Prior Domain Research Articles

Bayesian inference of real-time dynamics from lattice QCD

The computation of dynamical properties of nuclear matter, ranging from parton distribution functions of nucleons and nuclei to transport properties in the quark-gluon plasma, constitutes a central goal of modern theoretical physics. This real-time physics often defies a perturbative treatment and the most successful strategy so far is to deploy lattice QCD simulations. These numerical computations are based on Monte-Carlo sampling and formulated in an artificial Euclidean time. Real-time physics is most conveniently formulated in terms of spectral functions, which are hidden in lattice QCD behind an ill-posed inverse problem. I will discuss state-of-the art methods in the extraction of spectral functions from lattice QCD simulations, based on Bayesian inference and emphasize the importance of prior domain knowledge, vital to regularizing the otherwise ill-posed extraction task. With Bayesian inference allowing us to make explicit the uncertainty in both observations and in our prior knowledge, a systematic estimation of the total uncertainties in the extracted spectral functions is nowadays possible. Two implementations of the Bayesian Reconstruction (BR) method for spectral function extraction, one for MAP point estimates and one based on an open access Monte-Carlo sampler are provided. I will briefly touch on the use of machine learning for spectral function reconstruction and discuss some new insight it has brought to the Bayesian community.

Semi-supervised adversarial discriminative domain adaptation.

Domain adaptation is a potential method to train a powerful deep neural network across various datasets. More precisely, domain adaptation methods train the model on training data and test that model on a completely separate dataset. The adversarial-based adaptation method became popular among other domain adaptation methods. Relying on the idea of GAN, the adversarial-based domain adaptation tries to minimize the distribution between the training and testing dataset based on the adversarial learning process. We observe that the semi-supervised learning approach can combine with the adversarial-based method to solve the domain adaptation problem. In this paper, we propose an improved adversarial domain adaptation method called Semi-Supervised Adversarial Discriminative Domain Adaptation (SADDA), which can outperform other prior domain adaptation methods. We also show that SADDA has a wide range of applications and illustrate the promise of our method for image classification and sentiment classification problems.

A comprehensive artificial intelligence framework for dental diagnosis and charting

BackgroundThe aim of this study was to develop artificial intelligence (AI) guided framework to recognize tooth numbers in panoramic and intraoral radiographs (periapical and bitewing) without prior domain knowledge and arrange the intraoral radiographs into a full mouth series (FMS) arrangement template. This model can be integrated with different diseases diagnosis models, such as periodontitis or caries, to facilitate clinical examinations and diagnoses.MethodsThe framework utilized image segmentation models to generate the masks of bone area, tooth, and cementoenamel junction (CEJ) lines from intraoral radiographs. These masks were used to detect and extract teeth bounding boxes utilizing several image analysis methods. Then, individual teeth were matched with a patient’s panoramic images (if available) or tooth repositories for assigning tooth numbers using the multi-scale matching strategy. This framework was tested on 1240 intraoral radiographs different from the training and internal validation cohort to avoid data snooping. Besides, a web interface was designed to generate a report for different dental abnormalities with tooth numbers to evaluate this framework’s practicality in clinical settings.ResultsThe proposed method achieved the following precision and recall via panoramic view: 0.96 and 0.96 (via panoramic view) and 0.87 and 0.87 (via repository match) by handling tooth shape variation and outperforming other state-of-the-art methods. Additionally, the proposed framework could accurately arrange a set of intraoral radiographs into an FMS arrangement template based on positions and tooth numbers with an accuracy of 95% for periapical images and 90% for bitewing images. The accuracy of this framework was also 94% in the images with missing teeth and 89% with restorations.ConclusionsThe proposed tooth numbering model is robust and self-contained and can also be integrated with other dental diagnosis modules, such as alveolar bone assessment and caries detection. This artificial intelligence-based tooth detection and tooth number assignment in dental radiographs will help dentists with enhanced communication, documentation, and treatment planning accurately. In addition, the proposed framework can correctly specify detailed diagnostic information associated with a single tooth without human intervention.

