In recent years, recommender systems have been widely adopted in various applications. Such systems predict user preferences for items based on past behavior and attribute information, determining which items to present. However, if recommendations are repeatedly made based on limited data, users may be exposed only to a narrow selection of items. This can result in many items remaining undiscovered, preventing users from finding new interests and potentially leading to a less satisfying long-term experience. To address this, “Exploration” deliberately recommends items with uncertain user preferences, promoting discovery. While beneficial in the long term, exploration can negatively impact short-term user experience by suggesting items of lower immediate preference. Thus, accurately measuring exploration’s effects is crucial. Existing evaluation methods based on diversity, novelty, and serendipity face challenges as their appropriate definitions depend on user and item characteristics. This research introduces a novel approach by defining exploration from the system’s perspective using Bayesian Information Gain. We propose an evaluation metric that quantifies exploration based on how much the system improves its understanding of user preferences. This metric, relying on the uncertainty of estimated preference distributions, offers greater universality than conventional methods. Through experiments on artificial and real datasets, we demonstrate its effectiveness, providing a new, broadly applicable framework for assessing exploration in recommender systems. Implementations are available at: https://github.com/tishii2479/big-pu .

While pre-trained models, such as large language models, can achieve high performance with minimal fine-tuning, the source datasets used for pre-training often contain irrelevant or blackundant data, which can degrade performance on target tasks. Domain Adaptation Information Gain (DAIG)-based source data selection improves performance by pre-training on source data selected based on rough prior knowledge obtained from target data in advance. However, DAIG’s key component, the transition matrix, lacks flexibility and is limited to handling only single-label classification tasks. To address this limitation, we propose the Generalized DAIG (GDAIG)-guided selection process, a novel framework that extends DAIG to support multi-label classification. GDAIG introduces a soft transition matrix to capture inter-label dependencies and employs binary cross-entropy loss to enable adaptation to multi-label data. By leveraging “rough prior knowledge” from initial training on target data, GDAIG actively selects informative and task-relevant source data for pre-training. Experiments on medical image and general object classification datasets demonstrate that GDAIG consistently outperforms baseline approaches, with particularly significant improvements in scenarios involving label mismatch between source and target domains (partial or no label overlap), where conventional transfer learning methods suffer from noise caused by irrelevant source labels. These results highlight GDAIG’s ability to enhance the effectiveness of pre-trained models through strategic source data selection, thereby optimizing performance for specific target tasks. Our framework goes beyond existing approaches that rely solely on pre-trained models, emphasizing the direct utilization of task-relevant source data. Furthermore, GDAIG provides a practical and effective solution for domains with scarce labeled data, such as medical image analysis. • A GDAIG-guided data selection strategy for multi-label classification is proposed. • GDAIG improves target model performance through task-relevant multi-label data selection. • A probabilistic transition matrix captures inter-label dependencies. • “Rough prior” from target data effectively guides source data pre-training. • GDAIG outperforms conventional baselines across diverse multi-label scenarios.

Zero-shot domain-specific image classification is challenging in classifying real images without ground-truth in-domain training examples. Recent research involved knowledge from texts with a text-to-image model to generate in-domain training images in zero-shot scenarios. However, existing methods heavily rely on simple prompt strategies, limiting the diversity of synthetic training images, thus leading to inferior performance compared to real images. In this paper, we propose AttrSyn, which leverages large language models to generate attributed prompts. These prompts allow for the generation of more diverse attributed synthetic images. Experiments for zero-shot domain-specific image classification on two fine-grained datasets show that training with synthetic images generated by AttrSyn significantly outperforms CLIP’s zero-shot classification under most situations and consistently surpasses simple prompt strategies.

