Automated fact-checking (AFC) becomes a crucial task because it helps verify the truthfulness of the claim in written or spoken language to decrease the effects of misleading information on social media. Among various forms of information, tabular information plays an important role in fact-checking, as it is an efficient and compact method for storing and representing complex scenario information in the real world. Although large-language models (LLMs) demonstrate robustness in natural language processing, they struggle with understanding and reasoning in tabular data. Therefore, we propose TabV4FC (Tabular Verbalization for Fact Checking) - a simple framework that integrates the pre-trained language model (TAPEX - Table Pre-training via Learning a Neural SQL Executor) for table-to-text generation with robust LLMs, including Qwen, Llama, and DeepSeek-R1, to leverage the reasoning ability of LLMs on natural language text for solving the fact-checking task on tabular evidence. We also evaluate our proposed method on three different tabular fact-checking datasets. Our empirical results show that the description text generated from the table-to-text model significantly boosts the accuracy of LLMs. In addition, our framework achieves competitive results compared to baselines and other state-of-the-art methods, such as TART and ProTrix, even in zero-shot settings. Finally, from the error analysis, we show that there is room for further improvement of our framework, including improving the ability of table understanding via extracting latent information from the table and advanced prompting methods for LLMs.

Stochastic versions of recursive integrated climate-economy assessment models are essential for studying and quantifying policy decisions under uncertainty. However, as the number of state variables and stochastic shocks increases, solving these models via deterministic grid-based dynamic programming (e.g., value-function iteration / projection on a discretized grid over continuous state variables, typically coupled with discretized shocks) becomes computationally infeasible, and simulation-based methods are needed. The least-squares Monte Carlo (LSMC) method has become popular for solving optimal stochastic control problems in quantitative finance. In this paper, we extend the application of the LSMC method to stochastic climate-economy models. We exemplify this approach using a stochastic version of the DICE model with five key uncertainty sources highlighted in the literature. To address the complexity and high dimensionality of these models, we incorporate deep neural network approximations in place of standard regression techniques within the LSMC framework. Our results demonstrate that the deep LSMC method can be used to efficiently derive optimal policies for climate-economy models in the presence of uncertainty.

Functional Isolation Forest (FIF) is a recent state-of-the-art Anomaly Detection (AD) algorithm designed for functional data. It relies on a tree partition procedure where an abnormality score is computed by projecting each curve observation on a drawn dictionary through a linear inner product. Such linear inner product and the dictionary are a priori choices that highly influence the algorithm's performances and might lead to unreliable results, particularly with complex datasets. This work addresses these challenges by introducing Signature Isolation Forest, a novel AD algorithm class leveraging the rough path theory's signature transform. Our objective is to remove the constraints imposed by FIF through the proposition of two algorithms which specifically target the linearity of the FIF inner product and the choice of the dictionary. We provide several numerical experiments, including a real-world applications benchmark showing the relevance of our methods.

We tackle the task by using a pretrained large language model (LLM) and in-context learning with template-based instructions to guide the LLM. To improve generation quality, we employ a two-step procedure: sampling and selection. For the sampling step, we randomly sample a subset of the provided training data for the context of LLM prompting. Next, for the selection step, we map the LLM generated outputs into a vector space and employ the Gaussian kernel density estimation to select the most likely output. The results show that the approach can achieve a certain degree of performance and there is still room for improvement.

Cough is a common symptom associated with respiratory diseases and its analysis plays a crucial role in monitoring the health conditions of affected persons. Traditional cough detection approaches largely fail to identify single cough boundaries when continuous coughs are present, consequently limiting their suitability for effective cough monitoring. In this research, we propose a novel deep learning system for the efficient detection and monitoring of cough events in audio recordings. Our detection pipeline consists of three key steps. First, we perform voice activity detection to eliminate audio silences and focus on relevant segments. Next, we employ a cough classification technique to identify the presence of cough within those audio segments. Finally, we implement cough event detection using a high-performance classification-regression fusion method. Our approach differs from the traditional audio event detection methods in several notable ways: (1) we incorporate a teacher-student framework for the training of our detection model, (2) instead of relying on specific audio features such as MFCC or Mel Spectrogram, our end-to-end system takes the raw audio signal directly as input and outputs the cough boundary timings, (3) the proposed method is general enough to be used for various other sound event monitoring tasks. Our detection model demonstrated strong performance and robustness on both the in-house and public datasets, by achieving cough event detection error-rate scores of 0.31 and 0.32, respectively, which is several times lower than other models. The comparative cough monitoring evaluation of our approach against systems such as the Leicester Cough Monitor and XGBoost demonstrates our method’s superiority by achieving the lowest average hourly symmetric mean absolute error (sMAPE) of 8.48%. The code is available at https://github.com/FengZongyao/Neural-Cough-Counter.