Annotating Motion Primitives for Simplifying Action Search in Reinforcement Learning

Reinforcement learning in large-scale environments is challenging due to the many possible actions that can be taken in specific situations. We have previously developed a means of constraining, and hence speeding up, the search process through the use of motion primitives; motion primitives are sequences of pre-specified actions taken across a state series. As a byproduct of this work, we have found that if the motion primitives’ motions and actions are labeled, then the search can be sped up further. Since motion primitives may initially lack such details, we propose a theoretically viewpoint-insensitive and speed-insensitive means of automatically annotating the underlying motions and actions. We do this through a differential-geometric, spatio-temporal kinematics descriptor, which analyzes how the poses of entities in two motion sequences change over time. We use this descriptor in conjunction with a weighted-nearest-neighbor classifier to label the primitives using a limited set of training examples. In our experiments, we achieve high motion and action annotation rates for human-action-derived primitives with as few as one training sample. We also demonstrate that reinforcement learning using accurately labeled trajectories leads to high-performing policies more quickly than standard reinforcement learning techniques. This is partly because motion primitives encode prior domain knowledge and preempt the need to re-discover that knowledge during training. It is also because agents can leverage the labels to systematically ignore action classes that do not facilitate task objectives, thereby reducing the action space.

An End-to-End Curriculum Learning Approach for Autonomous Driving Scenarios

In this work, we combine Curriculum Learning with Deep Reinforcement Learning to learn without any prior domain knowledge, an end-to-end competitive driving policy for the CARLA autonomous driving simulator. To our knowledge, we are the first to provide consistent results of our driving policy on all towns available in CARLA. Our approach divides the reinforcement learning phase into multiple stages of increasing difficulty, such that our agent is guided towards learning an increasingly better driving policy. The agent architecture comprises various neural networks that complements the main convolutional backbone, represented by a ShuffleNet V2. Further contributions are given by (i) the proposal of a novel value decomposition scheme for learning the value function in a stable way and (ii) an ad-hoc function for normalizing the growth in size of the gradients. We show both quantitative and qualitative results of the learned driving policy.

Half-sibling regression meets exoplanet imaging: PSF modeling and subtraction using a flexible, domain knowledge-driven, causal framework

Context. High-contrast imaging of exoplanets hinges on powerful post-processing methods to denoise the data and separate the signal of a companion from its host star, which is typically orders of magnitude brighter. Aims. Existing post-processing algorithms do not use all prior domain knowledge that is available about the problem. We propose a new method that builds on our understanding of the systematic noise and the causal structure of the data-generating process. Methods. Our algorithm is based on a modified version of half-sibling regression (HSR), a flexible denoising framework that combines ideas from the fields of machine learning and causality. We adapted the method to address the specific requirements of high-contrast exoplanet imaging data obtained in pupil tracking mode. The key idea is to estimate the systematic noise in a pixel by regressing the time series of this pixel onto a set of causally independent, signal-free predictor pixels. We use regularized linear models in this work; however, other (nonlinear) models are also possible. In a second step, we demonstrate how the HSR framework allows us to incorporate observing conditions such as wind speed or air temperature as additional predictors. Results. When we applied our method to four data sets from the VLT/NACO instrument, our algorithm provided a better false-positive fraction than a popular baseline method in the field. Additionally, we found that the HSR-based method provides direct and accurate estimates for the contrast of the exoplanets without the need to insert artificial companions for calibration in the data sets. Finally, we present a first piece of evidence that using the observing conditions as additional predictors can improve the results. Conclusions. Our HSR-based method provides an alternative, flexible, and promising approach to the challenge of modeling and subtracting the stellar PSF and systematic noise in exoplanet imaging data.

A data-driven explainable case-based reasoning approach for financial risk detection

The rapid development of artificial intelligence methods contributes to their wide applications for forecasting various financial risks in recent years. This study introduces a novel explainable case-based reasoning (CBR) approach without a requirement of rich expertise in financial risk. Compared with other black-box algorithms, the explainable CBR system allows a natural economic interpretation of results. Indeed, the empirical results emphasize the interpretability of the CBR system in predicting financial risk, which is essential for both financial companies and their customers. In addition, our results show that the proposed automatic design CBR system has a good prediction performance compared to other artificial intelligence methods, overcoming the main drawback of a standard CBR system of highly depending on prior domain knowledge about the corresponding field.

A spatial attention guided deep learning system for prediction of pathological complete response using breast cancer histopathology images.

Predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in triple-negative breast cancer (TNBC) patients accurately is direly needed for clinical decision making. pCR is also regarded as a strong predictor of overall survival. In this work, we propose a deep learning system to predict pCR to NAC based on serial pathology images stained with hematoxylin and eosin and two immunohistochemical biomarkers (Ki67 and PHH3). To support human prior domain knowledge-based guidance and enhance interpretability of the deep learning system, we introduce a human knowledge-derived spatial attention mechanism to inform deep learning models of informative tissue areas of interest. For each patient, three serial breast tumor tissue sections from biopsy blocks were sectioned, stained in three different stains and integrated. The resulting comprehensive attention information from the image triplets is used to guide our prediction system for prognostic tissue regions. The experimental dataset consists of 26419 pathology image patches of 1000×1000 pixels from 73 TNBC patients treated with NAC. Image patches from randomly selected 43 patients are used as a training dataset and images patches from the rest 30 are used as a testing dataset. By the maximum voting from patch-level results, our proposed model achieves a 93% patient-level accuracy, outperforming baselines and other state-of-the-art systems, suggesting its high potential for clinical decision making. The codes, the documentation and example data are available on an open source at: https://github.com/jkonglab/PCR_Prediction_Serial_WSIs_biomarkers. Supplementary data are available at Bioinformatics online.

Power Line Communication and Sensing Using Time Series Forecasting.

Smart electrical grids rely on data communication to support their operation and on sensing for diagnostics and maintenance. Usually, the roles of communication and sensing equipment are different, i.e., communication equipment does not participate in sensing tasks and vice versa. Power line communication (PLC) offers a cost-effective solution for joint communication and sensing for smart grids. This is because the high-frequency PLC signals used for data communication also reveal detailed information regarding the health of the power lines that they travel through. Traditional PLC-based power line or cable diagnostic solutions are dependent on prior knowledge of the cable type, network topology, and/or characteristics of the anomalies. In this paper, we develop a power line sensing technique that can detect various types of cable anomalies without any prior domain knowledge. To this end, we design a solution that first uses time-series forecasting to predict the PLC channel state information at any given point in time based on its historical data. Under the approximation that the prediction error follows a Gaussian distribution, we then perform chi-squared statistical test to build an anomaly detector which identifies the occurrence of a cable fault. We demonstrate the effectiveness and universality of our sensing solution via evaluations conducted using both synthetic and real-world data extracted from low- and medium-voltage distribution networks.

Constraint Sampling Reinforcement Learning: Incorporating Expertise for Faster Learning

Online reinforcement learning (RL) algorithms are often difficult to deploy in complex human-facing applications as they may learn slowly and have poor early performance. To address this, we introduce a practical algorithm for incorporating human insight to speed learning. Our algorithm, Constraint Sampling Reinforcement Learning (CSRL), incorporates prior domain knowledge as constraints/restrictions on the RL policy. It takes in multiple potential policy constraints to maintain robustness to misspecification of individual constraints while leveraging helpful ones to learn quickly. Given a base RL learning algorithm (ex. UCRL, DQN, Rainbow) we propose an upper confidence with elimination scheme that leverages the relationship between the constraints, and their observed performance, to adaptively switch among them. We instantiate our algorithm with DQN-type algorithms and UCRL as base algorithms, and evaluate our algorithm in four environments, including three simulators based on real data: recommendations, educational activity sequencing, and HIV treatment sequencing. In all cases, CSRL learns a good policy faster than baselines.

Reducing the dependency of having prior domain knowledge for effective online information retrieval

AbstractSometimes Internet users struggle to find what they are looking for on the Internet due to information overload. Search engines intend to identify documents related to a given keyphrase on the Internet and provide suggestions. Having some background knowledge about a topic or a domain will help in building effective search keyphrases that will lead to accurate results in information retrieval. This is further pronounced among students that rely on the internet to learn about a new topic. Students might not have the required background knowledge to build effective keyphrases and find what they are looking for. In this research, we are addressing this problem, and aim to help students find relevant information online. This research furthers existing literature by enhancing information retrieval frameworks through keyphrase assignment, aiming to expose students to new terminologies, therefore reducing the dependency of having background knowledge about the domain under study. We evaluated this framework and identified how it can be enhanced to suggest more effective search keyphrases. Our proposed suggestion is to introduce a keyphrase Ranking Mechanism that will improve the keyphrase assignment part of the framework by taking into consideration the part‐of‐speech of the generated keyphrases. To evaluate the proposed approach, various data sets were downloaded and processed. The results obtained showed that our proposed approach produces more effective keyphrases than the existing framework.