Recent research on explainable recommendation generally frames the task as a standard text generation problem, and evaluates models simply based on the textual similarity between the predicted and ground-truth explanations. However, this approach fails to consider one crucial aspect of the systems: whether their outputs accurately reflect the users' (post-purchase) sentiments, i.e., whether and why they would like and/or dislike the recommended items. To shed light on this issue, we introduce new datasets and evaluation methods that focus on the users' sentiments. Specifically, we construct the datasets by explicitly extracting users' positive and negative opinions from their post-purchase reviews using an LLM, and propose to evaluate systems based on whether the generated explanations 1) align well with the users' sentiments, and 2) accurately identify both positive and negative opinions of users on the target items. We benchmark several recent models on our datasets and demonstrate that achieving strong performance on existing metrics does not ensure that the generated explanations align well with the users' sentiments. Lastly, we find that existing models can provide more sentiment-aware explanations when the users' (predicted) ratings for the target items are directly fed into the models as input. The datasets and benchmark implementation are available at: https://github.com/jchanxtarov/sent_xrec.

In recent years, Large Language Model (LLM) have seen significant advancements, attracting attention for their applications in various fields. These models have shown promising results in handling tabular data, especially in cases with limited datasets, by leveraging pre-trained knowledge. However, their effectiveness in addressing imbalanced data in tabular formats is less explored. To bridge this gap, our study introduces LLMOverTab, a novel approach using LLMs for oversampling in imbalanced tabular data. We conducted comprehensive experiments on diverse tabular datasets to assess the effectiveness of LLMOverTab, demonstrating its potential in improving the handling of imbalanced data. The study also explores application of LLMOverTab in zero-shot and few-shot learning contexts, providing insights into its adaptability. Additionally, we analyze the oversampled data, offering reflections on the quality of generated samples. Our research not only showcases the utility of LLMOverTab in managing imbalanced tabular data, but also opens new avenues for the application of language models in various tasks of tabular data. This study adds to the increasing interest in applying LLMs to various task domains. It provides new perspectives for the innovative use of LLMs in structured tabular data fields, highlighting their potential in a range of applications. • Introduces LLMOverTab for oversampling in imbalanced tabular data. • Surpasses traditional methods like SMOTE and other LLM approaches. • Uses prompt engineering to generate meaningful synthetic instances. • Finds LLMOverTab excels especially with LLM prediction models. • Suggests exploring different LLM architectures and prompt techniques.

Generating realistic synthetic tabular data is a crucial task for privacy-preserving data sharing, data augmentation, and learning from limited samples. However, existing methods often struggle with small sample sizes, heterogeneous feature types, and generalization in real-world scenarios. We propose TabularMDLM, a novel diffusion-based generative framework that integrates masked language modeling to synthesize high-quality tabular data. Unlike prior work, TabularMDLM applies noise only to feature values—preserving the semantic relationship between column names and values—and leverages pre-trained language models to iteratively reconstruct masked tokens during reverse diffusion. This design enhances generation quality while ensuring structural consistency. To evaluate the framework, we conduct extensive experiments on six tabular datasets of varying sizes, domains, and feature types. We compare TabularMDLM against recent baselines, including TabDDPM, CTGAN, and CTGAN+, under a privacy-conscious setting where only synthetic data is used for training and real data for testing. We also assess performance under class imbalance to validate generalization. Results show that TabularMDLM achieves consistently strong classification performance across accuracy, precision, recall, and F1 score, outperforming baselines in both balanced and imbalanced settings. In contrast to existing methods, TabularMDLM scales to diverse data types and low-resource regimes, offering practical advantages in privacy-sensitive applications. • Introduces TabularMDLM, a generative framework combining LLMs and diffusion-inspired refinement for synthetic tabular data. • Selective masking of feature values, while preserving feature names, ensures schema integrity and richer feature interactions. • Generates more diverse and representative samples than conventional oversampling methods. • Demonstrates improved predictive accuracy and robustness, even under challenging feature compositions and severe class imbalances. • Facilitates privacy-compliant data augmentation and fairer decision-making in data-sensitive and resource-constrained scenarios.

