The superior functionality and versatility of generative AI have raised expectations for the improvement of human society and concerns about the ethical and social risks associated with the use of generative AI. Many previous studies have presented risk issues as concerns associated with the use of generative AI, but since most of these concerns are from the user's perspective, they are difficult to lead to specific countermeasures. In this study, the risk issues presented by the previous studies were broken down into more detailed elements, and risk factors and impacts were identified. In this way, we presented information that leads to countermeasure proposals for generative AI risks.

Recently, various risks have been pointed out in artificial intelligence (AI) systems. In particular, AI security, AI ethics, and AI quality are considerable risks. To make AI systems trustworthy against these risks, risk assessment technology is needed to identify potential AI risks and decide which risks should be dealt with in priority. We propose a risk assessment technology that assesses three kinds of risks—AI security, AI ethics, and AI quality—which have been considered separately. Our technology follows the ISO 31000 framework, consisting of four phases: risk candidate identification, impact assessment, likelihood assessment, and priority derivation for countermeasures. To realize this technology, risk candidate identification is achieved by extending AI ethics impact assessment— an identification method of AI ethics risk—to AI security and AI quality. Impact and likelihood assessments are conducted by extending assessment methods for AI security to AI ethics and AI quality. We conducted a case study using our technology and confirmed that the risks were appropriately extracted, and the priority of the risks to be dealt with was derived.

Machine learning system analysis requires different approaches for each different task and domain. Selecting a proper set of analytic models can be challenging for a specific problem. This paper discusses the extensibility of the Multi-View Modeling Framework for ML Systems approach using process mapping and extensible metamodel. We conducted a case study to evaluate the feasibility of such extensibility by extending the approach to facilitate an activity-driven analysis for an optical character recognition system. Based on the result of the case study, we found that Multi-View Modeling Framework for ML Systems is likely to be extensible.

The mixing of other people's voices in daily conversations and remote meetings causes a large obstacle to the voice that you originally want to hear, regardless of the volume. In this paper, the voice of another specific speaker added to the voice of the speaker is regarded as noise, and a method of removing the noise and the experimental result are shown. We created a learning model for removing the mixed noise using U-Net and the data obtained by converting the voice data before and after the noise was mixed into an image by STFT. Then, the effect was confirmed by comparing the image data, and voice data obtained by the reverse STFT with the ones before the removal.

In order to extend the usability of BPLOT, a new version BPLOT4 was developed. BPLOT is the first tactile graphics production system for the blind that enables the blind to produce tactile graphics by themselves. BPLOT4 has obtained the following four new features: (1) The usage of variables and mathematical formulas is supported for the description of plotter control commands. (2) The production of ink print graphics is possible at the same time as that of tactile graphics. (3) A new GUI (Graphical User Interface) is developed for sighted users. (4) Everest-D V5 (Index Braille) is added to the list of supported Braille printers.

We present a novel mathematical model based on the concepts of progression and retrogression. The goal of the model is to provide a more general description of the motor learning process. The model structure is given as a set of two autonomous linear differential equations with two coefficients that characterize the learning process. The coefficients are estimated using data obtained from the learning process. This paper examines how well the data complied with the model and whether the model reflects the human learning process with three biofeedback learning methods. To evaluate them, trainees used our prepared motor learning program for the motion of “walking without vertical acceleration of the head.” 82.2% of the learning process data complied with the assumptions of the P-R model. To evaluate the impact of feedback, trainees used three motor learning programs. Audio feedback was perceived as “clearer,” the somatosensory feedback as “clear,” and the visual feedback as “fuzzy,” for learning the motion, and the results were evaluated based on the progression speed normalized by retrogression speed calculated by the model, which was 0.25 for the visual feedback learning program, 0.28 for the somatosensory feedback, and 0.69 for the audio feedback. These values had statistically significant differences. Thus, the descriptive model can express successive repeated learning processes under progression and retrogression components.