A Novel Deep Class-Imbalanced Semisupervised Model for Wind Turbine Blade Icing Detection.

Wind energy is of great importance for future energy development. In order to fully exploit wind energy, wind farms are often located at high latitudes, a practice that is accompanied by a high risk of icing. Traditional blade icing detection methods are usually based on manual inspection or external sensors/tools, but these techniques are limited by human expertise and additional costs. Model-based methods are highly dependent on prior domain knowledge and prone to misinterpretation. Data-driven approaches can offer promising solutions but require a massive amount of labeled training data, which are not generally available. In addition, the data collected for icing detection tend to be imbalanced because, most of the time, wind turbines operate under normal conditions. To address these challenges, this article presents a novel deep class-imbalanced semisupervised (DCISS) model for estimating blade icing conditions. DCISS integrates class-imbalanced and semisupervised learning (SSL) using a prototypical network that can rebalance features and measure the similarities between labeled and unlabeled samples. In addition, a channel calibration attention module is proposed to improve the ability to extract features from raw data. The proposed model has been evaluated using the blade icing datasets of three wind turbines. Compared to the classical anomaly detection and state-of-the-art SSL algorithms, DCISS shows significant advantages in terms of accuracy. Compared to five different class-imbalanced loss functions, the proposed DCISS is competitive. The generalization and practicability of the proposed model are further verified in the use case of online estimation.

Identifying and Characterizing New Expressions of Community Framing during Polarization

Chile experienced a series of important protests between October and December 2019. This social unrest, as it was called, was fueled by social inequity and radically affected the nation's status quo. A large portion of the population demanded a new Constitution and changes to the current government, whereas another part of the population rejected these social demands. This created a highly polarized scenario evidenced through online social media interactions. Analyzing controversial issues that emerge naturally from conversations in online communities can offer a more wide-scale understanding of today's political and societal discussions. Here, we analyze group polarization in social networks by studying the 2019 Chilean social unrest. Specifically, we propose an unsupervised approach for identifying and characterizing community framing (i.e., discovering and understanding polarized concepts). Our approach is based on the sequential application of community detection, topic modeling, and word embedding methods. The novelty of having an unsupervised approach is that it facilitates the performance of scalable and objective framing analyses with minimal human intervention, as it does not require prior domain or network knowledge. Using this methodology, we observe that an apparently similar conversation topic across communities can actually have completely different meanings to each group. We noted, for instance, that while an online community linked the term gente (people) with communism and terrorism, the other associated it with police and military oppression. In this direction, our work can help to contextualize real-world social issues in online platforms, describing how users discuss similar concepts with opposing views.

Community-based service ecosystem evolution analysis

Services are flourishing dramatically, and continuously increasing interactions among them are resulting in a new phenomenon called “service ecosystems,” which has become a focus of academia and industry. Driven by technology innovation, changes in regulations, and changes in the competitive strategies of individual businesses, service ecosystems are constantly evolving. Service ecosystem evolution analysis is an emerging research problem of great significance. By analyzing service ecosystem evolution history, common evolution patterns can be identified, underlying driving forces can be discovered, and future evolution trends can be predicated so that service providers can adjust their competitive strategies in a timely manner to adapt to evolution trends. In this paper, a framework for identifying service ecosystem evolution patterns from the service community perspective is presented. First, following the approaches of community detection and community evolution analysis, time-series community evolution traces are identified from historical service ecosystem evolution, and a service community evolution prediction model is trained in accordance with such traces. Second, the prediction model is explained to show how different factors affect the evolution of service communities. Finally, an approach for assisting service providers in making business decisions is presented according to interpretable prediction results and prior domain knowledge. Experiments on a real-world dataset showed that this work can indeed provide business-level insights on service ecosystem evolution. Additionally, all the data and source code have been made fully open-source for service ecosystem researchers.

