimization objective function, which indicate the similarity relationship among neighboring pixels. In this way, the obtained de-mixing matrix can produce the complete foreground compared with the previous ICA model. In addition, our method performs robust to the indoor illumination changes and features a high processing speed. Two sets of image sequences involving room lights switching on/of and door opening/closing are tested. The experimental results clearly demonstrate an improvement over the basic ICA model and the image difference method.
Online advertisers bidding in keyword auctions through Web search engines are experiencing fierce competition. Based on a model of advertisers' rational competitive preference, we propose a novel solution concept ...
详细信息
Online advertisers bidding in keyword auctions through Web search engines are experiencing fierce competition. Based on a model of advertisers' rational competitive preference, we propose a novel solution concept called the Upper Bound Nash Equilibrium(UBNE) targeting at modeling the competitive bidding dynamics. UBNE yields the best outcome for search engines, and provides a rational explanation of the bid inflation dynamics on keyword markets.
暂无评论