Inference Ability Research Articles

A Lightweight and Adaptive Knowledge Distillation Framework for Remaining Useful Life Prediction

For prognostics and health management of industrial systems, machine remaining useful life (RUL) prediction is an essential task. While deep learning-based methods have achieved great successes in RUL prediction tasks, large-scale neural networks are still difficult to deploy on edge devices owing to the constraints of memory capacity and computing power. In this paper, we propose a lightweight and adaptive knowledge distillation (KD) framework to alleviate this problem. Firstly, multiple teacher models are compressed into a student model through KD to improve the industrial prediction accuracy. Secondly, a dynamic exiting method is studied to enable an adaptive inference on the distilled student model. Finally, we develop a reparameterization scheme to further lessen the student network. Experiments on two turbofan engine degradation datasets and a bearing degradation dataset demonstrate that our method significantly outperforms the state-of-the-art KD methods and enables the distilled model with an adaptive inference ability.

Examining the relationship between English as a foreign language learners' cognitive abilities and L2 grit in predicting their writing performance

Research aimsThis study was designed to explore the interplay among English as a foreign language learners' cognitive abilities and grit in predicting writing performance along different task complexity through path analysis and multiple group analysis. MethodologyA total of 353 tertiary students from Western China were recruited to participate in this study. Their language aptitude and working memory capacity were evaluated by the LLAMA tests and the automated operation span task, respectively. They were required to complete writing tasks with different complexities: argumentation (complexity+) and narration (complexity-). Results and conclusionThe results revealed that (1) the effects of perseverance of effort and working memory on L2 writing performance were significant in the argumentative task, whereas the influences of perseverance of effort and consistency of interest on L2 writing performance were significant in the narrative one; (2) task complexity moderated the respective impacts of grammatical inferencing ability and working memory on L2 writing performance. Theoretical and pedagogical implications were discussed.

Branchy Deep Learning Based Real-Time Defect Detection Under Edge-Cloud Fusion Architecture

Many machine-learning-based defect detection methods, especially deep learning-based approaches, have high requirements on computing power and network. They lead to time delay, high cost, and energy consumption when applied to deal with the massive data in an autonomous manufacturing enterprise. So efficient detection in the end-edge-cloud architecture is a good solution to overcome the above challenges. A branchy deep learning detection model with early exit ability of inference is proposed, in which the main branch is deployed on the cloud server and the side branches are on edge equipment. The proposed method quickly and effectively detects the category and location of the defect in printed circuit boards since partial computing task is offloaded to the edge nodes. A prototype system is implemented based on a computer as the cloud server and a Raspberry Pi as an edge node in order to verify the feasibility of the proposed method. The experiment result manifests high detection accuracy and fast computing speed.

Risk prediction for petroleum exploration based on Bayesian network classifier

Effective prediction of petroleum exploration risks is critical for boosting oil and gas extraction efficiency and increasing economic benefits. To improve the accuracy and efficiency of risk prediction, this paper proposes a scalable k-dependence Bayesian network classifier (SKDB) method for solving the risk prediction problem of oil and gas exploration. The effectiveness and efficiency of the SKDB method were illustrated by taking the Jurassic Sangonghe Formation hydrocarbon exploration risk prediction in the Junggar Basin as an example. The model accuracy comparison results showed that SKDB has the highest accuracy compared with the state-of-the-art methods, and the accuracy was improved by 3.55∼7.63%. The oil and gas probability map prediction results indicated that the SKDB method not only showed high hydrocarbon probability in the reservoir areas, but also had better trend inference ability compared with other methods. Based on the SKDB prediction results, the remaining hydrocarbon favorable areas of the Sangonghe Formation reservoir were visualized and three types of favorable hydrocarbon distribution zones were preferentially selected, namely Type I (highly favorable), Type II (moderately favorable) and Type III (low favorable). The application results can provide a decision basis for the next oil and gas exploration and optimization strategy.

Social cognition and real-life functioning in patient samples with 22q11.2 deletion syndrome with or without psychosis, compared to a large sample of patients with schizophrenia only and healthy controls.

Patients with the 22q11.2 deletion syndrome (DS) show an increased risk of developing a psychotic illness lifetime. 22q11.2DS may represent a reliable model for studying the neurobiological underpinnings of schizophrenia. The study of social inference abilities in a genetic condition at high risk for psychosis, like 22q11.2DS, may shed light on the relationships between neurocognitive processes and patients' daily general functioning. The study sample consisted of 1736 participants, divided into four groups: 22q11.2DS patients with diagnosis of psychotic disorder (DEL SCZ, N = 20); 22q11.2DS subjects with no diagnosis of psychosis (DEL, N = 43); patients diagnosed with schizophrenia without 22q11.2DS (SCZ, N = 893); and healthy controls (HC, N = 780). Social cognition was assessed through The Awareness of Social Inference Test (TASIT) and general functioning through the Specific Levels of Functioning (SLoF) scale. We analysed data through regression analysis. The SCZ and DEL groups had similar levels of global functioning; they both had significantly lower SLoF Total scores than HC (p < .001); the DEL SCZ group showed significantly lower scores compared to the other groups (SCZ, p = .004; DEL, p = .003; HC, p < .001). A significant deficit in social cognition was observed in the three clinical groups. In the DEL SCZ and SCZ groups, TASIT scores significantly predicted global functioning (p < .05). Our findings of social cognition deficit in psychosis-prone patients point to the possible future adoption of rehabilitation programmes, like Social Skills Training and Cognitive Remediation, during premorbid stages of psychosis.