This paper describes a novel application of the bidirectional audio microphones in mobile phones to measure biosignals for ubiquitous health monitoring. The bidirectional microphone has two pressure-detecting ports. A pressure-detecting film receives the differential pressure between the ports. We apply this idea to biosignal measurements for a mobile phone. We used the microphone as 1) a sphygmograph at the fingertip, 2) a stethoscope in which a delay time sensor of pulse wave arrivals from the neck near the carotid or chest near the heart to the fingertip, and 3) a bed sensing device for biosignals. When used as a sphygmograph, it detects the wave similar to the second derivative of the conventional optical pulse oxymeter output wave. When used as a stethoscope, it detects cardiac sound and the pulse wave and respiration. With two microphones, the delay time from the neck carotid to the fingertip is detected.

This paper presents a method to extract cells from form documents of an unknown format based on intersection tracking and global optimization. Because of the increasing varieties of forms and complications in the forms, cell extraction has become a key factor in automatic form recognition. The proposed method produces cell candidates by using intersection features and tracing intersection points to form closed regions. Then, global optimization is carried out to determine a set of cells having the highest likelihood. Maintaining precision for 82 kinds of forms, the proposed method achieved a recall rate of 80.4% whereas the conventional method achieved a recall rate of 77.1%.

The judgment standards of R-K method include ambiguities and are thus compensated by subjective interpretations of sleep-stage scorers. This paper presents a novel method to compensate uncertainties in judgments by the subjective interpretations by the sleep-model estimation approach and by describing the judgments in probabilistic terms. Kalman filter based on the two sleep models with no body movement and with body movement was designed. Sleep stages judged by three different scorers were rejudged by the filter. The two sleep models were stochastically estimated from biosignals from 15 nights' data and the rejudged scores by the filter were evaluated by the data from 5 nights. The average values of kappa statistics, which show the degree of agreement, were 0.85, 0.89, and 0.81, respectively, for the original sleep stages. Because the new method provides probabilities on how surely the sleep belongs to each sleep stage, we were able to determine the most, second most, and third most probable sleep stage. The kappa statistics between the most probable sleep stages were improved to 0.90, 0.93, and 0.84, respectively. Those of sleep stages determined from the most and second most probable were 0.92, 0.94, and 0.89 and those from the most, second most, and third most probable were 0.95, 0.97, and 0.92. The sleep stages estimated by the filter are expressed by probabilistic manner, which are more reasonable in expression than those given by deterministic manner. The expression could compensate the uncertainties in each judgments and thus were more accurate than the direct judgments.

In this paper, a new threshold correction method for document image binarization that is forcused on ruled-line extraction is presented. This method enhances the binary image of a ruled line, which is often adversely influenced by adjacent text pixels or background noise. The threshold correction method consists of two submethods. One is a noise reduction method that is based on background determination, and the other is a threshold surface conversion method. Both these methods use the aspect of local straightness feature to distinguish ruled-line pixels from background pixels.

SUMMARY: BioCichlid is a 3D visualization system of time-course microarray data on molecular networks, aiming at interpretation of gene expression data by transcriptional relationships based on the central dogma with physical and genetic interactions. BioCichlid visualizes both physical (protein) and genetic (regulatory) network layers, and provides animation of time-course gene expression data on the genetic network layer. Transcriptional regulations are represented to bridge the physical network (transcription factors) and genetic network (regulated genes) layers, thus integrating promoter analysis into the pathway mapping. BioCichlid enhances the interpretation of microarray data and allows for revealing the underlying mechanisms causing differential gene expressions. AVAILABILITY: BioCichlid is freely available and can be accessed at http://newton.tmd.ac.jp/. Source codes for both biocichlid server and client are also available.

