Izumi Suzuki

The language observatory project (LOP)

The first part of the paper provides a brief description of the Language Observatory Project (LOP) and highlights the major technical difficulties to be challenged. The latter part gives how we responded to these difficulties by adopting UbiCrawler as a data collecting engine for the project. An interactive collaboration between the two groups is producing quite satisfactory results.

Monotone Increasing Binary Similarity and Its Application to Automatic Document-Acquisition of a Category

A technique that acquires documents in the same category with a given short text is introduced. Regarding the given text as a training document, the system marks up the most similar document, or sufficiently similar documents, from among the document domain (or entire Web). The system then adds the marked documents to the training set to learn the set, and this process is repeated until no more documents are marked. Setting a monotone increasing property to the similarity as it learns enables the system to 1) detect the correct timing so that no more documents remain to be marked and to 2) decide the threshold value that the classifier uses. In addition, under the condition that the normalization process is limited to what term weights are divided by a p-norm of the weights, the linear classifier in which training documents are indexed in a binary manner is the only instance that satisfies the monotone increasing property. The feasibility of the proposed technique was confirmed through an examination of binary similarity and using English and German documents randomly selected from the Web.

Proposal of fuzzy coverage region classifier as an extension of the naive Bayes classifier and improvement of its zero-one loss

A new classifying rule using a fuzzy coverage region classifier is introduced in this paper. The rule enables us to formally alter conditional probability distributions to improve the zero-one loss (misclassification rate) of the naive Bayes classifier. Altering the probability distribution is a justifiable variation for defining a fuzzy set from the probability distribution. By using this approach, the range for altering the probability distribution is identified, for example: the value of a distribution function is allowed to replace its value to the power of 1/p, where p is approximately 1 to infinity. Optimizing the parameters of p in each feature and each class to minimize the zero-one loss improves the performance of the fuzzy coverage region classifier (or that of the naive Bayes classifier). Also, it is suggested that the performance of the non-fuzzy coverage region classifier is hardly influenced by the bias of training data, if the training data only covers the range of the class object.

Parallel Computation of Rough Set Approximations in Information Systems with Missing Decision Data

The paper discusses the use of parallel computation to obtain rough set approximations from large-scale information systems where missing data exist in both condition and decision attributes. To date, many studies have focused on missing condition data, but very few have accounted for missing decision data, especially in enlarging datasets. One of the approaches for dealing with missing data in condition attributes is named twofold rough approximations. The paper aims to extend the approach to deal with missing data in the decision attribute. In addition, computing twofold rough approximations is very intensive, thus the approach is not suitable when input datasets are large. We propose parallel algorithms to compute twofold rough approximations in large-scale datasets. Our method is based on MapReduce, a distributed programming model for processing large-scale data. We introduce the original sequential algorithm first and then the parallel version is introduced. Comparison between the two approaches through experiments shows that our proposed parallel algorithms are suitable for and perform efficiently on large-scale datasets that have missing data in condition and decision attributes.

Interactive decoration design support system by fitness evaluation based on design knowledge and subjective evaluation

This paper proposes interactive decoration design support system introducing interactive evolutionary computation. Evaluation problem existing human can be evaluated synthesizing multi objectives. The proposed system made it importance dividing the human evaluation into two part, which is including regular level of design quality and users subjective evaluation by affective or Kansei image, and assumed to acquire good results by executing in serial order in a certain evaluation phase. From the experimental results, effectiveness of proposed methodology involving evolution by using fitness evaluation of decoration designs for quality and evolution by user evaluation are confirmed.

Semi-supervised based rough set to handle missing decision data

We have developed a rough set model for analyzing an information system in which some conditions as well as decision values, are missing. Current studies have focused mainly on the missing of condition data but seem to ignore the missing of decision data. The common approach is to remove objects with no decision values because such objects are apparently considered fruitless from the decision-making standpoint. However, this deletion may lead to the risk of information loss. We observe that such a situation is somewhat similar to the semi-supervised situation in the sense that some objects are characterized by complete decision data while some are not. Considering both kinds of objects from a probabilistic view, we predict potential candidates for missing values by comparing measurements of two factors, local decision belief and universal decision belief, with a parameter threshold α. These possible decision candidates help to form a relative dissimilarity relation, which measures the unlikeness of pairs of objects rather than their likeness. Contrasting with the other approaches, rough set definitions based on this relation do not approximate the target set but its complement instead. The knowledge acquisition induced by the common approach and the proposed approach is compared, and the result shows that the latter can overcome some limitations of the former. This approach is new and flexible to deal with missing decision information.

Measuring visual vocabulary appropriateness by dispersion index and its improvement by globalizing SIFT descriptors

Correct classification rates are often used to measure the appropriateness of a visual-word vocabulary. Appropriateness is also measured by the dispersion index, a technique in quantitative ecology that estimates the distribution pattern of individuals. In the bag-of-keypoints method of making a visual vocabulary, the size of the vocabulary is examined. In addition, a method to globalize a local feature is proposed. In this globalization method, each descriptor is modified to be similar to adjacent descriptors and therefore provides more opportunity for a visual word to form clumps on a test image. Although applying the index is limited by the classifier, we found that appropriateness is measured by estimating the average dispersion index for every visual word. In addition, this paper discusses another method for creating a visual-word vocabulary that is more appropriate for classification by globalization.

Pattern classification using bag-of-keypoints for improper object extraction

The classifications when a target is not properly extracted due to improper segmentation include the multi-class case, in which the target contains objects belonging to different classes. In this paper, a method is applied to transform the multiclass case to a single-label classification by creating merged classes. To train merged classes, each feature must be defined in a very small domain, and the range of each feature must be binary, i.e., {0, 1}. It is not a contradiction to consider that the range of each feature is binary when the naïve Bayes classifier is employed in the bag-of-keypoints method. Thus, a fuzzy extension technique is proposed that enables us to consider the range of each feature as continuous, i.e., [0, 1]. By using the weighted average operation of the fuzzy vector, the ordinary Bayes classifier can be applied to solve multiclass cases. The experimental results verify that the classifier correctly detects 1) multi-class targets, and 2) targets in the incomplete case, in which the target is not properly extracted.