Search

Persuading people to change their opinions is a common practice in online discussion forums on topics ranging from political campaigns to relationship consultation. Enhancing people's ability to write persuasive arguments could not only practice their critical thinking and reasoning but also contribute to the effectiveness and civility in online communication. It is, however, not an easy task in online discussion settings where written words are the primary communication channel. In this paper, we derived four design goals for a tool that helps users improve the persuasiveness of arguments in online discussions through a survey with 123 online forum users and interviews with five debating experts. To satisfy these design goals, we analyzed and built a labeled dataset of fine-grained persuasive strategies (i.e., logos, pathos, ethos, and evidence) in 164 arguments with high ratings on persuasiveness from ChangeMyView, a popular online discussion forum. We then designed an interactive visual system, Persua, which provides example-based guidance on persuasive strategies to enhance the persuasiveness of arguments. In particular, the system constructs portfolios of arguments based on different persuasive strategies applied to a given discussion topic. It then presents concrete examples based on the difference between the portfolios of user input and high-quality arguments in the dataset. A between-subjects study shows suggestive evidence that Persua encourages users to submit more times for feedback and helps users improve more on the persuasiveness of their arguments than a baseline system. Finally, a set of design considerations was summarized to guide future intelligent systems that improve the persuasiveness in text., Comment: This paper will appear in CSCW 2022

Recently, large pretrained language models have achieved compelling performance on commonsense benchmarks. Nevertheless, it is unclear what commonsense knowledge the models learn and whether they solely exploit spurious patterns. Feature attributions are popular explainability techniques that identify important input concepts for model outputs. However, commonsense knowledge tends to be implicit and rarely explicitly presented in inputs. These methods cannot infer models' implicit reasoning over mentioned concepts. We present CommonsenseVIS, a visual explanatory system that utilizes external commonsense knowledge bases to contextualize model behavior for commonsense question-answering. Specifically, we extract relevant commonsense knowledge in inputs as references to align model behavior with human knowledge. Our system features multi-level visualization and interactive model probing and editing for different concepts and their underlying relations. Through a user study, we show that CommonsenseVIS helps NLP experts conduct a systematic and scalable visual analysis of models' relational reasoning over concepts in different situations., Comment: This paper is accepted by IEEE VIS, 2023. To appear in IEEE Transactions on Visualization and Computer Graphics (IEEE TVCG). 14 pages, 11 figures

Livestream e-commerce integrates live streaming and online shopping, allowing viewers to make purchases while watching. However, effective marketing strategies remain a challenge due to limited empirical research and subjective biases from the absence of quantitative data. Current tools fail to capture the interdependence between live performances and feedback. This study identified computational features, formulated design requirements, and developed LiveRetro, an interactive visual analytics system. It enables comprehensive retrospective analysis of livestream e-commerce for streamers, viewers, and merchandise. LiveRetro employs enhanced visualization and time-series forecasting models to align performance features and feedback, identifying influences at channel, merchandise, feature, and segment levels. Through case studies and expert interviews, the system provides deep insights into the relationship between live performance and streaming statistics, enabling efficient strategic analysis from multiple perspectives., Comment: Accepted by IEEE VIS 2023

Generative text-to-image models have gained great popularity among the public for their powerful capability to generate high-quality images based on natural language prompts. However, developing effective prompts for desired images can be challenging due to the complexity and ambiguity of natural language. This research proposes PromptMagician, a visual analysis system that helps users explore the image results and refine the input prompts. The backbone of our system is a prompt recommendation model that takes user prompts as input, retrieves similar prompt-image pairs from DiffusionDB, and identifies special (important and relevant) prompt keywords. To facilitate interactive prompt refinement, PromptMagician introduces a multi-level visualization for the cross-modal embedding of the retrieved images and recommended keywords, and supports users in specifying multiple criteria for personalized exploration. Two usage scenarios, a user study, and expert interviews demonstrate the effectiveness and usability of our system, suggesting it facilitates prompt engineering and improves the creativity support of the generative text-to-image model., Comment: Accepted full paper for IEEE VIS 2023

