A classification model must deal with incremental changes of classification tasks in practical use, which is the aim of continual learning. In particular, many existing studies focus on class incremental learning, which requires a model to learn new classes while maintaining the ability to recognize the classes the model has already learned. However, a neural network model completely forgets the ability to recognize all the learned classes when learning new classes. This phenomenon is called catastrophic forgetting and known as the main problem in continual learning. To overcome catastrophic forgetting, various continual learning methods are proposed. However, there is little understanding of their mechanism for mitigating forgetting. In this paper, we propose a novel analytical method based on classification complexity in representation spaces to reveal the properties of class incremental learning methods. To evaluate classification complexity, we design new metrics based on Local Set Cardinality average, which is the existing complexity metric. Our analytical method reveals the properties of class incremental learning methods through the evaluation of various classification complexities such as the complexity of the classification among learned classes. To verify the usefulness of our analytical method, we analyze three typical class incremental learning techniques on the two-task and five-task class incremental learning setups.

This paper considers optimal scheduling for a vertical transport machine. We propose a scheduling method with MaxSAT which is an optimal version of Boolean satisfiability problem (SAT). The movement of the luggage in the machine is described as a set of Boolean formulas, and those costs are represented by a set of weighted Boolean formulas. The MaxSAT solver finds out an optimal scheduling as a model satisfying the formulas with the minimum cost. Experimental results show that the proposed method can solve the practical problems in reasonable time. We succeeded to improve transport capacity by constructing the optimal table with MaxSAT.

Intelligent Transport Systems (ITS) play an important role in achieving smooth and safe travel on urban road networks. ITS provide software-based traffic management based on traffic prediction, so they can’t manage traffic properly without accurate traffic prediction. Recently, spatio-temporal graph neural networks (STGNNs) have achieved significant improvements in traffic prediction by taking into account spatial and temporal dependencies in traffic data. However, although the length of congestion queues is one of the most important statistics in ITS because it can be used for proactive signal control and information providing, it has not been a prediction target in existing studies. In addition, the relationships between multimodal traffic variables have been ignored. Moreover, due to the significant impact of ITS on the real world, ITS tend to prefer explainable methods over black-box methods. In this study, we propose a Queueing-theory-based Neural Network (QTNN) for queue length prediction. QTNN combines data-driven STGNN methods with queueing-theory-based traffic engineering domain knowledge to make predictions accurate and explainable. Our experiments on queue length prediction using real-world data showed that QTNN outperformed the baseline methods, including state-of-the-art STGNNs, by 12.6% and 9.9% in RMSE and MAE, respectively.

Compositionality is one of the most important properties of human language. In the study of emergent communication, methods to induce compositionality in emergent language learned by neural network agents have been actively studied. Humans have individually articulated prior knowledge about the environment, and it is said to be important for compositional human language by communicating based on this prior knowledge. We implement this idea as a Reassembly game, a reconstruction task-based Lewis signaling game in which communicating agents have different pre-trained VAE. Experimental results show that the emergent language learned in the Reassembly game showed high compositionality on several metrics than the emergent language learned in the other reconstruction task-based Lewis signaling game settings. In addition, compared to other factors inducing compositional emergent language such as the length of message and the number of vocabulary, the emergent language learned in the Reassembly game achieve a high level of both communication success rate and compositionality.

This research aims to develop a knowledge sharing system for supporting cancer survivors who face life reconstruction challenges caused by Chemotherapy-Induced Peripheral Neuropathy (CIPN). This paper presents the intricate details of knowledge modeling and the system’s development. The life reconstruction knowledge has been described in four knowledge modules, each employing a uniform decomposition structure. The relationships between knowledge modules are described through two types of linking nodes. The system developed based on the knowledge model provides a searching function by specifying problems in daily life from multifaceted perspectives. The search results provide diverse solutions to the specified problems, which had remained currently personal.

Music classification is a fundamental task in the field of Music Information Retrieval. This paper focuses on composer classification, a specific task within music classification. Compressive techniques are commonly employed in such music classification tasks. In this study, we propose a method to apply the Computing Information Quantity using Maximum Probability partitioning to music classification. To evaluate the effectiveness of our proposed method, we perform composer classification, specifically distinguishing between Haydn and Mozart, who are well-known for their stylistic similarities. The experimental results demonstrate that our proposed approach outperforms traditional compression-based classification methods. Furthermore, we compare our method with non-compressive techniques, discussing the significance of feature extraction methods. Our proposed method is a parameter-free classification approach that does not require domain-specific musical expertise or feature extraction based on such expertise.