Talent recommendation based on attentive deep neural network and implicit relationships of resumes

Talent recruitment has become a crucial issue for companies since finding suitable candidates from the massive data on potential candidates from online talent platforms is a challenging task. However, extant studies mainly focus on the scalability and inference ability of models, while the dynamic variability of the importance of each feature in different scenarios is barely addressed. Besides, there is a lack of research on how to depict the hidden potential preference of the job which cannot be derived from job requirements. In this paper, we propose a two-stage resume recommendation model based on deep learning and attention mechanisms, especially considering the latent preference information hidden in the hired employee resumes, named the Attentive Implicit Relationship-Aware Neural Network (AIRANN) model. Specifically, a novel mechanism is proposed herein to extract the hidden potential preference of the corresponding job by deriving the implicit relationship between the target resume and hired employees’ resumes. Existing studies have not considered such resume implicit relationships. Moreover, we propose a Feature Co-Attention mechanism to capture the dynamic interactive importance within the non-text features of both resumes and jobs. For different jobs, the suitability of resumes would be valued from different aspects, including resume implicit relationships, as well as textual and non-textual features. Accordingly, an Aspect-attention mechanism is designed herein to automatically adjust the variant importance of each aspect. Finally, extensive experiments are conducted on a real-world company dataset. The experiment results of ablation studies demonstrate the effectiveness of each mechanism in the proposed AIRANN model. The experiment results also show that the proposed AIRANN model outperforms other baseline methods, showing a general improvement of 13.31%, 12.49%, 6.5% and 7.17% over the state-of-the-art baseline under F1@6, F1@15, NDCG@6 and NDCG@15, respectively.

추론적 이해 능력 향상을 위한 학습활동 구안 및 지도 방안

Purpose: Inferential comprehension plays a central role in the process of reading comprehension. In order to improve this inference ability, it is necessary to establish a teaching and learning plan thateffectively reflects the thinking base of inference, and to implement learning activities appropriately. Methods: In order to plan learning activities for improving inference comprehension ability, we graspedthe concept and types of inference comprehension, and the stylistic characteristics of newspaper articlesthat can be used as learning materials. Results: Combining stylistic elements such as headlines and subtitles of newspaper articles, specializations,texts, and types of reasoning, learning activities and teaching strategies for improving reasoningcomprehension ability were devised. Conclusion: Headlines and subtitles of newspaper articles are sufficient as clues for reasoning asteaching and learning materials for improving deductive comprehension ability. Learning activities usingthe headlines and subtitles of newspaper articles can be typed according to the direction of reasoning,the scope of reasoning, and the method of reasoning. Specifically, forward reasoning and backwardreasoning, core content reasoning and supporting content reasoning, background knowledge and articlecontent utilization reasoning, and other areas. can be classified as typing categories. Newspaper articles,in addition to their unique stylistic elements such as headlines and subtitles, are highly utilized interms of access to reading, interest and motivation as reading materials closely related to daily life. Therefore, it can be an efficient way to improve the reasoning comprehension ability by activelyutilizing this to plan learning activities and to construct teaching plans.

Exploration of Vehicle Target Detection Method Based on Lightweight YOLOv5 Fusion Background Modeling

Due to the explosive increase per capita in vehicle ownership in China brought about by the continuous development of the economy and society, many negative impacts have arisen, making it necessary to establish the smart city system that has rapidly developing vehicle detection technology as its data acquisition system. This paper proposes a lightweight detection model based on an improved version of YOLOv5 to address the problem of missed and false detections caused by occlusion during rush hour vehicle detection in surveillance videos. The proposed model replaces the BottleneckCSP structure with the Ghostnet structure and prunes the network model to speed up inference. Additionally, the Coordinate Attention Mechanism is introduced to enhance the network’s feature extraction and improve its detection and recognition ability. Distance-IoU Non-Maximum Suppression replaces Non-Maximum Suppression to address the issue of false detection and omission when detecting congested targets. Lastly, the combination of the five-frame differential method with VIBE and MD-SILBP operators is used to enhance the model’s feature extraction capabilities for vehicle contours. The experimental results show that the proposed model outperforms the original model in terms of the number of parameters, inference ability, and accuracy when applied to both the expanded UA-DETRAC and a self-built dataset. Thus, this method has significant industrial value in intelligent traffic systems and can effectively improve vehicle detection indicators in traffic monitoring scenarios.