Classification based topic extraction using domain-specific vocabulary: a supervised approach

Recently, a probabilistic topic modelling approach, <span>latent dirichlet allocation (LDA), has been extensively applied in the arena of document classification. However, classical LDA is an unsupervised algorithm implemented using a fixed number of topics without prior domain knowledge and generates different outcomes with the change in the order of documents. This article presents a comprehensive framework to evade the order effect and unsupervised probabilistic nature. First, the framework creates the vocabulary specific to the category using a weight-dependent model that extracts distinctive features suitable for supervised classification. Then, it transforms a classified cluster of documents from the domain corpus to the relevant topic making it more robust to noise. The framework was tested on a comprehensive collection of benchmark news datasets that vary in sample size, class characteristics, and classification tasks. In contrast to the conventional classification methods, the proposed framework achieved 95.56% and 95.23% accuracy when applied on two datasets, indicating that the proposed algorithm has a better classification capability. Furthermore, the topics extracted from the classified clusters are highly relevant to domain categories.</span>

The evolution of metacognitive strategy use in an open-ended learning environment: Do prior domain knowledge and motivation play a role?

There is a growing interest in viewing self-regulated learning as events unfolding over time, especially when students perform learning tasks in computer-based environments. Metacognitive activities are critical events in self-regulated learning. This study investigated the evolution of metacognitive strategy use in an open-ended computer-based learning environment, Betty's Brain. The data were from 93 sixth graders who used Betty's Brain to learn about climate change for four days. We extracted indicators of metacognitive strategy use from action logs. A knowledge test and self-report questionnaire were administrated before students started using Betty's Brain to assess prior domain knowledge and motivation, respectively. Results showed that metacognitive strategy use increased from the first to the second day and remained stable from the second to the fourth day of the study. The evolution of these behaviors varied across students. Task value and prior domain knowledge partially explained the individual differences in this evolution. Task value and prior domain knowledge also predicted the use of metacognitive strategies. Self-efficacy did not influence metacognitive strategy use. These results suggest the need for further investigation into the role of motivation and prior domain knowledge in the temporal evolution of metacognitive events.

A contrastive learning approach for\xa0ICU false arrhythmia\xa0alarm reduction

The high rate of false arrhythmia alarms in Intensive Care Units (ICUs) can lead to disruption of care, negatively impacting patients’ health through noise disturbances, and slow staff response time due to alarm fatigue. Prior false-alarm reduction approaches are often rule-based and require hand-crafted features from physiological waveforms as inputs to machine learning classifiers. Despite considerable prior efforts to address the problem, false alarms are a continuing problem in the ICUs. In this work, we present a deep learning framework to automatically learn feature representations of physiological waveforms using convolutional neural networks (CNNs) to discriminate between true vs. false arrhythmia alarms. We use Contrastive Learning to simultaneously minimize a binary cross entropy classification loss and a proposed similarity loss from pair-wise comparisons of waveform segments over time as a discriminative constraint. Furthermore, we augment our deep models with learned embeddings from a rule-based method to leverage prior domain knowledge for each alarm type. We evaluate our method using the dataset from the 2015 PhysioNet Computing in Cardiology Challenge. Ablation analysis demonstrates that Contrastive Learning significantly improves the performance of a combined deep learning and rule-based-embedding approach. Our results indicate that the final proposed deep learning framework achieves superior performance in comparison to the winning entries of the Challenge.

DKDFN: Domain Knowledge-Guided deep collaborative fusion network for multimodal unitemporal remote sensing land cover classification