The fast reactor is gradually being paid attention to in nuclear power due to a better breeding effect, and its economy is also directly followed. Due to the restriction of the core power density, the core power level is directly related to the volume of the core active area. When the power level and the volume of the core gradually increase, the mass of uranium and 235U in the core at the beginning of life will gradually increase on the basis of a critical uranium inventory. At the same time, due to the gradual increase in the breeding effect, the utilization rate of uranium resources at the end of life will gradually increase. Due to the characteristics of the large critical uranium mass and high breeding effect, the mass of uranium and 235U in the low power level core at the beginning of life cannot be further reduced because of the critical uranium inventory. The lifetime of the high power level core should be extended to use the wealthy reactivity of the core because of the improvement in the breeding effect. However, due to the requirements of the nuclear power refuel time and the limitation to fuel consumption, the breeding effect brought by the improvement in the core power level and the size of the active zone may be less effective. In this article, the power level and economy of lead-based modular nuclear power are analyzed, and the economy of different power level fast reactors with 2,000 EFPD lifetime such as 100 MWt, 300 MWt, 500 MWt, 700 MWt, and 1000 MWt thermal power is designed and analyzed. By analyzing the invention of uranium and 235U and the fuel utilization of different core schemes, the best economic analysis of the single reactor power of the current lead-based modular reactor is proposed and the foundation of the lead-based modular nuclear power is provided.

Natural language interfaces (NLIs) provide users with a convenient way to interactively analyze data through natural language queries. Nevertheless, interactive data analysis is a demanding process, especially for novice data analysts. When exploring large and complex SQL databases from different domains, data analysts do not necessarily have sufficient knowledge about different data tables and application domains. It makes them unable to systematically elicit a series of topically-related and meaningful queries for insight discovery in target domains. We develop a NLI with a step-wise query recommendation module to assist users in choosing appropriate next-step exploration actions. The system adopts a data-driven approach to suggest semantically relevant and context-aware queries for application domains of users' interest based on their query logs. Also, the system helps users organize query histories and results into a dashboard to communicate the discovered data insights. With a comparative user study, we show that our system can facilitate a more effective and systematic data analysis process than a baseline without the recommendation module., 14 pages, 6 figures

Ubiquitination is an important reversible post-translational modification. Many viruses hijack the host ubiquitin system to enhance self-replication. In the present study, we found that Avibirnavirus VP3 protein was ubiquitinated during infection and supported virus replication by ubiquitination. Mass spectrometry and mutation analysis showed that VP3 was ubiquitinated at residues K73, K135, K158, K193, and K219. Virus rescue showed that ubiquitination at sites K73, K193, and K219 on VP3 could enhance the replication abilities of infectious bursal disease virus (IBDV), and that K135 was essential for virus survival. Binding of the zinc finger domain of TRAF6 (TNF receptor associated factor 6) to VP3 mediated K11- and K33-linked ubiquitination of VP3, which promoted its nuclear accumulation to facilitate virus replication. Additionally, VP3 could inhibit TRAF6-mediated NFKB/NF-κB (nuclear factor kappa B) activation and IFNB/IFN-β (interferon beta) production to evade host innate immunity by inducing TRAF6 autophagic degradation in an SQSTM1/p62 (sequestosome 1)-dependent manner. Our findings demonstrated a macroautophagic/autophagic mechanism by which Avibirnavirus protein VP3 blocked NFKB-mediated IFNB production by targeting TRAF6 during virus infection, and provided a potential drug target for virus infection control. Abbreviations: ATG: autophagy related; BafA1: bafilomycin A1; CALCOCO2/NDP52: calcium binding and coiled-coil domain 2; Cas9: CRISPR-associated protein 9; CHX: cycloheximide; Co-IP: co-immunoprecipitation; CRISPR: clustered regularly interspaced short palindromic repeats; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; GST: glutathione S-transferase; IBDV: infectious bursal disease virus; IF: indirect immunofluorescence; IFNB/IFN-β: interferon beta; mAb: monoclonal antibody; MAP1LC3/LC3: microtubule associated protein 1 light chain 3; MOI: multiplicity of infection; MS: mass spectrometry; NFKB/NF-κB: nuclear factor kappa B; NBR1: NBR1 autophagy cargo receptor; OPTN: optineurin; pAb: polyclonal antibody; PRRs: pattern recognition receptors; RNF125: ring finger protein 125; RNF135/Riplet: ring finger protein 135; SQSTM1/p62: sequestosome 1; TAX1BP1: tax1 binding protein1; TCID50: 50% tissue culture infective dose; TRAF3: TNF receptor associated factor 3; TRAF6: TNF receptor associated factor 6; TRIM25: tripartite motif containing 25; Ub: ubiquitin; Wort: wortmannin; WT: wild type.