• Domain adaptation information gain-based target-related data selection is proposed. • Proposed DAIG improves target model accuracy via target-related data selection. • DAIG effectively gathers relevant data from source to benefit target tasks. • DAIG extracts “rough prior” from target data for source data pre-training. • DAIG-guided selection outperforms baselines in multiple experimental settings. In recent years, due to the explosive popularity of large-scale pre-trained models such as large language models, pre-training approaches that use a massive amount of source data and can be applied to various target tasks are becoming more popular. Pre-trained models allow us to learn highly accurate target models by fine-tuning them with target data, even when their volume is insufficient. However, the source data used to train a pre-training model is generally a large and miscellaneous data set obtained in the wild without being aware of the target task, and it highly possibly contains much data that does not contribute to relearning the target task. This study defines a novel paradigm as “target-aware source data selection,” which uses the source data itself instead of a pre-training model and selects source data for pre-training and aims to increase its quality, effectiveness, and robustness. Our proposal fundamentally differs from the current studies addressing the lack of target data and conventional transfer learning approaches, improving source data quality using the novel Domain Adaptation Information Gain criteria. Specifically, the target model is pre-trained while actively selecting only informative data from the source data using the “rough-prior knowledge” obtained from the target data training before the pre-training. Finally, fine-tuning the model with the target data results in a highly accurate model for the target (downstream) task. The effectiveness of our proposed paradigm has been demonstrated through multifaceted experiments using multiple pairs of target data and source data with different strengths of their relevance.

Abstract In the world of data-driven decision-making, tabular data reigns supreme as the most prevalent and crucial format, especially in business contexts. However, data scarcity remains a recurring challenge. In this context, transfer learning has emerged as a potent solution. This study explores the untapped potential of transfer learning in the realm of tabular data analysis, with a focus on leveraging deep learning models—especially the Transformer model—that have garnered significant recognition. Our research investigates the intricacies of tabular data and illuminates the shortcomings of conventional attention mechanisms in the Transformer model when applied to such structured datasets. This highlights the pressing requirement need for specialized solutions tailored to tabular data. We introduce an innovative transfer learning method based on series of thoroughly designed experiments across diverse business domains. This approach harnesses Transformer-based models enhanced with optimized sparse attention mechanisms, offering a groundbreaking solution for tabular data analysis. Our findings reveal the remarkable effectiveness of enhancing the attention mechanism within the Transformer in transfer learning. Specifically, pre-training with sparse attention proves increasingly powerful as data volumes increase, resulting in superior performance on large datasets. Conversely, fine-tuning with full attention becomes more impactful when data availability decreases in downstream tasks, ensuring adaptability in situations with limited data. The empirical results presented in this study provide compelling evidence of the revolutionary potential of our approach. Our optimized sparse attention model emerges as a powerful tool for researchers and practitioners seeking highly effective solutions for tabular data tasks. As tabular data remain the backbone of business operations, our study promises to revolutionize data analysis in critical domains. This work bridges the gap between limited data availability and the requirement for effective analysis in business settings, marking a significant step forward in the field of tabular data analysis.

Abstract Understanding products and customers is a critical challenge for efficient business operations. While various machine learning-based analytical methods have been proposed, most rely on objective metrics such as evaluation scores or tags. However, estimating subjective evaluation scores is also an essential aspect of understanding customers, yet research in this area remains limited. Moreover, it is well-known that directly evaluating the subjective scores of targets is challenging. Consequently, traditional methods have used pairwise comparisons between targets to estimate true evaluation scores. However, as the number of targets increases, the required number of pairwise comparisons grows exponentially, making it difficult to estimate subjective evaluations for a large number of targets using conventional methods. To address this issue, this study proposes a scalable model for subjective evaluation score estimation by completing pairwise comparison data using a deep learning model trained on a limited number of annotations. Specifically, the deep learning model is trained on pairwise comparison results from a subset of evaluation target combinations annotated by humans, and the model’s predictions are used to complete the pairwise comparison matrix. The effectiveness and practical applicability of the proposed method are demonstrated through applications to multiple real-world datasets.