Land use and land cover maps provide fundamental information that has been used in different types of studies, ranging from public health to carbon cycling. However, the existing remote sensing image classification methods thus far suffer from the insufficient usage of multiple modalities, underconsideration of prior domain knowledge, and poor performance on minority classes. To alleviate these problems, we propose a novel domain knowledge-guided deep collaborative fusion network (DKDFN) with performance boosting for minority categories for land cover classification. More specifically, the DKDFN adopts a multihead encoder and a multibranch decoder structure. The architecture of the encoder probablizes sufficient mining of complementary information from multiple modalities, which are Sentinel-2, Sentinel-1, and SRTM Digital Elevation Data (SRTM) in our case. The multibranch decoder enables land cover classification in a multitask learning setup, performing semantic segmentation and reconstructing multimodal remote sensing indices, which are selected as representatives of domain knowledge. This design incorporates domain knowledge in an effective end-to-end manner. The training stage of our DKDFN is supervised by our proposed asymmetry loss function (ALF), which boosts performance on nearly all categories, especially the categories with a low frequency of occurrence. Ablation studies of the network suggest that our design logic is worth testing in any network with an encoder-decoder structure. The study is conducted in Hunan, China and is verified using a self-labeled multimodal unitemporal remote sensing image dataset. The comparative experiments between DKDFN and 6 state-of-the-art models (U-Net, SegNet, PSPNet, DeepLab, HRNet, MP-ResNet) testify to the superiority of our method and suggest its potential to be applied more widely to map land cover in other geographical areas given the availability of Sentinel-2, Sentinel-1, and SRTM data. The dataset can be downloaded by https://github.com/LauraChow/HunanMultimodalDataset.

Insulator Hydrophobic Image Edge Detection Algorithm considering Deconvolution and Deblurring Algorithm

In this paper, the Gram matrix is used to calculate the correlation of the filter response sets under different scale kernels learned by each layer of the network in the deconvolution, and the loss between the corresponding feature response correlations in the multilayer network is calculated. Linear summation is used to obtain a stable, multiscale image model representation. This paper extracts the contours of the salient areas of the image and adjusts the parameters of the deconvolution network to learn the salient area patterns of the image. At the same time, for the image to be generated, a shape template is used to limit the range of the area to be generated in order to obtain a shape image with similar patterns. When the spatial relative relationship characteristics of the image constituent objects are obvious, we appropriately add high-level semantic feature activation values for reinforcement. This paper solves the estimation of the unknown blur kernel by using image prior knowledge, filtering and gradient domain algorithms and other different technologies to obtain image jitter or scene movement information and estimate the size, location, and density of the blur kernel. This paper studies a relatively robust deconvolution model, which is insensitive to random noise, has stable effects, and can overcome the water ripple effect caused by the usual convolution process. This paper attempts to study the fuzzy model with variable space. The usual blur is a spatial invariant model; that is, a single kernel is used to describe the motion of all pixels on the image. By selecting different characteristic parameters, this paper conducts experimental research on some existing hydrophobic indicator function methods and calculates the relationship between characteristic parameters and hydrophobicity when different hydrophobic indicator functions are adopted. One characteristic of the hydrophobic image of composite insulators is low contrast. The traditional method of converting color images to grayscale images cannot improve the image contrast. This paper analyzes the hydrophobic image of the composite insulator, and the extracted B channel component image of the hydrophobic image improves the contrast of the image and facilitates the subsequent segmentation of water traces and background. In this paper, the water repellent image's watermark area is counted, and connected-domain wave processing is used to limit the area of water droplets retained, thereby improving the efficiency of filtering water droplets without having a big impact on the image as a whole. The problem of uneven illumination is an unavoidable problem in the field of image processing, and the resulting reflection problem brings difficulties to image processing. This article regards the reflective area of the watermark as a “hole” and uses the idea of “hole filling” to eliminate the reflective point, which weakens the reflection problem to a certain extent.

Selecting Workers Wisely for Crowdsourcing When Copiers and Domain Experts Co-exist

Crowdsourcing integrates human wisdom to solve problems. Tremendous research efforts have been made in this area. However, most of them assume that workers have the same credibility in different domains and workers complete tasks independently. This leads to an inaccurate evaluation of worker credibility, hampering crowdsourcing results. To consider the impact of worker domain expertise, we adopted a vector to more accurately measure the credibility of each worker. Based on this measurement and prior task domain knowledge, we calculated fine-grained worker credibility on each given task. To avoid tasks being assigned to dependent workers who copy answers from others, we conducted copier detection via Bayesian analysis. We designed a crowdsourcing system called SWWC composed of a task assignment stage and a truth discovery stage. In the task assignment stage, we assigned tasks wisely to workers based on worker domain expertise calculation and copier removal. In the truth discovery stage, we computed the estimated truth and worker credibility by an iterative method. Then, we updated the domain expertise of workers to facilitate the upcoming task assignment. We also designed initialization algorithms to better initialize the accuracy of new workers. Theoretical analysis and experimental results showed that our method had a prominent advantage, especially under a copying situation.