In the era of large language models, exploring the capability of a large language model is a trending research direction where one is prompt engineering aiming at drafting the best instructions to ask a large language model. We assume that a selected large language model has a certain knowledge and capability of a given task, and the way to ask is strongly related to the task’s performance. This is certainly useful when we don’t have sufficient resource for finetuning or aligning the large language model. Following that direction, we present our approach towards improving the effectiveness of prompting a large language model by finding an optimal ensemble of instructions by using the large language model’s self-generated instructions and labeled data. A case study on sentiment analysis is carried out in a preliminary experiment. The positive results of our case study shows that the approach is promising.

Accurate temperature forecasting is critical for various sectors, yet traditional methods struggle with complex atmospheric dynamics. Deep neural networks (DNNs), especially transformer-based DNNs, offer potential advantages, but face challenges with domain adaptation across different geographical regions. We evaluated the effectiveness of DNN-based domain adaptation for daily maximum temperature forecasting in experimental low-resource settings. We used an attention-based transformer deep learning architecture as the core forecasting framework and used kernel mean matching (KMM) for domain adaptation. Domain adaptation significantly improved forecasting accuracy in most experimental settings, thereby mitigating domain differences between source and target regions. Specifically, we observed that domain adaptation is more effective than exclusively training on a small amount of target-domain training data. This study reinforces the potential of using DNNs for temperature forecasting and underscores the benefits of domain adaptation using KMM. It also highlights the need for caution when using small amounts of target-domain data to avoid overfitting. Future research includes investigating strategies to minimize overfitting and to further probe the effect of various factors on model performance.

Social media platforms have emerged as a significant source of public opinion, offering a massive user-generated data in which user-opinions are valuable if obtainable. Large language models (LLMs) have been in the spotlight recently with suggestions on the emergent abilities to solve tasks that are not explicitly trained for. Thus, this study explores the potential of utilizing LLMs for opinion mining on social media data by asking LLMs difficult questions, instead of simply asking whether the text's sentiment polarity is either positive, negative, or neutral. This study compares the LLM response statistics and the corresponding public surveys related to COVID-19, including the intention to take vaccination and the stress check. The results indicate that it is promising, but also challenging, to utilize LLMs for the tasks.

We devised a data-driven framework for uncovering hidden control strategies used by an evolutionary system described by an evolutionary probability distribution. This innovative framework enables deciphering of the concealed mechanisms that contribute to the progression or mitigation of such situations as the spread of COVID-19. Novel algorithms are used to estimate the optimal control in tandem with the parameters for evolution in general dynamical systems, thereby extending the concept of model predictive control. This marks a significant departure from conventional control methods, which require knowledge of the system to manipulate its evolution and of the controller’s strategy or parameters. We use a generalized additive model, supplemented by extensive statistical testing, to identify a set of predictor covariates closely linked to the control. Using real-world COVID-19 data, we delineate the descriptive behaviors of the COVID-19 epidemics in five prefectures in Japan and nine countries. We compare these nine countries and group them on the basis of shared profiles, providing valuable insights into their pandemic responses. Our findings underscore the potential of our framework as a powerful tool for understanding and managing complex evolutionary processes.

As global warming progresses, it is increasingly important to monitor and analyse spatio-temporal patterns of heat waves and other extreme climate-related events that impact urban areas. In this work, we present a novel dynamic spatio-temporal model by combining a state space model (SSM) and a generalised hyperbolic distribution to flexibly describe a spatial–temporal profile of the tail behaviour, skewness and kurtosis of the local urban temperature distribution of the greater Tokyo metropolitan area. Such a model can be used to study local dynamics of temperature effects, specifically those that characterise extreme heat or cold. The focus of the application in this paper will be heat wave events in the greater Tokyo metropolitan area which is known to be prone to some of the most severe heat wave events that have one of the largest population exposures due to high density living in Tokyo city. The advantages the proposed model offers are as follows: it accommodates skewed and fat-tail distributions for temperature profiles; the model can be expressed as a location-scale linear Gaussian SSM which allows the development of an efficient Monte Carlo mixture Kalman Filter solution for the estimation. The proposed model is compared with the Gaussian SSM through application to maximum temperature data in the Tokyo metropolitan area between 1978–2016. The result suggests that the proposed model estimates the temperature distribution more accurately than the conventional linear Gaussian SSM and that the predictive variance of our method tends to be smaller than that obtained from the conventional spate time linear Gaussian SSM benchmark model.

