Particle Swarm Optimization (PSO) is one of the most effective optimization methods for the black-box optimization problem. PSO involves a large number of particles sharing information with each other to search for the optimal solution. The method in which a large number of search individuals cooperate to search for the optimal solution is called swarm intelligence optimization. In the group intelligence optimization, the balance between exploration and exploitation is important. However, in PSO, it is unclear to what extent each parameter affects exploration and exploitation. Therefore, we proposed a deterministic PSO without probabilistic elements and analyzed the dynamics of PSO using dynamical systems theory. Each particle in deterministic PSO has its motion determined by its eigenvalue. In order to make this motion clearer, a canonical deterministic PSO on a regularized phase space was proposed. The results of these analyses clarified what is attributed to the parameters for exploration and exploitation, i.e., global and local search capabilities. Based on this fact, we proposed a nonlinear map optimization (NMO) with improved local search capability. In this paper, we present the background of our proposal and consider the solution-search capability of nonlinear map optimization.

In order to realize a barrier free society, it is important to realize sign language recognition on smartphones. Several methods have been proposed for fingersplling recognition, which is the basis of sign language recognition. In this paper, we propose a simple still fingerspelling recognition system. The system apply the two dimensional hand point information obtained by MediaPipe. The obtained information is identified by a linear SVM. The proposed system requires only a smaller amount of input data than the previous ones, and the processing is realized by SVM which requires a small amount of computation. Although the proposed system requires only a small amount of computation, it achieves extremely recognition accuracy for still fingerspellings.

To find the shortest tour of a transporting vehicle in a bike sharing system (BSS), a bike sharing system routing problem (BSSRP) has been proposed. In the BSSRP, a single vehicle restores the number of bicycles in stations. However, in a real system, the restoration of the number of the bicycles in each station is carried out by multiple vehicles. To decide the shortest tour of the multiple vehicles, we proposed a mathematical optimization model called the multiple-vehicles bike sharing system routing problem (mBSSRP). We present a construction method and local search methods to solve the large-size mBSSRP. The result of numerical experiments shows that the proposed method can find good solutions in a short time.

In this study, using k-means and UMAP, we verified that the sentence vectors generated by Sentence-BERT as distributed representations of sentences capture the meaning of sentences well. To this end, we visualized the sentence vectors by generating images matching the meaning of the sentence from the sentence vectors generated by Sentence-BERT. The results confirm that although there were differences in the information represented by each dimension as distributional features of the sentence vector, this information overlapped substantially.

An encoder-decoder model consists of an encoder that encodes the input into a low-dimensional latent variable and a decoder that decodes the obtained latent variable to the same dimension as the input. The encoder-decoder model performs representation learning to automatically extract features of the data, but the model is a black box and it is not clear what features are extracted. We focused on whether including a skip connection between the encoder and decoder increased accuracy. It is generally believed that this skip connection plays a role in conveying high-resolution information. However, its actual role remains unclear. In this study, we focused on this concatenation. We experimentally clarified the role of the latent variables conveyed by this concatenation when the images given to the input and output were the same or different during training.

Reservoir computing is a machine learning model that is widely used for time series tasks due to its advantages of low cost and fast learning. However, conventional reservoir computing often do not achieve the memory capacity and nonlinearity required for the task. To solve this problem, we proposed hysteresis reservoir computing which conventional reservoir neurons are replaced by hysteresis neurons. The model generates various output sequences by changing their parameters. In addition, it has the potential to memorize time series because it presents complex periodic solutions. In this paper, we confirm the dynamics generated by changing the hysteresis reservoir computing parameters. The experimental results show that changing the parameters improves the learning ability and can represent specific series of data. This indicates an important dynamics in terms of memory capacity and the ability to represent.