In recent years, the task of learned point cloud compression has gained prominence. An important type of point cloud, LiDAR point cloud, is generated by spinning LiDAR on vehicles. This process results in numerous circular shapes and azimuthal angle invariance features within the point clouds. However, these two features have been largely overlooked by previous methodologies. In this paper, we introduce a model-agnostic method called Spherical-Coordinate-based learned Point cloud compression (SCP), designed to fully leverage the features of circular shapes and azimuthal angle invariance. Additionally, we propose a multi-level Octree for SCP to mitigate the reconstruction error for distant areas within the Spherical-coordinate-based Octree. SCP exhibits excellent universality, making it applicable to various learned point cloud compression techniques. Experimental results demonstrate that SCP surpasses previous state-of-the-art methods by up to 29.14% in point-to-point PSNR BD-Rate.

In recent years, it has become common for consumers to familiarize themselves with the latest fashion trends through the internet and engage in their own fashion-inspired shopping activities. Therefore, making fashion-inspired shopping and browsing activities (internet surfing in the fashion domain) comfortable is essential because it leads to interactions in the fashion industry. However, fashion is a fuzzy and complex domain that contains many abstract elements, and this ambiguity and complexity can hinder users’ deep interest in the fashion industry. Therefore, we define a novel technology and domain called “fashion intelligence” and propose a system based on a visual-semantic embedding method for automatically learning and interpreting fashion and obtaining answers to users’ questions. Our proposed method can embed the abundant abstract tag information in the same projective space as outfit images. Mapping of images and tags in a projective space helps search for outfit images using fashion-specific abstract words. In addition, visually estimating the degree of relevance between images and tags helps interpret abstract words. As a result, this research helps decrease fashion-specific ambiguity and complexity and supports the marketing activities and fashion choices of both experts and non-experts.

Error-correcting output coding (ECOC) is a method for constructing a multi-valued classifier using a combination of given binary classifiers. ECOC can estimate the correct category by other binary classifiers even if the output of some binary classifiers is incorrect based on the framework of the coding theory. The code word table representing the combination of these binary classifiers is important in ECOC. ECOC is known to perform well experimentally on real data. However, the complexity of the classification problem makes it difficult to analyze the classification performance in detail. For this reason, theoretical analysis of ECOC has not been conducted. In this study, if a binary classifier outputs the estimated posterior probability with errors, then this binary classifier is said to be noisy. In contrast, if a binary classifier outputs the true posterior probability, then this binary classifier is said to be noiseless. For a theoretical analysis of ECOC, we discuss the optimality for the code word table with noiseless binary classifiers and the error rate for one with noisy binary classifiers. This evaluation result shows that the Hamming distance of the code word table is an important indicator.

A novel technology named fashion intelligence system, which quantifies ambiguous expressions unique to fashion, such as “casual,” “adult-casual,” and “office-casual,” was previously proposed to support users in their understanding of fashion. However, the existing visual-semantic embedding (VSE) model, which forms the basis of the system, does not support images that are composed of multiple parts, such as those containing hair, tops, trousers, skirts, and shoes. Therefore, we propose a partial VSE (PVSE) model, which enables fine-grained learning of each part of the fashion outfit. The proposed model learns embedded representations via angular-based contrastive learning. This helps in retaining three existing practical functionalities and further enables image-retrieval tasks where changes are only made to specified parts and image-reordering tasks focusing on the specified parts. In other words, the proposed model enables five types of practical functionalities, even with a simple structure. Through qualitative and quantitative experiments, we demonstrate that the proposed model is superior to conventional models, without increasing computational complexity.

Consider M-valued (M$\geq$3) classification systems realized by combination of N(N$\geq\lceil\log_{2}$M$\rceil$) binary classifiers. Such a construction method is called an Error Correcting Output Code (ECOC). First, focusing on a Reed-Muller (RM) code, we derive a modified RM (mRM) code to make it suitable for the ECOC. Using the mRM code and the Hadamard matrix, we introduce a simplex code which is one of the powerful equidistant codes. Next, from the viewpoint of system evaluation model, we evaluate the ECOC by using constructive coding described above. We show that they have desirable properties such as Flexible, Elastic, and Effective Elastic as M becomes large, by employing analytical formulas and experiments.