Addressing biases in observed data is a major challenge in statistical and machine learning applications. This challenge also exists in recommendation systems, and various methods based on causal inference are being investigated. We investigate a collaborative filtering technique that robustly predicts ratings from biased observation. Utilizing the proportion of unbiased ratings in the different data sources, we extend collaborative filtering by adding a term for the proportion of ratings estimates from collaborative filtering to be closer to the proportion of ratings in unbiased data. Our aim is to obtain less biased estimates from observations that include bias. The proposed method is based on collaborative filtering with binomial matrix factorization, which treats observations and predictions as discrete variables. By treating the proportion of ratings in the unbiased case as a probability distribution, we introduce a constraint term that minimizes the KL divergence with the estimates by collaborative filtering. The binomial matrix factorization allows for direct calculation of the KL term due to the discrete assumption. The simple extension by adding the constraint term can be straightforwardly combined with various existing methods, such as inverse propensity weighting matrix factorization. Experimental results show that the standalone proposed method improves MSE from the conventional method, and the proposed method combined with inverse propensity score weighting also still improves slightly.

Introduction The worldwide COVID-19 pandemic, which began in December 2019 and has lasted for almost 3 years now, has undergone many changes and has changed public perceptions and attitudes. Various systems for predicting the progression of the pandemic have been developed to help assess the risk of COVID-19 spreading. In a case study in Japan, we attempt to determine whether the trend of emotions toward COVID-19 expressed on social media, specifically Twitter, can be used to enhance COVID-19 case prediction system performance. Methods We use emoji as a proxy to shallowly capture the trend in emotion expression on Twitter. Two aspects of emoji are studied: the surface trend in emoji usage by using the tweet count and the structural interaction of emoji by using an anomalous score. Results Our experimental results show that utilizing emoji improved system performance in the majority of evaluations.

A practical algorithm has been developed for closeness analysis of sequential data that combines closeness testing with algorithms based on the Markov chain tester. It was applied to reported sequential data for COVID-19 to analyze the evolution of COVID-19 during a certain time period (week, month, etc.).

The COVID-19 pandemic, which began in December 2019, progressed in a complicated manner and thus caused problems worldwide. Seeking clues to the reasons for the complicated progression is necessary but challenging in the fight against the pandemic. We sought clues by investigating the relationship between reactions on social media and the COVID-19 epidemic in Japan. Twitter was selected as the social media platform for study because it has a large user base in Japan and because it quickly propagates short topic-focused messages (“tweets”). Analysis using Japanese Twitter data suggested that reactions on social media and the progression of the COVID-19 epidemic may have a close relationship. Analysis of the data for the past waves of COVID-19 in Japan revealed that the relevant reactions on Twitter and COVID-19 progression are related repetitive phenomena. We propose using observations of the reaction trend represented by tweet counts and the trend of COVID-19 epidemic progression in Japan and a deep neural network model to capture the relationship between social reactions and COVID-19 progression and to predict the future trend of COVID-19 progression. This trend prediction would then be used to set up a susceptible-exposed-infected-recovered model for simulating potential future COVID-19 cases. Experiments to evaluate the potential of using tweets to support the prediction of how an epidemic will progress demonstrated the value of using epidemic-related social media data. Our findings provide insights into the relationship between user reactions on social media, particularly Twitter, and epidemic progression, which can be used to fight pandemics.

We propose a new framework that generalizes the parameters of neural network models to C*-algebra-valued ones. C*-algebra is a generalization of the space of complex numbers. A typical example is the space of continuous functions on a compact space. This generalization enables us to combine multiple models continuously and use tools for functions such as regression and integration. Consequently, we can learn features of data efficiently and adapt the models to problems continuously. We apply our framework to practical problems such as density estimation and few-shot learning and show that our framework enables us to learn features of data even with a limited number of samples. Our new framework highlights the potential possibility of applying the theory of C*-algebra to general neural network models.

Hybrid recommendation, which is based on collaborative filtering and supplemented with auxiliary content information, is being actively researched due to its ability to overcome the cold-start problem. Many proposed hybrid methods make recommendations using Gaussian distribution-based collaborative filtering even though they handle variables that tend to be non-Gaussian, such as the number of interactions. We present a method that uses a hybrid recommendation framework based on collaborative filtering that models the number of interactions as a Poisson-distributed and variational autoencoder-based content information generation process that shares latent variables with collaborative filtering. As a prior for the shared latent variables, we use a gamma distribution, which is a conjugate prior of a Poisson distribution. An implicit-derivative-based reparameterization trick enables the use of a gamma distribution in a variational autoencoder. The latent variables in the generative model are inferred using the stochastic gradient variational Bayes approach, taking the number of interactions corresponding to users and items and content information as input. In accordance with the inference, unobserved interactions between users and items are predicted for recommendation. The use of a neural-network-based generative model for content information enables the framework to handle various types of content information. Experimental results show that the proposed method utilizes content information effectively for predicting the number of interactions and that it should aid in overcoming the cold-start problem.

