UAVs (Unmanned Aerial Vehicles) are promising tools for disaster monitoring, by obtaining valuable information of important PoIs (Points of Interest) with onboard cameras. Since people trapped at some PoIs are more likely in danger than people in other PoIs, different PoIs have different monitoring priorities, so that the PoIs with high monitoring priorities should be visited more often than those with low priorities. Unlike existing studies that assumed a UAV is required to fly to the location of a PoI to monitor the PoI, we observe that the a UAV can monitor a PoI as long as it hovers at any location around the PoI (e.g., 200 m away horizontally), thereby reducing the flying time of the UAV. In this paper, we first study a problem of finding a sequence of monitoring tours for an energy-constrained UAV to monitor PoIs in a disaster area for a monitoring period T (e.g., 72 hours) persistently, such that the maximum weighted monitoring interval of PoIs is minimized, where the weight associated with a PoI is its monitoring priority, and the monitoring interval of a PoI is the longest time between its two consecutive visits in period T. We then propose a novel approximation algorithm for the problem. We finally evaluate the algorithm performance based on both a real testbed and the simulation. The experimental results show that the maximum weighted monitoring interval by the proposed algorithm is up to 30% shorter than those by existing algorithms.

The deep operator networks (DON), a class of neural operators that learn mappings between function spaces, have recently emerged as surrogate models for parametric partial differential equations (PDEs). However, their full potential for accurately approximating general black-box PDEs remains underexplored due to challenges in training stability and performance, primarily arising from difficulties in learning mappings between low-dimensional inputs and high-dimensional outputs. Furthermore, inadequate encoding of input functions and query positions limits the generalization ability of DONs. To address these challenges, we propose the Dynamical Coupled Operator (DCO), which incorporates temporal dynamics to learn coupled functions, reducing information loss and improving training robustness. Additionally, we introduce an adaptive spectral input function encoder based on empirical mode decomposition to enhance input function representation, as well as a hybrid location encoder to improve query location encoding. We provide theoretical guarantees on the universal expressiveness of DCO, ensuring its applicability to a wide range of PDE problems. Extensive experiments on real-world, high-dimensional PDE datasets demonstrate that DCO significantly outperforms DONs.

Decoding semantic meanings from brain activity has attracted increasing attention. Neurolinguists have found that semantic perception is open to multisensory stimulation, as word meanings can be delivered by both auditory and visual inputs. Prior work which decodes semantic meanings from neuroimaging data largely exploits brain activation patterns triggered by stimulation in cross-modality (i.e. text-audio pairs, text-picture pairs). Their goal is to develop a more sophisticated computational model to probing what information from the act of language understanding is represented in human brain. While how the brain receiving such information influences decoding performance is underestimated. This study dissociates multisensory integration of word understanding into written text, spoken text and image perception respectively, exploring the decoding efficiency and reliability of unisensory information in the brain representation. The findings suggest that, in terms of unisensory, decoding is most successful when semantics is represented in pictures, but the effect disappears in the case of congeneric words which share a related meaning. These results reveal the modality dependence and multisensory enhancement in the brain decoding methodology.

Brain activation can reflect semantic information elicited by natural words and concepts. Increasing research has been conducted on decoding such neural activation patterns using representational semantic models. However, prior work decoding semantic meaning from neurophysiological responses has been largely limited to ECoG, fMRI, MEG, and EEG techniques, each having its own advantages and limitations. More recently, the functional near infrared spectroscopy (fNIRS) has emerged as an alternative hemodynamic-based approach and possesses a number of strengths. We investigate brain decoding tasks under the help of fNIRS and empirically compare fNIRS with fMRI. Primarily, we find that: 1) like fMRI scans, activation patterns recorded from fNIRS encode rich information for discriminating concepts, but show limits on the possibility of decoding fine-grained semantic clues; 2) fNIRS decoding shows robustness across different brain regions, semantic categories and even subjects; 3) fNIRS has higher accuracy being decoded based on multi-channel patterns as compared to single-channel ones, which is in line with our intuition of the working mechanism of human brain. Our findings prove that fNIRS has the potential to promote a deep integration of NLP and cognitive neuroscience from the perspective of language understanding. We release the largest fNIRS dataset by far to facilitate future research.