With the development of information technology, a wide variety of data have been accumulated, and there are many methods for analyzing such data. In this study, we model the input data and the metrics between the data based on the assumption that each metric is generated from a continuous latent variable. Specifically, we assume that the input data are generated using low-dimensional latent variables and their projection matrices. We describe a method for estimating the latent variables. Because the generative model defined in this study cannot obtain the Q function analytically, we use the Monte Carlo EM algorithm to approximate the Q function and investigate an efficient parameter estimation method. Experiments using artificial data and the 20 newsgroups dataset demonstrate the effectiveness of the proposed method.

In recent years, explainable recommendation has been a topic of active study. This is because the branch of the machine learning field related to methodologies is enabling human understanding of the reasons for the outputs of recommender systems. The realization of explainable recommendation is widely expected to increase both user satisfaction and the demand for explainable recommendation systems. Explainable recommendation utilizes a wealth of side information (such as sellers, brands, user ages and genders, and bookmark information, among others) to expound the decision-making reasoning applied by recommendation models. In explainable recommendation, although learning side information containing numerous variables leads to rich interpretability, learning too many variables presents a challenge because decreases the amount of learning that a given computational resource can perform, and the accuracy of the recommendation model may be degraded. However, numerous and diverse variables are included in the side information stored by the actual companies operating massive real-world services. Hence, to realize practical applications of this valuable information, it is necessary to resolve problems such as computational cost. In this study, we propose a new framework for explainable recommendation based on an improved knowledge graph attention network model, which utilizes the side information of items and realizes high recommendation accuracy. The proposed framework enables direct interpretation by visualizing the reasons for the recommendations provided. Experimental results show that the proposed framework reduced computational time requirements by approximately 80%, while maintaining recommendation accuracy by enabling the model to learn the probabilistically given edges included in the graph structure. Moreover, the results show that the proposed framework exhibited richer interpretability than the conventional model. Finally, a multifaceted analysis suggests that the proposed framework is not only effective as an explainable recommendation model but also provides a powerful tool for planning various marketing strategies.

With the development of IoT technology, it has become possible to accumulate and regularly measure multidimensional time-series data. In this study, we focus on the usage of multidimensional time-series data from printer products' log data and propose a method for its analysis. In addition to the number of sheets printed by each customer, the log data includes various time-series information such as the amount of remaining toner, the number of stoppages that occur, and the activation times. To utilize these data for business purposes, it is desirable to construct a model for predicting future changes in use characteristics for each customer. In this study, we apply the random forest algorithm to predict such changes. However, if all measurable features of the problem are included, the model becomes complex and cannot be interpreted. Although the accuracy is relatively high if an appropriate learning algorithm is applied, the complex model tends to overfit the training data. In this paper, we propose a method to select the modeling features that can be interpreted by graph mining while maintaining accuracy. This would enable us to interpret the data at the field level and discover the hypotheses that are necessary for planned marketing policies. Finally, the proposed method is applied to real data and its efficacy is demonstrated.

To solve multiple classification problems with $M (\geq$ 3) categories, many studies have been devoted using $N (\geq\ \lceil\log_{2}M\rceil)$ binary $(\{0,1\})$ classifiers, where these systems are known as binary Error-Correcting Output Codes (binary ECOC). As an extended version of the binary ECOC, the ternary $(\{0,\ *,\ 1\})$ ECOC have also been discussed, where ternary classifiers classify data into positive examples when the element is 1, into negative examples when the element is 0, and no classification when the element is $*$. In this paper, we discuss the ternary ECOC system from the view point of the system evaluation model based on rate-distortion function. First, we discuss a table of M code words with length N which is given by a ternary matrix W of M rows and N columns. Next, by leveraging the benchmark data for multiclass document classification which is widely used in Japan, the relationships between the probability of classification error Peand the number of the ternary classifiers N for a given M are experimentally investigated. In addition, by assuming the M-dimensional Normal distribution for a classification data model, the relationship between Peand N for a given M is also examined. Finally, we show by the system evaluation model that the ternary ECOC systems have desirable properties such as “Flexible”, “Elastic”, and “Effective Elastic”, when M becomes large.