The promise of multimodal models for real-world applications has inspired research in visualizing and understanding their internal mechanics with the end goal of empowering stakeholders to visualize model behavior, perform model debugging, and promote trust in machine learning models. However, modern multimodal models are typically black-box neural networks, which makes it challenging to understand their internal mechanics. How can we visualize the internal modeling of multimodal interactions in these models? Our paper aims to fill this gap by proposing MultiViz, a method for analyzing the behavior of multimodal models by scaffolding the problem of interpretability into 4 stages: (1) unimodal importance: how each modality contributes towards downstream modeling and prediction, (2) cross-modal interactions: how different modalities relate with each other, (3) multimodal representations: how unimodal and cross-modal interactions are represented in decision-level features, and (4) multimodal prediction: how decision-level features are composed to make a prediction. MultiViz is designed to operate on diverse modalities, models, tasks, and research areas. Through experiments on 8 trained models across 6 real-world tasks, we show that the complementary stages in MultiViz together enable users to (1) simulate model predictions, (2) assign interpretable concepts to features, (3) perform error analysis on model misclassifications, and (4) use insights from error analysis to debug models. MultiViz is publicly available, will be regularly updated with new interpretation tools and metrics, and welcomes inputs from the community., Comment: ICLR 2023. Code available at: https://github.com/pliang279/MultiViz

Existing generative pre-trained language models (e.g., GPT) focus on modeling the language structure and semantics of general texts. However, those models do not consider the numerical properties of numbers and cannot perform robustly on numerical reasoning tasks (e.g., math word problems and measurement estimation). In this paper, we propose NumGPT, a generative pre-trained model that explicitly models the numerical properties of numbers in texts. Specifically, it leverages a prototype-based numeral embedding to encode the mantissa of the number and an individual embedding to encode the exponent of the number. A numeral-aware loss function is designed to integrate numerals into the pre-training objective of NumGPT. We conduct extensive experiments on four different datasets to evaluate the numeracy ability of NumGPT. The experiment results show that NumGPT outperforms baseline models (e.g., GPT and GPT with DICE) on a range of numerical reasoning tasks such as measurement estimation, number comparison, math word problems, and magnitude classification. Ablation studies are also conducted to evaluate the impact of pre-training and model hyperparameters on the performance., 8 pages, 3 figures

Recently, organic-inorganic halide perovskite solar cells (PSCs) have received extensive research in the field of optoelectronic materials due to their unique optical and electrical properties, especially lead-based PSCs. However, the toxicity and stability of these devices, as well as the expensive hole transport layer (HTL) and other factors inhibit their commercial production. In this work, the non-toxic tin was applied as the battery material, the perovskite solar cell adopts an inverted HTL-free structure, and the one-dimensional solar cell capacitor simulator SCAPS-1D (Solar Cell Capacitance Simulator) was adopted for numerical simulation and found that FTO/CH3 NH3 SnI3 /C60 /Au structure PSCs also showed excellent photovoltaic performance. We studied the influence of the thickness of the absorber layer, the defect density, the doping concentration of different layers, and the thickness of the electron transport layer (ETL) under different directions of illumination on the battery performance. The simulation results show that the optimized inverted HTL-free tin-based PSCs based on C60 are with inspiring performance: a short-circuit current density (JSC ) of 30.1646 mA/cm 2 , open-circuit voltage (VOC ) of 1.0465V, fill factor (FF) of 59.49% and power conversion efficiency (PCE) of 18.78%. We also introduced light from different directions to irradiate PSCs, and the results show that the HTL-free perovskite adopting an inverted structure can retain the light intensity of the irradiated perovskite layer to the greatest extent and exhibit superior performance. Based on the inverted HTL-free tin-based PSCs, we also investigated the performance parameters of ETL batteries with different materials. This work provides new ideas for PSCs development in the future.