This paper consider the penalized least squares estimators with convex penalties or regularization norms. We provide sparsity oracle inequalities for the prediction error for a general convex penalty and for the particular cases of Lasso and Group Lasso estimators in a regression setting. The main contribution is that our oracle inequalities are established for the more general case where the observations noise is issued from probability measures that satisfy a weak spectral gap（or Poincaré） inequality instead of Gaussian distributions. We illustrate our results on a heavy tailed example and a sub Gaussian one; we especially give the explicit bounds of the oracle inequalities for these two special examples.

We describe a multi-source and unsupervised domain adaptation method using Sinkhorn barycenters, which, given the labeled data in multi-source domains and unlabeled data in a target domain, uses the optimal transport Sinkhorn distance to measure gaps between data distributions in the source and target domains. For end-to-end classification learning, the feature extractor and classifier are simultaneously estimated on the basis of two criteria: the minimization of the Sinkhorn distance for the source and target domains and the minimization of the classification loss for the source domains. The first criterion is based on the assumptions that domain-invariant features would be captured in a latent feature space obtained by minimizing the Sinkhorn distance among all domains and that the space would be close to the Sinkhorn barycenter. Experiments on image classification using the Digit-Five dataset, which is comprised of digit datasets from five different domains, demonstrated that our method outperforms other state-of-the-art methods.

The classical DICE model is a widely accepted integrated assessment model for the joint modeling of economic and climate systems, where all model state variables evolve over time deterministically. We reformulate and solve the DICE model as an optimal control dynamic programming problem with six state variables (related to the carbon concentration, temperature, and economic capital) evolving over time deterministically and affected by two controls (carbon emission mitigation rate and consumption). We then extend the model by adding a discrete stochastic shock variable to model the economy in the stressed and normal regimes as a jump process caused by events such as the COVID-19 pandemic. These shocks reduce the world gross output leading to a reduction in both the world net output and carbon emission. The extended model is solved under several scenarios as an optimal stochastic control problem, assuming that the shock events occur randomly on average once every 100 years and last for 5 years. The results show that, if the world gross output recovers in full after each event, the impact of the COVID-19 events on the temperature and carbon concentration will be immaterial even in the case of a conservative 10\% drop in the annual gross output over a 5-year period. The impact becomes noticeable, although still extremely small (long-term temperature drops by $0.1^\circ \mathrm{C}$), in a presence of persistent shocks of a 5\% output drop propagating to the subsequent time periods through the recursively reduced productivity. If the deterministic DICE model policy is applied in a presence of stochastic shocks (i.e. when this policy is suboptimal), then the drop in temperature is larger (approximately $0.25^\circ \mathrm{C}$), that is, the lower economic activities owing to shocks imply that more ambitious mitigation targets are now feasible at lower costs.

A class of models for non-Gaussian spatial random fields is explored for spatial field reconstruction in environmental and sensor network monitoring. The family of models explored utilises a class of transformation functions known as Tukey g-and-h transformations to create a family of warped spatial Gaussian process models which can support various desirable features such as flexible marginal distributions, which can be skewed, leptokurtic and/or heavy-tailed. The resulting model is widely applicable in a range of spatial field reconstruction applications. To utilise the model in applications in practice, it is important to carefully characterise the statistical properties of the Tukey g-and-h random fields. In this work, we study both the properties of the resulting warped Gaussian processes as well as using the characterising statistical properties of the warped processes to obtain flexible spatial field reconstructions. In this regard we derive five different estimators for various important quantities often considered in spatial field reconstruction problems. These include the multi-point Minimum Mean Squared Error (MMSE) estimators, the multi-point Maximum A-Posteriori (MAP) estimators, an efficient class of multi-point linear estimators based on the Spatial-Best Linear Unbiased (S-BLUE) estimators, and two multi-point threshold exceedance based estimators, namely the Spatial Regional and Level Exceedance estimators. Simulation results and real data examples show the benefits of using the Tukey g-and-h transformation as opposed to standard Gaussian spatial random fields in a real data application for environmental monitoring.