In recent years, the paradigm of social coding in software development has attracted attention to developers all over the world, and GitHub which is a social coding tool has spread to the area like education. There are many cases using it as a platform of PBL (Project Based Learning). However, since GitHub is not a tool for education, it is difficult to evaluate learners. This research focuses on the contribution of learners and proposes a system that teachers can grasp the contribution of learners.

In this manuscript, we investigate a collaborative filtering method to characterize consumption behavior of customers and services with various attributes for marketing. We assume that each customer and service have the invisible attribute which is called latent class. Assuming a combination of attribute values of a customer and service is classified to a latent class, furthermore, we propose a new Bayesian statistical model that consumption behavior is probabilistically arise based on a latent class combination of a customer, service and attribute values. Then, we show the method to estimate parameters of a statistical model based on the variational Bayes method and the mean field approximation. Consequently, we show the effectiveness of the proposed model and the estimation method by simulation.

In this manuscript, we investigate a collaborative filtering method to characterize consumption behavior (or evaluation) of customers (or users) and services (or items) for marketing. Assuming that each customer and service have the invisible attribute, which is called latent class, we propose a new Bayesian statistical model that consumption behavior is probabilistically arise based on a latent class combination of a customer and service. Then, we show the method to estimate parameters of a statistical model based on the variational Bayes method and the mean field approximation. Consequently, we show the effectiveness of the proposed model and the estimation method by simulation and analyzing actual data.

In this paper, we focus on pattern recognition based on the vector space model. As one of the methods, distance metric learning is known for the learning metric matrix under the arbitrary constraint. Generally, it uses iterative optimization procedure in order to gain suitable distance structure by considering the statistical characteristics of training data. Most of the distance metric learning methods estimate suitable metric matrix from all pairs of training data. However, the computational cost is considerable if the number of training data increases in this setting. To avoid this problem, we propose the way of learning distance metric by using the each category centroid. To verify the effectiveness of proposed method, we conduct the simulation experiment by using benchmark data.

This paper proposes a new latent class model to represent user segments in a marketing model of electric commerce sites. The aspect model proposed by T. Hofmann is well known and is also called the probabilistic latent semantic indexing (PLSI) model. Although the aspect model is one of effective models for information retrieval, it is difficult to interpret the meaning of the probability of latent class in terms of marketing models. It is desirable that the probability of latent class means the size of customer segment for the purpose of marketing research. Through this formulation, the simulation analysis to dissect the several situations become possible by using the estimated model. The impact of the strategy that we contact to the specific customer segment and make effort to increase the number of customers belonging to this segment can be predicted by using the model demonstrating the size of customer segment. This paper proposes a new model whose probability parameter of latent variable means the rate of users with the same preference in market. By applying the proposed model to the data of an internet portal site for job hunting, the effectiveness of our proposal is verified.

F. P. Preparata et al. proposed a fault diagnosis model (PMC model) to find all fault units in the multicomputer system by using outcomes that each unit tests some other units. T. Kohda proposed a highly structured(HS) system and the syndrome analysis method(SAM) to diagnose from local testing results. In this paper, we introduce the maximum a posteriori probability algorithm(MAPDA) for the HS system in the probabilistic fault model. Analyzing the MAPDA, we show that the SAM is closer to the MAPDA as the fault probability becomes smaller. Finally, we show the robustness of the SAM in the HS system.

In this paper, we focus on pattern recognition based on the vector space model with the high dimensional and sparse data. One of the pattern recognition methods is metric learning which learns a metric matrix by using the iterative optimization procedure. However most of the metric learning methods tend to cause overfitting and increasing computational time for high dimensional and sparse settings. To avoid these problems, we propose the method of l1regularized metric learning by using the algorithm of alternating direction method of multiplier (ADMM) in the supervised setting. The effectiveness of our proposed method is clarified by classification experiments by using the Japanese newspaper article and UCI machine learning repository. And we show proposed method is the special case of the statistical sparse covariance selection.