In the field of automatic inspection and factory quality inspection, it is often necessary to analyze the readings of many different types of meters to understand the working status of the factory. Therefore, it is particularly important to get an accurate reading. To solve this problem, an automatic meter reading system based on core technologies such as image recognition, deep learning, internet of things and intelligent machine car was designed. We adopted YOLOV3, SIFT, FLANN, Hough transform and other efficient algorithms to achieve the function of reading meter. Our system is competent for the reading of digital meter as well as pointer meter. Experiments show that our system has a good effect.

In this paper, we propose a closed-form IMU error state covariance integration method for optimization-based visual- inertial state estimator. We derive a closed-form solutions of the IMU error state covariance propagation in pre-integration process, yielding improved accuracy of IMU residual information matrix. Our visual-inertial state estimator is based on a tightly-coupled, sliding-window optimization framework, which jointly estimate the IMU states and landmarks and performing marginalization to limit the computational cost. Finally, the system are validated in park environment dataset, the result shows our proposed method is effective.

In this paper, based on the DEAP dataset, we studied the effects of different EEG channel selection methods on the accuracy of emotion recognition for different frequency bands. First, the discrete wavelet transform method is used to divide the EEG signals into four bands of gamma, beta, alpha and theta, and extract the entropy and energy of each band as classification features. Then, the following three channel selection methods are compared to select the best EEG channel combination for the four emotions classification susing, the channel selection method based on experience, the indirect channel selection method based on the mRMR feature selection algorithm, and the direct channel selection method based on the mRMR feature selection algorithm. Finally, the extreme learning machine with kernel is used to verify the effectiveness of the channel selection method. The results show that based on the mRMR feature selection algorithm, the channel selection method taking each channel as a whole is more powerful in balancing the number of channels and classification accuracy. In the beta band, the number of channels is reduced from 32 to 22, which is only 1.37% (from 80.83% to 79.37%) lower than the best classification accuracy, and the emotion recognition performance remains at a high level. Compared with the results of others, this paper can use less channels to achieve similar or higher emotional recognition performance than others, which further proves the effectiveness of the method. In addition, we also found that high frequencies (gamma and beta bands) are better for emotional recognition. This study provides a reference for channel and band selection in EEG-based emotion recognition.

A fast transient response low dropout (LDO) regulator based on push-pull structure is presented in this paper. The push-pull circuit sampling the output of the LDO and adjusting the gate voltage of the power MOSFET by push-pull current. The slew rate of the power MOSFET gate is accelerated and the transient response performance is improved. With a 1μF decoupling under 180 nm CMOS technology, the proposed LDO exhibits an excellent transient performance. When the workload changes steeply between 100μA and 20mA in 100ns, the overshoot is 10.176mV and undershoot is 10.795mV with the recovery time are 1.272μs and 0.888μs respectively.

