Acquiring plausible pathways on high-dimensional structural distributions is beneficial in several domains. For example, in the drug discovery field, a protein conformational pathway, i.e. a highly probable sequence of protein structural changes, is useful to analyze interactions between the protein and the ligands, helping to create new drugs. Recently, a state-of-the-art method in drug discovery was presented, which efficiently computes protein pathways using latent variables obtained from an isometric auto-encoding of the space of 3D density maps associated to protein conformations. However, our preliminary experiments show that there is room to significantly reduce the computing time. In this study, we use the Mapper algorithm, which is a Topological Data Analysis method, and present a novel variant to extract plausible conformational pathways from the isometric latent space with comparatively short running time. The extracted pathways are visualized as paths on the resulting Mapper graph. The methodological novelties are described as follows: firstly, the filter function of the Mapper algorithm is optimized so as to extract the pathways via minimization of an energy loss defined on the Mapper graph itself, while filter functions taken in the classical Mapper algorithm are fixed beforehand. The optimization is with respect to parameters of a deep neural network in the filter. Secondly, the clustering method, which defines the vertices and edges of the Mapper graph, of our algorithm is designed by incorporating domain prior knowledge to assist the extraction. In our numerical experiments, based on an isometric latent space built on the common 50S-ribosomal dataset, the resulting Mapper graph successfully includes all the well-recognized plausible pathways. Moreover, our running time is much shorter than the above state-of-the-art counterpart.

The training data distribution is often biased towards objects in certain orientations and illumination conditions. While humans have a remarkable capability of recognizing objects in out-of-distribution (OoD) orientations and illuminations, Deep Neural Networks (DNNs) severely suffer in this case, even when large amounts of training examples are available. Neurons that are invariant to orientations and illuminations have been proposed as a neural mechanism that could facilitate OoD generalization, but it is unclear how to encourage the emergence of such invariant neurons. In this paper, we investigate three different approaches that lead to the emergence of invariant neurons and substantially improve DNNs in recognizing objects in OoD orientations and illuminations. Namely, these approaches are (i) training much longer after convergence of the in-distribution (InD) validation accuracy, i.e., late-stopping, (ii) tuning the momentum parameter of the batch normalization layers, and (iii) enforcing invariance of the neural activity in an intermediate layer to orientation and illumination conditions. Each of these approaches substantially improves the DNN's OoD accuracy (more than 20% in some cases). We report results in four datasets: two datasets are modified from the MNIST and iLab datasets, and the other two are novel (one of 3D rendered cars and another of objects taken from various controlled orientations and illumination conditions). These datasets allow to study the effects of different amounts of bias and are challenging as DNNs perform poorly in OoD conditions. Finally, we demonstrate that even though the three approaches focus on different aspects of DNNs, they all tend to lead to the same underlying neural mechanism to enable OoD accuracy gains - individual neurons in the intermediate layers become invariant to OoD orientations and illuminations. We anticipate this study to be a basis for further improvement of deep neural networks' OoD generalization performance, which is highly demanded to achieve safe and fair AI applications.

Stream-based active learning (AL) is an efficient training data collection method, and it is used to reduce human annotation cost required in machine learning. However, it is difficult to say that the human cost is low enough because most previous studies have assumed that an oracle is a human with domain knowledge. In this study, we propose a method to replace a part of the oracle’s work in stream-based AL by self-training with weak labeling using a robot arm. A camera attached to a robot arm takes a series of image data related to a streamed object, which should have the same label. We use this information as a weak label to connect a pseudo-label (estimated class label) and a target instance. Our method selects two data from a series of image data; high confidence data for correcting pseudo-labels and low confidence data for improving the performance of the classifier. We paired a pseudo-label provided to high confidence data with a target instance (low confidence data). By using this technique, we mitigate the inefficiency in self-training, that is, difficulty in creating pseudo-labeled training data with a high impact on the target classifier. In the experiments, we employed the proposed method in the classification task of objects on a belt conveyor. We evaluated the performance against human cost on multiple scenarios considering the temporal variation of data. The proposed method achieves the same or better performance as the conventional methods while reducing human cost.

In recent years, one of the most widely used authentication methods is a password-based authentication method. In this method, users are required to create a secure (i.e.\ difficult to crack) and memorable (i.e.\ easy to remember) password when they create one. Taking account of these two important requirements, a mnemonic phrase-based password has been proposed. However, it is easy to crack a password if the users adopt famous phrases (e. g.\ music lyrics, movie quotes) to create a mnemonic phrase-based password. In this paper, we propose an associating image-based password creating method to create a password which is difficult to crack and easy to remember. Furthermore, we propose and develop a password creating support system for our method.

In this paper, our research objective is to develop a database virtualization technique so that data analysts or other users who apply data mining methods to their jobs can use all ubiquitous databases in the Internet as if they were recognized as a single database, thereby helping to reduce their workloads such as data collection from the Internet databases and data cleansing works. In this study, firstly we examine XML scheme advantages and propose a database virtualization method by which such ubiquitous databases as relational databases, object-oriented databases, and XML databases are usable, as if they all behaved as a single database. Next, we show the method of virtualization of ubiquitous databases can describe ubiquitous database schema in a unified fashion using the XML schema. Moreover, it consists of a high-level concept of distributed database management of the same type and of different types, and also of a location transparency feature. Finally, we propose a database incompatibility trouble-recovery technique for use in a virtualized ubiquitous database use environment.

In recent years, one of the most widely used authentication methods is a password-based authentication method. When using the password-based authentication method, users are required to create a secure password according to various requirements such that the password is long enough, or it does not contain dictionary words nor user's personal information. As a result, it is difficult for users to create memorable passwords. To resolve the problem, we propose a new password based authentication scheme using multiple passphrases and its arrangement information. By experiment, we show our proposed scheme attains good usability and memorability. We also show it is invulnerable against shoulder surfing, dictionary attack and brute-force attack.

A DoS (denial of service) attack is one of the most serious threats in the Internet. It is important to protect the resources and services from the DoS attack, but it is difficult to distinguish normal traffic and DoS attack traffic because the DoS attackers generally hide their true identities/origins. In this paper, we propose a technique to reduce the influence of the DoS attack without disturbing the demand of the regular users by allocating the information, when DoS attack occurs, to the filtering rules. This can be done by using DNS request replies.

Since digital contents are easy to copy or process, they tend to be unfairly used. A watermarking method is one of the effective countermeasures against such unfairness. In this paper, we propose watermarking schemes for Music XML, which represents a musical score by XML and able to display it and play back the corresponding music using a Web browser, and implement and evaluate the proposed schemes.

In multiagent system, it is important for agents to assess the situation prevailing in the system, especially to anticipate other agents' intentions. We argue in favor of cooperation among agents and propose a new method to utilize potential field as a tool for estimation of the environment. In our method, potential of environment gives agents some criteria to assess environmental situations from their own perspective. The potential of each object represents its influence on the environment and the environmental potential, i.e., summation of each object's potential, represents global situation of the environment. Agents' decision of their behavior will be done by refining the policy obtained from potential. We use a trash collecting problem as an example to show the effectiveness of our method. Furthermore, we show the efficiency of our method by some sets of experiments of the trash collecting problem. We also discuss the applicability of our method to hybrid systems, environments where available information has a margin of error, or environments where agents' range of vision are limited.