Real human language mechanisms and the artificial intelligent language processing methods are two independent systems. Exploring the relationship between the two can help develop human-like language models and is also beneficial to reveal the neuroscience of the reading brain. The flourishing research in this interdisciplinal research field calls for surveys to systemically study and analyze the recent successes. However, such a comprehensive review still cannot be found, which motivates our work. This article first briefly introduces the interdisciplinal research progress, then systematically discusses the task of brain decoding from the perspective of simple concepts and complete sentences, and also describes main limitations in this field and put forward with possible solutions. Finally, we conclude this survey with certain open research questions that will stimulate further studies.

SUMMARY: Sampling of control variants having matched properties with input variants is widely used in enrichment analysis of genome-wide association studies/quantitative trait loci and negative data construction for pathogenic/regulatory variant prediction methods. Spurious enrichment results because of confounding factors, such as minor allele frequency and linkage disequilibrium pattern, can be avoided by calibration of statistical significance based on matched controls. Here, we presented vSampler which can generate sets of randomly drawn variants with comprehensive choices of matching properties, such as tissue/cell type-specific epigenomic features. Importantly, the development of a novel data structure and sampling algorithms for vSampler makes it significantly fast than existing tools. AVAILABILITY AND IMPLEMENTATION: vSampler web server and local program are available at http://mulinlab.org/vsampler. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Contextualized representations trained over large raw text data have given remarkable improvements for NLP tasks including question answering and reading comprehension. There have been works showing that syntactic, semantic and word sense knowledge are contained in such representations, which explains why they benefit such tasks. However, relatively little work has been done investigating commonsense knowledge contained in contextualized representations, which is crucial for human question answering and reading comprehension. We study the commonsense ability of GPT, BERT, XLNet, and RoBERTa by testing them on seven challenging benchmarks, finding that language modeling and its variants are effective objectives for promoting models' commonsense ability while bi-directional context and larger training set are bonuses. We additionally find that current models do poorly on tasks require more necessary inference steps. Finally, we test the robustness of models by making dual test cases, which are correlated so that the correct prediction of one sample should lead to correct prediction of the other. Interestingly, the models show confusion on these test cases, which suggests that they learn commonsense at the surface rather than the deep level. We release a test set, named CATs publicly, for future research.

Deep learning has led to significant improvement in text summarization with various methods investigated and improved ROUGE scores reported over the years.However, gaps still exist between summaries produced by automatic summarizers and human professionals.Aiming to gain more understanding of summarization systems with respect to their strengths and limits on a fine-grained syntactic and semantic level, we consult the Multidimensional Quality Metric 1 (MQM) and quantify 8 major sources of errors on 10 representative summarization models manually.Primarily, we find that 1) under similar settings, extractive summarizers are in general better than their abstractive counterparts thanks to strength in faithfulness and factual-consistency; 2) milestone techniques such as copy, coverage and hybrid extractive/abstractive methods do bring specific improvements but also demonstrate limitations; 3) pre-training techniques, and in particular sequence-to-sequence pre-training, are highly effective for improving text summarization, with BART giving the best results.

Machine reading is a fundamental task for testing the capability of natural language understand- ing, which is closely related to human cognition in many aspects. With the rising of deep learning techniques, algorithmic models rival human performances on simple QA, and thus increasingly challenging machine reading datasets have been proposed. Though various challenges such as evidence integration and commonsense knowledge have been integrated, one of the fundamental capabilities in human reading, namely logical reasoning, is not fully investigated. We build a comprehensive dataset, named LogiQA, which is sourced from expert-written questions for testing human Logical reasoning. It consists of 8,678 QA instances, covering multiple types of deductive reasoning. Results show that state-of-the-art neural models perform by far worse than human ceiling. Our dataset can also serve as a benchmark for reinvestigating logical AI under the deep learning NLP setting. The dataset is freely available at https://github.com/lgw863/LogiQA-dataset.