Spectral imaging apparatus in current use are often cumbersome, costly and slow in operation, which becomes a major obstacle to extensive use of spectral imaging in several application areas. In recent years, the technical and commercial success of color filter array (CFA) based imaging systems has motivated researchers to generalise and expand the concept of CFA to achieve efficient spectral imaging through the use of the spectral filter array (SFA). This dissertation expounds the research into the filter array approach to spectral imaging based on a simulation framework, from the development of demosaicking methods to the design and evaluation at the system level. The dissertation first presents the development of the field of spectral imaging from its roots in spectroscopy and imaging, and explores the state-of-the-art solutions based on SFA from design to realisation. It then proposes a simulation framework composed of the major parts in a typical imaging pipeline. Based on this, the influence of chromatic aberration on CFA demosaicking and the impact of filter bandwidth on spectral reconstruction were evaluated. The results helped to better understand the delicate interactions between the components in the pipeline and verify the validity of the simulation framework. On the basis of the framework, three novel SFA demosaicking methods were developed and evaluated. The methods differ fundamentally and thus featuring distinct properties, as confirmed by the experimental results. The key to understanding the differences lies in the way demosaicking methods deal with the spatial and spectral correlation between pixels in a mosaicked image. An evaluation of the colorimetric performance shows that a properly designed SFA-based imaging system may also be useful for colour image acquisition. Lastly, performance of the proposed and conventional demosaicking methods were scrutinised, given the characteristics and parameters of a real-world SFA sensor design. We conclude that, for a successful SFA-based spectral imaging system design, it is important to consider carefully the joint influence of all the involved modules as well as the requirements and constraints of applications. And we hope that the use of SFA based spectral imaging is expected to be wider in the foreseeable future in the light of technological advances and market demand.

International audience; In every aspect, filter design plays an important role in an image acquisition system based on a single image sensor and a colour filter array (CFA) mounted onto the sensor. Complementary CFAs are used by some colour cameras in the interest of higher sensitivity, which motivated us to employ filters of wide pass bands in the effort to adapt CFA for multispectral image acquisition. In this context, filter design has an effect on the accuracy of spectrum reconstruction in addition to other aspects. The results show that wider bandwidths in general result in more faithful spectrum reconstruction and higher signal-to-noise performance.

Ozonation is one of the most important steps in drinking water treatment plants. The efficiency of ozonation is directly related to the treated water quality. However, ozone dosing process is very difficult to be controlled because of its complicated nonlinear behaviour. In this paper, an advanced control scheme based on model predictive control (MPC) is proposed for the ozone dosing process. With the proposed control scheme, a RBF neural network model is established which could be used for MPC. The control strategy of maintaining a constant ozone exposure is adopted for effective disinfection requirements. The full-scale experimental results demonstrate the effectiveness of this advanced control method.

This is an Open Access article. This is the publisher’s PDF originally published in Journal of the International Colour Association: http://aic-colour-journal.org/index.php/JAIC/article/view/149 In every aspect, spectral characteristics of filters play an important role in an image acquisition system. For a colorimetric system, traditionally, it is believed that narrow-band filters give rise to higher accuracy of colour reproduction, whereas wide-band filters, such as complementary colour filters, have the advantage of higher sensitivity. In the context of multispectral image capture, the objective is very often to retrieve an estimation of the spectral reflectance of the captured objects. The literature does not provide a satisfactory answer to which configuration yields the best results. It is therefore of interest to verify which type of filters performs the best in estimating the reflectance spectra for the purpose of multispectral image acquisition. A series of experiments were conducted on a simulated imaging system, with six types of filters of varying bandwidths paired with three linear reflectance estimation methods. The results show that filter bandwidth exerts direct influence on the accuracy of reflectance estimation. Extremely narrowband filters did not perform well in the experiment and the relation between bandwidth and reflectance estimation accuracy is not monotonic. Also it is indicated that the optimal number of filters depends on the spectral similarity metrics employed.

The localization of non-cooperative illuminators is a hot and difficult problem in the modern war. The arithmetic of Inverse Diffraction Parabolic Equation has a good applicability and stability characteristic for the passive localization in complex environment. Based on the localization method in the symmetrical atmosphere, the paper analyzed localization of single and couple illuminators on the conditions of standard atmosphere and terrain defilade, proposing the methods of data points selection and error ellipse protraction and having investigated the relationship between parameters and localization errors. Simulation analysis proved the characteristics of the algorithm. In the end, some parameters selection measure is presented to increase the accuracy of the localization algorithm and extend its application.

We have recreated and expressed two known natural mutations within the LCAT gene which were reported on both alleles in a single case of familial LCAT deficiency. We demonstrate that the Ala-93-->Thr mutation is responsible for the biochemical defect while the Arg-158-->Cys mutation is a co-inherited natural polymorphism of LCAT which results in normal enzyme function.