This study proposes a novel technique to display tactile sensation using electro-rheological fluid (ERF). ERF changes its rheological characteristics according to the electric field applied. The ERF used in this study generates relatively high yield stress and behaves as a solid when subjected to a strong electric field. Using this solid-liquid phase transition, we propose a novel device which provides a tactile sensation, i.e., tactile bump display. Applying electric field at a specific position in an ERF chamber, the corresponding ERF behaves as solid at the position. This solid-state ERF gives tactile sensation like a physical bump to a user. We fabricate in this study a prototype, which may realize the above-mentioned idea, and characterize it. We obtained the following results by experiments. First, the tactile bump display could make the users recognize their finger position on a flat surface by creating a tactile bump. Second, we confirmed that the tactile bump might significantly improve the accuracy and the precision of touch typing.

This paper discusses a new similarity measure between documents on a vector space model from the view point of distance metric learning. The documents are represented by points in the vector space by using the information of frequencies of words appearing in each document. The similarity measure between two different documents is useful to recognize the relationship and can be applied to classification or clustering of the data. Usually, the cosine similarity and the Euclid distance have been used in order to measure the similarity between points in the Euclidean space. However, these measures do not take the correlation among words which appear in documents into consideration on an application of the vector space model to document analysis. Generally speaking, many words which appear in documents have correlation to one another depending on the sentence structures, topics and subjects. Therefore, it is effective to build a suitable metric measure taking the correlation of words into consideration on the vector space in order to improve the performance of document classification and clustering. This paper presents a new effective method to acquire a distance measure on the document vector space based on an extended cosine measure. In addition, the way of distance metric learning is proposed to acquire the proper metric from the view point of supervised learning. The effectiveness of our proposal is clarified by simulation experiments for the text classification problems of the customer review which is posted on the web site and the newspaper article.

In this paper, we present a new mathematical model based on a “Vector Space Model” and consider its implications. The proposed method is evaluated by performing several experiments. In these experiments, we classify newspaper articles from the English Reuters-21578 data set, and Taiwanese China Times 2005 data set using the proposed method. The Reuters-21578 data set is a benchmark data set for automatic text categorization. It is shown that FRAM has good classification accuracy. Specifically, the micro-averaged F-measure of the proposed method is 94.5% for English. However, that is 78.0% for Taiwanese. Though the proposed method is language-independent and provides a new perspective, our future work is to improve classification accuracy for Taiwanese.

In our previous paper, we proposed a new classification technique called the Frequency Ratio Accumulation Method (FRAM). This is a simple technique that adds up the ratios of term frequencies among categories, and it is able to use index terms without limit. Then, we adopted the Character N-gram to form index terms, thereby improving FRAM. However, FRAM did not have a satisfactory mathematical basis. Therefore, we present here a new mathematical model based on a “Vector Space Model” and consider its implications. The proposed method is evaluated by performing several experiments. In these experiments, we classify newspaper articles from the English Reuters-21578 data set, a Japanese CD-Mainichi 2002 data set using the proposed method. The Reuters-21578 data set is a benchmark data set for automatic text categorization. It is shown that FRAM has good classification accuracy. Specifically, the micro-averaged F-measure of the proposed method is 92.2% for English. The proposed method can perform classification utilizing a single program and it is language-independent.

In this paper, we consider the Bayesian approach for representation of a set of documents. In the field of representation of a set of documents, many previous models, such as the latent semantic analysis (LSA), the probabilistic latent semantic analysis (PLSA), the semantic aggregate model (SAM), the Bayesian latent semantic analysis (BLSA), and so on, were proposed. In this paper, we formulate the Bayes optimal solutions for estimation of parameters and selection of the dimension of the hidden latent class in these models and analyze it's asymptotic properties.

In this paper, we discuss the difference in code lengths between the code based on the minimum description length (MDL) principle (the MDL code) and the Bayes code under the condition that the same prior distribution is assumed for both codes. It is proved that the code length of the Bayes code is smaller than that of the MDL code by o(1) or O(1) for the discrete model class and by O(1) for the parametric model class. Because we can assume the same prior for the Bayes code as for the code based on the MDL principle, it is possible to construct the Bayes code with equal or smaller code length than the code based on the MDL principle. From the viewpoint of mean code length per symbol unit (compression rate), the Bayes code is asymptotically indistinguishable from the MDL two-stage codes.