This article presents our generative model for rhythm action games together with applications in business operation. Rhythm action games are video games in which the player is challenged to issue commands at the right timings during a music session. The timings are rendered in the chart, which consists of visual symbols, called notes, flying through the screen. We introduce our deep generative model, GenéLive!, which outperforms the state-of-the-art model by taking into account musical structures through beats and temporal scales. Thanks to its favorable performance, GenéLive! was put into operation at KLab Inc., a Japan-based video game developer, and reduced the business cost of chart generation by as much as half. The application target included the phenomenal "Love Live!", which has more than 10 million users across Asia and beyond, and is one of the few rhythm action franchises that has led the online era of the genre. In this article, we evaluate the generative performance of GenéLive! using production datasets at KLab as well as open datasets for reproducibility, while the model continues to operate in their business. Our code and the model, tuned and trained using a supercomputer, are publicly available.

In the operation of information technology (IT) services, operators monitor the equipment-issued alarms, to locate the cause of a failure and take action. Alarms generate simultaneously from multiple devices with physical/logical connections. Therefore, if the time and location of the alarms are close to each other, it can be judged that the alarms are likely to be caused by the same event. In this paper, we propose a method that takes a novel approach by correlating alarms considering event units using a Bayesian network based on alarm generation time, generation place, and alarm type. The topology information becomes a critical decision element when doing the alarm correlation. However, errors may occur when topology information updates manually during failures or construction. Therefore, we show that event-by-event correlation with 100% accuracy is possible even if the topology information is 25% wrong by taking into location information other than topology information.

In the course of network operations, telecommunications carriers must have accurate topology information of the network related to the failure to identify quickly causes of service failures and determine their impacts. However, telecommunications carrier networks are divided into multiple layers according to their roles, and each layer has a different management system, making it difficult to detect the topology information between different layers. Therefore, there is a need for a technology that can assure accurate topology information of the multiple layers network. We propose a new topology discovery method based on the consistency of the traffic of mutually connected interfaces. We verify the effectiveness of the proposed method using traffic data from network equipment in a commercial network and show that the proposed method is capable of detecting topologies with high accuracy.

Deploying artificial intelligence (AI) to network operations have long been an issue. AI has been expected to help automating network operations, especially in those requiring human decisions, such as handling failures which involve complex decisions. Since current AIs do not have enough parameters for individual failures cases, their accuracy is not enough to fully rely on their decisions. Thus, handing unusual fault cases are still dominated by skilled operators. This paper proposes an extended Business Process Model and Notation (BPMN) which uses Bayesian networks to represent operator's decisions, to connect AIs and workflow engines (WFE) which automates handing atypical failures.

As for the service assurance operation of telecommunications carriers, accurate information about a network topology which indicates the connection relationships between pieces of network equipment is necessary. However, the network of a telecommunications carrier has several hundreds of thousands of equipment, and its topology is frequently supplemented and modified due to daily construction work and troubleshooting. It is a therefore a problem when incorrect topology information is mixed into the overall topology information. In this paper, we propose a method that can discover the topology between equipment by using alarm information issued by those equipment during construction work or when a failure occurs. The proposed method was evaluated using alarm information generated under in certain commercial configurations (sections containing specific routers), and it was confirmed that the current topology could be discovered with 100% accuracy, even though only 1.7% of the total topology was evaluated in one day.

For the service assurance operation of telecommunications carriers, accurate information about a network topology that indicates the connection relationships between pieces of network equipment is necessary. However, the network of a telecommunications carrier has several hundreds of thousands of pieces of equipment. Furthermore, its topology is frequently supplemented and modified due to daily construction work and troubleshooting. As a result, this causes incorrect topology information to be mixed into the overall topology information. In this paper, we propose a method that can discover the topology between equipment by using alarm information issued by those equipment during construction work or when a failure occurs. The proposed method was evaluated using alarm information generated on a commercial network. The experimental results show that the proposed method discovers topology with higher accuracy than classical topology discovery approaches.

Telecoms need to change their operations more flexibly in accordance with diverse requirements for service quality. They have long operated networks with high availability (24/7) to support the emergency services, however their networks now provide ever more diverse and newer services. Telecoms can reduce operating expenses and even lower service prices if they make their operations more flexible on the basis of services' requirements instead of simply pursuing high availability. Thus, we researched the concept of Service Level Agreement (SLA) Driven Operation, which enables various decisions about failure recoveries to be automated on the basis of the service quality requirements.In this paper, to implement SLA Driven Operation to telecom operation, we focus on automating decisions about "how," "who," and "when" to take action for failure recovery. Moreover, we propose a method to automatically decide the optimal start time by assessing the schedule in terms of response cost, SLA violation cost, and lost profits. Then we estimate the effects and the risks that occur when the response start time is decided automatically. As a result, we demonstrate that applying the proposed method can be expected to sufficiently reduce cost with a very low risk of availability decline.

Network Function Virtualization (NFV)/ Software Defined Network (SDN) technologies are introduced by content providers in advance, furthermore telecom carriers are also starting examining. However, it is necessary to examine the operation policy considering the difference characteristics telecom carriers and content providers. Specifically, it is the historical background which emphasizes the geographical dispersion and high reliability of the facilities. This paper clarifies that the conversion to the operation style which focus on the quality required for each service is necessary, after analyzing the present state of telecom network operation. Also, this paper proposes a concept “Service Level Agreement (SLA) Driven Operation” aiming to make decisions automatically for assurance work based on service quality (SLA) to be satisfied by operation.

We present a platform on all-programmable SoC for micro autonomous robots for probing, exploring, rescuing, etc. Contrast to challenges for auto drive cars having relatively rich power supply and high-performance computing platform, our challenge is to develop technologies for autonomous robots with tight restrictions on size, weight, and energy consumption. We utilize all-programmable SoCs for the purpose and develop a system including camera interface, image processing, recognition, action planning, and motor control. The key techniques are to optimize the dataflow between software (and main memory) and hardware for efficiency and to adopt a standard stream interface in hardware modules for productivity. The system can be utilized as a common platform for micro autonomous robots. The system is implemented as a robot car named ZybotR2-Z2 and achieves 3 to 37 frame/sec image recognition and car control with single Zynq-7020 device.