BACKGROUND: Analysis of within-patient HIV evolution under anti-HIV therapy is crucial to a better understanding the possible mechanisms of HIV drug-resistance acquisition. The high evolutionary rate of HIV allows us to trace its evolutionary process in real time by analyzing virus samples serially collected from the same patient. However, such studies are still uncommon due to the lack of powerful computational methods designed for serial virus samples. In this study, we develop a computational method, vSPA (viral Sequential Pathway Analysis), which groups viral sequences from the same sampling time into clusters and traces the evolution between clusters over sampling times. The method makes use of information of different sampling times and traces the evolution of important amino acid mutations. Second, a permutation test at the codon level is conducted to determine the threshold of the correlation coefficient for clustering viral quasispecies. We applied vSPA to four large data sets of HIV-1 protease and reverse transcriptase genes serially collected from two AIDS patients undergoing anti-HIV therapy over several years. RESULTS: The results show that vSPA can trace within-patient HIV evolution by detecting many amino acid changes, including important drug-resistant mutations, and by classifying different viral quasispecies coexisting during different periods of the therapy. CONCLUSION: Given that many new anti-HIV drugs will be available in the near future, vSPA may be useful for quickly providing information on the acquisition of HIV drug-resistant mutations by monitoring the within-patient HIV evolution under anti-HIV therapy as a computational approach.

Protein-protein interaction networks (PINs) are rich sources of information that enable the network properties of biological systems to be understood. A study of the topological and statistical properties of budding yeast and human PINs revealed that they are scale-rich and configured as highly optimized tolerance (HOT) networks that are similar to the router-level topology of the Internet. This is different from claims that such networks are scale-free and configured through simple preferential-attachment processes. Further analysis revealed that there are extensive interconnections among middle-degree nodes that form the backbone of the networks. Degree distributions of essential genes, synthetic lethal genes, synthetic sick genes, and human drug-target genes indicate that there are advantageous drug targets among nodes with middle- to low-degree nodes. Such network properties provide the rationale for combinatorial drugs that target less prominent nodes to increase synergetic efficacy and create fewer side effects.

This study is intended to build a controller of redundant manipulator with the simultaneous ability of trajectory tracking and obstacle avoidance without any preparation of path planning to achieve full automation even for one production of one kind, while keeping the avoidance ability high and keeping its shape away from obstacle to reduce the possibility that the manipulator crashes to the obstacle. To evaluate the avoidance ability of the intermediate links, a scalar value of avoidance manipulability shape index (AMSI) was proposed, which is independent of the obstacle's shape. On the other hand, the danger to crash to the obstacle is depending on the shape of the obstacle, which could be evaluated by the potential field set around the obstacle. This paper just proposes a control method of the manipulator's shape based on avoidance manipulability shape index with potential (AMSIP) to simultaneously avoid obstacle and keep the avoidance ability high while tracking the hand trajectory

This paper proposes trajectory tracking and obstacle avoidance control using avoidance manipulability of redundant manipulators. We propose two evaluation indexes of whole shape of the redundant manipulators. First one is sum of singular value of avoidance matrix. Second one is sum of the volume of avoidance manipulability ellipsoid. We evaluate whole manipulator's shape by two indexes and compare the ability. We found the second one is better than the first. We call the second index avoidance manipulability shape index (AMSI). Finally, by using AMSI, we construct trajectory tracking/obstacle avoidance control system. And we confirm the performance of trajectory tracking by simulations.

A novel concept for a protein-ligand docking simulator using Virtual Reality (VR) technologies, in particular the tactile sense technology, was designed and a prototype was developed. Most conventional docking simulators are based on numerical differential calculations of the total energy between a protein and a ligand. However, the basic concept of our method differs from that of conventional simulators. Our design utilizes the force between a ligand and a protein instead of the total energy. The most characteristic function of the system is its ability to enable the user to 'touch' and sense the electrostatic potential field of a protein molecule. The user can scan the surface of a protein using a globular probe, which is given an electrostatic charge, and is controlled by a force feedback device. The electrostatic force between the protein and the probe is calculated in real time and immediately fed back into the force feedback device. The user can easily search interactively for positions where the probe is strongly attracted to the force field. Such positions can be regarded as candidate sites where functional groups of ligands corresponding to the probe can bind to the target protein. Certain limitations remain; for example, only twenty protein atoms can be used to generate the electrostatic field. Furthermore, the system can only use globular probes, preventing drug molecules or small chemical groups from being simulated. These limitations are the result of our insufficient computer resources. However, our prototype system has the potential to become a novel application method as well as being applicable to conventional VR technologies, especially to force feedback technologies.

