ÐThe primary goal of pattern recognition is supervised or unsupervised classification. Among the various frameworks in which pattern recognition has been traditionally formulated, the statistical approach has been most intensively studied and used in practice. More recently, neural network techniques and methods imported from statistical learning theory have been receiving increasing attention. The design of a recognition system requires careful attention to the following issues: definition of pattern classes, sensing environment, pattern representation, feature extraction and selection, cluster analysis, classifier design and learning, selection of training and test samples, and performance evaluation. In spite of almost 50 years of research and development in this field, the general problem of recognizing complex patterns with arbitrary orientation, location, and scale remains unsolved. New and emerging applications, such as data mining, web searching, retrieval of multimedia data, face recognition, and cursive handwriting recognition, require robust and efficient pattern recognition techniques. The objective of this review paper is to summarize and compare some of the well-known methods used in various stages of a pattern recognition system and identify research topics and applications which are at the forefront of this exciting and challenging field.

Numerous e orts have been made in developing \intelligent" programs based on the Von Neumann's centralized architecture. However, these e orts have not been very successful in building general-purpose intelligent systems. Inspired by biological neural networks, researchers in a number of scienti c disciplines are designing arti cial neural networks (ANNs) to solve a variety of problems in decision making, optimization, prediction, and control. Arti cial neural networks can be viewed as parallel and distributed processing systems which consist of a huge number of simple and massively connected processors. There has been a resurgence of interest in the eld of ANNs for several years. This article intends to serve as a tutorial for those readers with little or no knowledge about ANNs to enable them to understand the remaining articles of this special issue. We discuss the motivations behind developing ANNs, basic network models, and two main issues in designing ANNs: network architecture and learning process. We also present one of the most successful application of ANNs, namely automatic character recognition.

No abstract

Display advertising has been a significant source of revenue for publishers and ad networks in online advertising ecosystem. One important business model in online display advertising is Ad Exchange marketplace, also called nonguaranteed delivery (NGD), in which advertisers buy targeted page views and audiences on a spot market through real-time auction. In this paper, we describe a bid landscape forecasting system in NGD marketplace for any advertiser campaign specified by a variety of targeting attributes. In the system, the impressions that satisfy the campaign targeting attributes are partitioned into multiple mutually exclusive samples. Each sample is one unique combination of quantified attribute values. We develop a divide-andconquer approach that breaks down the campaign-level forecasting problem. First, utilizing a novel star-tree data structure, we forecast the bid for each sample using non-linear regression by gradient boosting decision trees. Then we employ a mixture-of-log-normal model to generate campaignlevel bid distribution based on the sample-level forecasted distributions. The experiment results of a system developed with our approach show that it can accurately forecast the bid distributions for various campaigns running on the world's largest NGD advertising exchange system, outperforming two baseline methods in term of forecasting errors.

Abstruct-A nonlinear projection method is presented to visualize higb-dimensional data as a two-dimensional image. The proposed method b based on the topotogV p " mpp-ping algorithm d Kohonen [13H16]. The tapology preserving mapping algorithm is used to trpin a two-dimensional network structure. Then the interpoint dbtances in tbe feature space between the units in the network are graphidly cusplayea to show the underlying StruCtuFe of the data. Fartheimore, we will present and discuss a new method to qnadfy how well a topologv preserving mapping algorithm maps the bigbdbensiod input data onto the network stmeture, This will be used to compare our projection method with a well-k~~own method of Sa"on [ S I. Experiments indicate that the performance of the Koho-nen projection method is con~pambk or better than Sammon's method for the purpose of clparsurine dasEcnd data. Another advantage of the metbod is that its tbe-complesity only depends on the resolution of the outpot irmrse, and not on the size of the dataset. A disadvantage, however, is the large amount of CPU time required.