A humanoid robot distributes leaflets of GMO Internet Group at the company's booth during the Japan Drone Expo 2026 in Chiba, in June | REUTERS GMO Internet has made an urgent request for the Japanese government to implement "active cyber defense" operations following an unauthorized access incident, sources said Saturday. GMO Internet has revealed that there was unauthorized access to its servers for automated FX currency margin trading. A total of 26,739 such servers may have been affected. Since some users have reported losses from unauthorized transactions on their accounts following the incident, the company said Saturday that it will compensate users for these losses. The company also plans to make a consolation payment of ¥5,000 to all users, regardless of whether they have suffered financial damage. On Oct. 1, the government began operations under the active cyber defense system, which involves the Self-Defense Forces and police pre-emptively breaking into an attacker's server to neutralize it and prevent damage.

Government and company websites across the world went offline this week after the disclosure of critical vulnerabilities in a widely used programming language prompted millions of hacking attempts. Japan has said it will set up a team to simulate cyberattacks using AI to identify weaknesses in technology systems as organizations scramble to curtail risks from flaws in the PHP programming language that underpins much of the internet. Government and company websites across the world went offline this week after the disclosure of critical vulnerabilities in the widely used programming language prompted millions of hacking attempts, forcing website operators to rush to protect or temporarily shut down their systems. Highlighting the extent of the disruption, Japan's third-largest mobile carrier, SoftBank, said on Friday it was suspending its emergency alert system for mobile users. In a time of both misinformation and too much information, quality journalism is more crucial than ever. By subscribing, you can help us get the story right.

SoftBank Group's market value halved from its peak in June. SoftBank Group's market value halved from its peak in June, as higher borrowing costs fuel investor concerns over the sustainability of its massive investments in artificial intelligence. Shares of the Tokyo-based company fell 6.9% on Thursday, taking their drop from an all-time high struck on June 2 to 50%, with earnings from U.S. megacaps failing to ease the angst. After briefly overtaking Toyota as Japan's biggest company four months ago, its market capitalization is now about ¥24.7 trillion ($157 billion), smaller than that of Hitachi and Recruit. Global concerns over AI spending had accelerated after U.S.-based cloud-computing provider CoreWeave's bonds fell to levels associated with distress, exposing financing risks across the industry. "SoftBank's vulnerability stands out as investors increasingly scrutinize the financial health of AI-related companies, given the group's highly leveraged financial structure and growing borrowing costs," said Kazuyuki Muramatsu, head of investment management at Nagomi Capital.

The industry ministry said Friday that it has selected 18 regions as green transformation strategy areas. The ministry will support the regions by easing regulations on land use and providing subsidies, aiming to attract companies and investment from Japan and abroad as part of efforts to strengthen the industrial competitiveness of regional economies. Demand for data centers is growing with the spread of generative artificial intelligence, while operating them requires enormous amounts of electricity. The ministry aims to attract large-scale data centers requiring around 1 gigawatt of power to regions rich in decarbonized energy sources to revitalize domestic investment and local industries. The ministry designated nine of the 18 regions as areas for concentrating data centers.

U.S. defense firm Anduril in talks for Nissan plant to build drones in Japan, sources say U.S. defense firm Anduril Industries is in talks to acquire the plant to build military drones in Japan, sources say. U.S. defense firm Anduril Industries is in talks to acquire Nissan Motor's Oppama assembly plant near Tokyo as the maker of autonomous weapons looks to build military drones in Japan, according to three sources familiar with the matter. While they say no decision has been made, any deal could transform one of Japan's first large-scale postwar car factories, long a symbol of its industrial revival, into an arms-making hub. The talks over Oppama, which are being reported for the first time, come as Prime Minister Sanae Takaichi's government seeks to expand defense manufacturing amid growing concern that a Taiwan Strait crisis could draw in Japan and run down weapons stocks. In a time of both misinformation and too much information, quality journalism is more crucial than ever. By subscribing, you can help us get the story right.

Terra Drone's Terra A1 interceptor drone has entered active combat use in Ukraine after being deployed to a military unit tasked with countering Russian uncrewed aerial systems. Japanese drone company Terra Drone said its Terra A1 interceptor -- developed with its Ukrainian partner Amazing Drones -- has moved from the lab to the front lines, entering active combat use in Ukraine against Russian-made Shahed drones. "Deployment for defense purposes has already begun with a military unit, and evaluation and feedback collection under actual operating conditions are currently under way," the Tokyo-based firm, which recently made a strategic investment in the Ukrainian startup, said in a recent statement. Terra Drone explained that this initial "real-world operational deployment" -- carried out via its local partner -- will follow a phased rollout, where new equipment is first issued to a single unit and then expanded to further deployments depending on evaluations from the field. In a time of both misinformation and too much information, quality journalism is more crucial than ever.