We propose a new computing model, computational living systems. We use the chemical system in our model as it is the basic system of living things. The real biochemical system is so complicated that it is hard to deal with. We use an abstract chemical system by using the multiset rewriting system, Abstract Rewriting System on Multisets (ARMS) that is also a multiset transform system. Considering that a membrane is an important structure for a living thing to separate 'self' from its environment as well as many organelles inside are composed of membranes, we introduce the membrane structure in ARMS. We further develop an artificial cell system (ACS) and investigate the behavior of ACS under various environments by introducing a genetic method and using genetic programming.

Describes a handwritten character recognition system that integrates offline recognition requiring a bitmap image and online recognition involving an input pattern as a sequence of x-y coordinates. Offline recognition performs well for painted or overwritten patterns (for which online recognition would not be suited), whereas online recognition is suitable for very deformed patterns (for which offline recognition is not suited). Because each method has different recognition capabilities, the methods complement each other when integrated together. We have implemented a hybrid handwritten character recognition system in which the recognition results of the offline and online recognizer are integrated to create an improved product. After testing several integration methods for a handwritten character database, we found that the best method increased the recognition rate from 73.8% (offline) and 84.8% (online) to 87.6% (integrated).

For efficient and integrated network management of SDH/PDH mixed large-scale transport networks, KDD has developed a TMN-based integrated management system. This paper describes an architecture of the system and its implementation. The features of the architecture are: (1) to employ a distributed system configuration based on TMN multi-layered functional architecture, (2) to newly define an integrated management information model which unifies different management information models for many kinds of NEs (Network Elements) and which enables us to manage a network with mixed SDH and PDH hierarchies of, and (3) to introduce a new alarm correlation mechanism for high system performance.

In order to acquire knowledge from databases, there have been proposed several methods of inductive learning, such as ID3 family and AQ family. These methods are applied to discover meaningful knowledge from large databases, and their usefulness is e

Automated knowledge acquisition is an important research issue in machine learning. Several methods of inductive learning, such as ID3 family and AQ family, have been applied to discover meaningful knowledge from large databases and their usefulness is assured in several aspects. However, since their methods are of a deterministic nature and the reliability of acquired knowledge is not evaluated statistically, these methods are ineffective when applied to domains essentially probabilistic in nature, such as medical domains. Extending concepts of rough set theory to a probabilistic domain, we introduce a new approach to knowledge acquisition, which induces probabilistic rules based on rough set theory (PRIMEROSE) and develop a program that extracts rules for an expert system from a clinical database, using this method. The results show that the derived rules almost correspond to those of the medical experts.

Medical data consist of many kinds of data from different resources, such as natural language data, sound data from physical examinations, numerical data from laboratory examinations, time-series data from monitoring systems, and medical images (e.g. X-ray, Computer Tomography, and Magnetic Resonance Image). Therefore it has been pointed out that medical databases should be implemented as multidatabases. However, there have been few systems which integrate these data into multidatabases. In this paper, we report a system called COBRA (Computer-Operated Birth-defect Recognition Aid), which supports diagnosis and information retrieval of congenital malformation diseases and which also integrates natural language data, sound data, numerical data, and medical images into multidatabases on syndrome of congenital malformation. The results show that object-oriented scheme makes it easy to implement and integrate these knowledge-databases in COBRA, which suggests that these clinical databases should be implemented as object-oriented databases.

Describes the anatomical-map (A-Map), a knowledge model of the human brain atlas, and shows its application to the interpretation of CT images acquired under varying degrees of individual differentiation. The A-Map is employed to integrate the information obtained from CT image data with that derived from knowledge about anatomy, physiology, and radiology. The authors present the results of an experiment whereby the A-Map was used to interpret anatomical parts in CT brain images.>