Atsuyoshi Nakamura

Partitioning of Web graphs by community topology

We introduce a stricter Web community definition to overcome boundary ambiguity of a Web community defined by Flake, Lawrence and Giles [2], and consider the problem of finding communities that satisfy our definition. We discuss how to find such communities and hardness of this problem.We also propose Web page partitioning by equivalence relation defined using the class of communities of our definition. Though the problem of efficiently finding all communities of our definition is NP-complete, we propose an efficient method of finding a subclass of communities among the sets partitioned by each of n-1 cuts represented by a Gomory-Hu tree [10], and partitioning a Web graph by equivalence relation defined using the subclass.According to our preliminary experiments, partitioning by our method divided the pages retrieved by keyword search into several different categories to some extent.

Improvements to the Linear Programming Based Scheduling of Web Advertisements

We propose and evaluate a number of improvements to the linear programming formulation of web advertisement scheduling, which we have proposed elsewhere together with our colleagues [Langheinrich et al., 9]. In particular, we address a couple of important technical challenges having to do with the estimation of click-through rates and optimization of display probabilities (the exploration–exploitation trade-off and the issue of data sparseness and scalability), as well as practical aspects that are essential for successful deployment of this approach (the issues of multi-impressions and inventory management). We propose solutions to each of these issues, and assess their effectiveness by running large-scale simulation experiments.

Algorithms for adversarial bandit problems with multiple plays

Adversarial bandit problems studied by Auer et al. [4] are multi-armed bandit problems in which no stochastic assumption is made on the nature of the process generating the rewards for actions. In this paper, we extend their theories to the case where k(≥1) distinct actions are selected at each time step. As algorithms to solve our problem, we analyze an extension of Exp3 [4] and an application of a bandit online linear optimization algorithm [1] in addition to two existing algorithms (Exp3, ComBand [5]) in terms of time and space efficiency and the regret for the best fixed action set. The extension of Exp3, called Exp3.M, performs best with respect to all the measures: it runs in O(K(log k+1)) time and O(K) space, and suffers at most O(√kTK log(K/k)) regret, where K is the number of possible actions and T is the number of iterations. The upper bound of the regret we proved for Exp3.M is an extension of that proved by Auer et al. for Exp3.

Improvements in practical aspects of optimally scheduling web advertising

We addressed two issues concerning the practical aspects of optimally scheduling web advertising proposed by Langheinrich et al. [5], which scheduling maximizes the total number of click-throughs for all banner advertisements. One is the problem of multi-impressions in which two or more banner ads are impressed at the same time. The other is inventory management, which is important in order to prevent over-selling and maximize revenue. We propose efficient methods which deal with these two issues.

On the Minimum ℓ 1 -Norm Signal Recovery in Underdetermined Source Separation

This paper studied the minimum l1-norm signal recovery in underdetermined source separation, which is a problem of separating n sources blindly from m linear mixtures for n>m. Based on our previous result of submatrix representation and decision regions, we describe the property of the minimum l1-norm sequence from the viewpoint of source separation, and discuss how to construct it geometrically from the observed sequence and the mixing matrix, and the unstability for a perturbation of mixing matrix.

Bagging, Random Subspace Method and Biding

In recent years, many approaches for achieving high performance by combining some classifiers have been proposed. We exploit many random replicates of samples in the bagging, and randomly chosen feature subsets in the random subspace method. In this paper, we introduce a method for selecting both samples and features at the same time and demonstrate the effectiveness of the method. This method includes a parametric bagging and a parametric random subspace method as special cases. In some experiments, this method and the parametric random subspace method showed the best performance.

Classification by reflective convex hulls

A set of convex bodies including samples of a single class only is used for classification. The convex body is defined by some facets (hyper-planes) that separate the class from the other classes. This paper describes an algorithm to find a set of such convex bodies efficiently and examine the performance of a classifier using them. The relationship to the support vector machines is also discussed.

CCIC: Consistent Common Itemsets Classifier

We propose a novel approach which extracts consistent (100% confident) rules and builds a classifier with them. Recently, associative classifiers which utilize association rules have been widely studied. Indeed, the associative classifiers often outperform the traditional classifiers. In this case, it is important to collect high quality (association) rules. Many algorithms find only high support rules, because decreasing the minimum support to be satisfied is computationally demanding. However, it may be effective to collect low support but high confidence rules. Therefore, we propose an algorithm that produces a wide variety of 100% confident rules including low support rules. To achieve this goal, we adopt a specific-to-general rule searching strategy, in contrast to the previous many approaches. Our experimental results show that the proposed method achieves higher accuracies in several datasets taken from UCI machine learning repository.

Collaborative filtering using restoration operators

We propose a new collaborative filtering method that uses restoration operators. The problem of restoration by operators was originally studied in the field of digital image restoration [9]. We also consider the problem of selecting items that users should be asked to rate in order to achieve a small expected squared error, and we propose a greedy method as a solution of this problem. According to our experimental results, prediction performance of restoration operators is good when the number of observed ratings is small, and our greedy method outperforms random query item selection.

What Sperner Family Concept Class is Easy to Be Enumerated

We study the problem of enumerating concepts in a Sperner family concept class using subconcept queries, which is a general problem including maximal frequent itemset mining as its instance. Though even the theoretically best known algorithm needs quasi-polynomial time to solve this problem in the worst case, there exist practically fast algorithms for this problem. This is because many instances of this problem in real world have low complexity in some measures. In this paper, we characterize the complexity of Sperner family concept class by the VC dimension of its intersection closure and its characteristic dimension, and analyze the worst case time complexity on the enumeration problem of its concepts in terms of the VC dimension. We also showed that the VC dimension of real data used in data mining is actually small by calculating the VC dimension of some real datasets using a new algorithm closely related to the introduced two measures, which does not only solve the problem but also let us know the VC dimension of the intersection closure of the target concept class.