Feedback tools help people to monitor information about themselves to improve their health, sustainability practices, or personal well-being. Yet reasoning about personal data (e.g., pedometer counts, blood pressure readings, or home electricity consumption) to gain a deep understanding of your current practices and how to change can be challenging with the data alone. We integrate quantitative feedback data within a personal digital calendar; this approach aims to make the feedback data readily accessible and more comprehensible. We report on an eight-week field study of an on-calendar visualization tool. Results showed that a personal calendar can provide rich context for people to reason about their feedback data. The on-calendar visualization enabled people to quickly identify and reason about regular patterns and anomalies. Based on our results, we also derived a model of the behavior feedback process that extends existing technology adoption models. With that, we reflected on potential barriers for the ongoing use of feedback tools.

Data surrounds each and every one of us in our daily lives, ranging from exercise logs, to archives of our interactions with others on social media, to online resources pertaining to our hobbies. There is enormous potential for us to use these data to understand ourselves better and make positive changes in our lives. Visualization (Vis) and visual analytics (VA) offer substantial opportunities to help individuals gain insights about themselves, their communities and their interests; however, designing tools to support data analysis in non-professional life brings a unique set of research and design challenges. We investigate the requirements and research directions required to take full advantage of Vis and VA in a personal context. We develop a taxonomy of design dimensions to provide a coherent vocabulary for discussing personal visualization and personal visual analytics. By identifying and exploring clusters in the design space, we discuss challenges and share perspectives on future research. This work brings together research that was previously scattered across disciplines. Our goal is to call research attention to this space and engage researchers to explore the enabling techniques and technology that will support people to better understand data relevant to their personal lives, interests, and needs.

A human-robot cooperative approach to reliable planning and execution is presented. The human-robot system consists of three components: human user, wheelchair robot and the noninvasive brain-computer interface (BCI) which can represent limit types of user's intention patterns based on EEG signals, with insufficient decoding accuracy and time delay. To achieve efficient navigation and positioning under condition of decoding uncertainties of the BCI, three cooperative modes are proposed for specific situations based on trade-off of robot's autonomy and user's flexibility. The coding protocol in each mode is elucidated in detail, and strategies of mode switching are developed. To achieve continuous and smooth motion, a look-ahead visual feedback is applied, so that the user can adjust the intention and/or actively correct extraction error of the BCI before the robot reaches current path node, and consequently, reliable planning and execution are ensured. The effectiveness of the strategies is evaluated by simulations.

Abstract Project schedules are effectively represented by Gantt charts, but comparing multiple versions of a schedule is difficult. To compare versions with current methods, users must search and navigate through multiple large documents, making it difficult to identify differences. We present two novel visualization techniques to support the comparison of Gantt charts. First, we encode two Gantt charts in one view by overlapping them to show differences. Second, we designed an interactive visual technique, the ‘TbarView’, that allows users to compare multiple schedules within one single view. We evaluated the overlap and TbarView techniques via a user study. The study results showed that our design provided a quick overview of the variances among two or more schedules, and the techniques also improved efficiency by minimizing view switching. Our visual techniques for schedule comparison could be combined with other resource analysis tools to help project teams identify and resolve errors and problems in project schedules.

In wireless sensor networks, sensors should have some mechanisms to learn their locations since sensed data without associated location information may be meaningless. While many sensor localization algorithms have been proposed, security issues in sensor localization are usually not addressed in their original design. Secure sensor localization is very challenging due to limited computation and energy resources in sensors. It is highly desirable that a localization scheme is robust and is able to detect malicious attacks without using complex cryptographic operations. In this paper, we present and analyze detection methods purely based on geometric constraints in sensor networks. Our detection methods can strengthen the localization algorithm to be resilient to malicious attacks by detecting and eliminating the negative impact of fake information.