Eisuke Kita’s research while affiliated with Nagoya University and other places

In image-posting social networking services, such as Instagram, recommendation of appropriate hashtags for posts is vital. In the existing methods, a hashtag is searched using the names of object labels included in images added to posts as hashtags, and a relevance prediction model is applied to hashtags that appear most frequently among those attached to posts obtained from the search. Hashtags that are considered highly relevant to the post are then recommended to the user. However, it is difficult to recommend adequate hashtags relevant to a post containing a label that refers to different objects, such as “mouse,” which can refer to a “computer input device” and an “animal.” In this study, we developed algorithms (Algorithms 1 and 2) that employ additional labels related to object labels in posts to solve this problem. As additional labels, Algorithm 1 uses the other labels in the same object category in the Microsoft Common Objects in Context (COCO) image database, and Algorithm 2 uses words translated into six other languages. We also developed Algorithm 3, which is a hybrid of Algorithms 1 and 2. Based on user questionnaires, the hashtags suggested by Algorithms 1 and 2 are highly relevant to the posts: compared to an existing algorithm, the relevance of the hashtags improved by 18% and 64%, respectively.

XGBoost is well-known as an open-source software library that provides a regularizing gradient boosting framework. Although it is widely used in the machine learning field, its performance depends on the determination of hyper-parameters. This study focuses on the optimization algorithm for hyper-parameters of XGBoost by using Stochastic Schemata Exploiter (SSE). SSE, which is one of Evolutionary Algorithms, is successfully applied to combinatorial optimization problems. SSE is applied for optimizing hyper-parameters of XGBoost in this study. The original SSE algorithm is modified for hyper-parameter optimization. When comparing SSE with a simple Genetic Algorithm, there are two interesting features: quick convergence and a small number of control parameters. The proposed algorithm is compared with other hyper-parameter optimization algorithms such as Gradient Boosted Regression Trees (GBRT), Tree-structured Parzen Estimator (TPE), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), and Random Search in order to confirm its validity. The numerical results show that SSE has a good convergence property, even with fewer control parameters than other methods.

Creating a control program for robots needs explicit definition of each action to handle inputs from sensors and produce outputs to actuator. The aim of this project is to use Grammatical Evolution (GE) to automate the process of producing multiple actions as strategy. This algorithm can separate searching space and generated program, thus eliminating human criteria bias when creating a robot controller while able to find more optimized and complex actions strategy. The evaluation is done through partially observed maze problem which objective to find goal without defining starting orientation and discoverable path along the way.

The Stochastic Schemata Exploiter (SSE), one of the Evolutionary Algorithms, is designed to find the optimal solution of a function. SSE extracts common schemata from individual sets with high fitness and generates individuals from the common schemata. For hyper-parameter optimization, the initialization method, the schema extraction method, and the new individual generation method, which are characteristic processes in SSE, are extended. In this paper, an SSE-based multi-objective optimization for AutoML is proposed. AutoML gives good results in terms of model accuracy. However, if only model accuracy is considered, the model may be too complex. Such complex models cannot always be allowed because of the long computation time. The proposed method maximizes the stacking model accuracy and minimizes the model complexity simultaneously. When compared with existing methods, SSE has interesting features such as fewer control parameters and faster convergence properties. The visualization method makes the optimization process transparent and helps users understand the process.

According to a survey by the Ministry of Land, Infrastructure, Transport and Tourism, as much as 40% of the annual travel time is wasted in traffic jams. In addition, some studies have shown that the gas emitted by idling causes air pollution, and this has become a major social problem. One of the efforts to improve these problems caused by traffic congestion is the research on platooning. Most of the conventional research assumes a straight road as the experimental condition, and there is no discussion on the behavior at intersections scattered in the general roadway. In this study, a velocity control model is studied to avoid collisions with pedestrians when platoon vehicles make a turn at the intersection. A vehicle following model is defined according to Helly model which controls the vehicle acceleration according to relative speed and distance with the vehicle in front. The validity of the model is discussed from the computer simulation. Simulations were performed in some cases. The model could control the vehicles individually without collisions with pedestrians. However, these experiments were conducted only in an ideal environment by simulation. Therefore, in the future, the effectiveness of the model should be improved in actual traffic flow by conducting experiments using real vehicles.