Gaussian process ($GP$) regression is a widely used non-parametric modeling tool, but its cubic complexity in the training size limits its use on massive data sets. A practical remedy is to predict using only the nearest neighbours of each test point, as in Nearest Neighbour Gaussian Process ($NNGP$) regression for geospatial problems and the related scalable $GPnn$ method for more general machine-learning applications. Despite their strong empirical performance, the large-$n$ theory of $NNGP/GPnn$ remains incomplete. We develop a theoretical framework for $NNGP$ and $GPnn$ regression. Under mild regularity assumptions, we derive almost sure pointwise limits for three key predictive criteria: mean squared error ($MSE$), calibration coefficient ($CAL$), and negative log-likelihood ($NLL$). We then study the $L_2$-risk, prove universal consistency, and show that the risk attains Stone's minimax rate $n^{-2α/(2p+d)}$, where $α$ and $p$ capture regularity of the regression problem. We also prove uniform convergence of $MSE$ over compact hyper-parameter sets and show that its derivatives with respect to lengthscale, kernel scale, and noise variance vanish asymptotically, with explicit rates. This explains the observed robustness of $GPnn$ to hyper-parameter tuning. These results provide a rigorous statistical foundation for $NNGP/GPnn$ as a highly scalable and principled alternative to full $GP$ models.

Damped sinusoidal oscillations are widely observed in many physical systems, and their analysis provides access to underlying physical properties. However, parameter estimation becomes difficult when the signal decays rapidly, multiple components are superposed, and observational noise is present. In this study, we develop an autoencoder-based method that uses the latent space to estimate the frequency, phase, decay time, and amplitude of each component in noisy multi-component damped sinusoidal signals. We investigate multi-component cases under Gaussian-distribution training and further examine the effect of the training-data distribution through comparisons between Gaussian and uniform training. The performance is evaluated through waveform reconstruction and parameter-estimation accuracy. We find that the proposed method can estimate the parameters with high accuracy even in challenging setups, such as those involving a subdominant component or nearly opposite-phase components, while remaining reasonably robust when the training distribution is less informative. This demonstrates its potential as a tool for analyzing short-duration, noisy signals.

We consider the problem of conditional independence (CI) testing and adopt a kernel-based approach. Kernel-based CI tests embed variables in reproducing kernel Hilbert spaces, regress their embeddings on the conditioning variables, and test the resulting residuals for marginal independence. This approach yields tests that are sensitive to a broad range of conditional dependencies. Existing methods, however, rely heavily on kernel ridge regression, which is computationally expensive when properly tuned and yields poorly calibrated tests when left untuned, which limits their practical usefulness. We propose the Generalised Kernel Covariance Measure (GKCM), a regression-model-agnostic kernel-based CI test that accommodates a broad class of regression estimators. Building on the Generalised Hilbertian Covariance Measure framework (Lundborg et al., 2022), we characterise conditions under which GKCM satisfies uniform asymptotic level guarantees. In simulations, GKCM paired with tree-based regression models frequently outperforms state-of-the-art CI tests across a diverse range of data-generating processes, achieving better type I error control and competitive or superior power.

We present mlr3mbo, a comprehensive and modular toolbox for Bayesian optimization in R. mlr3mbo supports single- and multi-objective optimization, multi-point proposals, batch and asynchronous parallelization, input and output transformations, and robust error handling. While it can be used for many standard Bayesian optimization variants in applied settings, researchers can also construct custom BO algorithms from its flexible building blocks. In addition to an introduction to the software, its design principles, and its building blocks, the paper presents two extensive empirical evaluations of the software on the surrogate-based benchmark suite YAHPO Gym. To identify robust default configurations for both numeric and mixed-hierarchical optimization regimes, and to gain further insights into the respective impacts of individual settings, we run a coordinate descent search over the mlr3mbo configuration space and analyze its results. Furthermore, we demonstrate that mlr3mbo achieves state-of-the-art performance by benchmarking it against a wide range of optimizers, including HEBO, SMAC3, Ax, and Optuna.