This paper describes an interpretation support system for classification patterns extracted from deep learning with texts using HMM, and verified its effectiveness. It is well known that classification patterns by deep learning models are often difficult to interpret the reasons derived. In the proposed system, the content of deep learning results is extracted using HMMs, and classification patterns are provided for the system users to interpret the learned features. Then, the system displays learned network structures so that anyone can easily understand learning results. In verification experiments to confirm the effectiveness of the system, based on the learning result of deep learning classifying sentences, in the experiment, the subjects were divided into two groups. One group used the proposed system. The other group used the system that displays words with high TFIDF values. The both groups were instructed to give meanings of classification patterns peculiar to each output. The results show that the subjects who used the proposed system were able to understand the meanings of the classification patterns of deep learning with texts more deeply than those who used the comparison system.

Particle swarm optimization (PSO) has been extended and shown to perform well on discrete optimization. Some of the extended PSOs handle continuous parameters of probability distributions over variable values of candidate solutions instead of directly handling discrete variables. These distribution-based discrete PSOs (DDPSOs) sample a variable value from a probability distribution for each variable to generate a candidate solution. This procedure can be considered as a kind of random local search centering on an intended solution, which has the highest generation probability. It has drawbacks the step size increases proportionally and the probability of producing solutions close to the intended solution decreases exponentially in high-dimensional problems. This paper proposes a novel sampling method (NS) to control the step size for DDPSO and determined the step sizes according to Lévy distribution in a similar way to Lévy flight. NS is applied to three representative DDPSOs, and they were compared in integer and categorical versions of function optimization problems and NK landscape problems. The results show NS improves the efficiency and robustness of all the DDPSO variants. In addition, NS was tested on feature selection experiments and was achieved superior results compared to some evolutionary algorithms designed for feature selection.

The purpose of this paper is to observe whether if robot expressions are affecting subjects’ stress and physiological status during the exam. This project proposed two kinds of robot expressions in this experiment, robots express themselves differently when subjects answer correctly or incorrectly. Subject heart rate and brainwaves are monitored by sensors during the experiment. The experiment discovered when subjects face hard questions and answered incorrectly, there are different fluctuations in subject stress which in turn is shown by robotic expressions. This proves that robot expressions influence subject stress under different circumstances which in turn can be applied as a feedback mechanism for influencing stress with students, in the future may substitute humans as teachers.

To investigate how gestures a robot expressed affect people’s acceptance of the encouragement or requests from the robot, experiments where a robot showed different expressions have been conducted. Acceptance tendencies were analyzed from questionnaires and biosensors, including heartbeat and brain waves, in this paper. With the increasing needs of robots in daily life, the situations that robots induce humans to take actions and perform tasks collaboratively must be considered in studies of human–robot interaction. A multi-modal expression that includes utterance content and nonverbal information is crucial to promote the acceptance of inducement and request. Therefore, this study focuses on the gesture, an important nonverbal expression, and has investigated the effects on humans. The experiment compared the three conditions; M-condition (gestures that match the emotion of the speech content), R-condition (gestures that are opposite to the emotion of the speech content), and N-condition (no gesture). The result showed that the people in M-condition and R-conditions were more likely than the N-condition to accept robot requests and feel relaxing after the robot asking.