Evaluating Deep Learning Techniques for Natural Language Inference

Natural language inference (NLI) is one of the most important natural language understanding (NLU) tasks. NLI expresses the ability to infer information during spoken or written communication. The NLI task concerns the determination of the entailment relation of a pair of sentences, called the premise and hypothesis. If the premise entails the hypothesis, the pair is labeled as an “entailment”. If the hypothesis contradicts the premise, the pair is labeled a “contradiction”, and if there is not enough information to infer a relationship, the pair is labeled as “neutral”. In this paper, we present experimentation results of using modern deep learning (DL) models, such as the pre-trained transformer BERT, as well as additional models that relay on LSTM networks, for the NLI task. We compare five DL models (and variations of them) on eight widely used NLI datasets. We trained and fine-tuned the hyperparameters for each model to achieve the best performance for each dataset, where we achieved some state-of-the-art results. Next, we examined the inference ability of the models on the BreakingNLI dataset, which evaluates the model’s ability to recognize lexical inferences. Finally, we tested the generalization power of our models across all the NLI datasets. The results of the study are quite interesting. In the first part of our experimentation, the results indicate the performance advantage of the pre-trained transformers BERT, RoBERTa, and ALBERT over other deep learning models. This became more evident when they were tested on the BreakingNLI dataset. We also see a pattern of improved performance when the larger models are used. However, ALBERT, given that it has 18 times fewer parameters, achieved quite remarkable performance.

On the Performance of Bayesian Generalized Dissimilarity Model Estimator

The Generalized Dissimilarity Model (GDM) is an extension of Generalized Linear Model (GLM) that is used to describe and estimate biological pairwise dissimilarities following a binomial process in response to environmental gradients. Some improvement has been made to accommodate the uncertainty quantity of GDM by applying resampling scheme such as Bayesian Bootstrap (BBGDM). Because there is an ecological assumption in the GDM, it is reasonable to use a proper Bayesian approach rather than resampling method to obtain better modelling and inference results. Similar to other GLM techniques, the GDM also employs a link function, such as the logit link function that is commonly used for the binomial regression model. By using this link, a Bayesian approach to GDM framework which called Bayesian GDM (BGDM) can be constructed. In this paper, we aim to evaluate the estimators' performance of Bayesian Generalized Dissimilarity Model (BGDM) in relative to BBGDM. Our study revealed that the performance of BGDM estimator outperformed that of BBGDM, especially in term of unbiasedness and efficiency. However, the BGDM estimators failed to meet consistency property. Moreover, the application of the BGDM to a real case study indicates that its inferential abilities are superior to the preceding model.

Dynamic Slimmable Denoising Network.

Recently, tremendous human-designed and automatically searched neural networks have been applied to image denoising. However, previous works intend to handle all noisy images in a pre-defined static network architecture, which inevitably leads to high computational complexity for good denoising quality. Here, we present a dynamic slimmable denoising network (DDS-Net), a general method to achieve good denoising quality with less computational complexity, via dynamically adjusting the channel configurations of networks at test time with respect to different noisy images. Our DDS-Net is empowered with the ability of dynamic inference by a dynamic gate, which can predictively adjust the channel configuration of networks with negligible extra computation cost. To ensure the performance of each candidate sub-network and the fairness of the dynamic gate, we propose a three-stage optimization scheme. In the first stage, we train a weight-shared slimmable super network. In the second stage, we evaluate the trained slimmable super network in an iterative way and progressively tailor the channel numbers of each layer with minimal denoising quality drop. By a single pass, we can obtain several sub-networks with good performance under different channel configurations. In the last stage, we identify easy and hard samples in an online way and train a dynamic gate to predictively select the corresponding sub-network with respect to different noisy images. Extensive experiments demonstrate our DDS-Net consistently outperforms the state-of-the-art individually trained static denoising networks.

Altered sensory integration from body and language development in children with autism spectrum disorder.

Although atypical sensory motor processing has been investigated in children with autism spectrum disorder (ASD), whether or not atypical sensory motor processing is related to altered language function in children with ASD remains unclear. This study examined the relationship between sensory motor processing and language conceptual inference ability in 3-10-year-old children with (n = 61) and without (n = 114) ASD. Language performance was assessed using the language conceptual inference task of the Kaufman Assessment Battery for Children (K-ABC). Sensory processing was assessed using the Caregiver Sensory Profile. In children with ASD, altered processing of the fine motor/perceptual factor scored by sensory profile was found to be significantly related to language conceptual inference ability in the K-ABC, representing the integrated abilities of language comprehension and language expression, which reflect language semantic concept formation. For children with ASD, the results suggest a relationship between difficulties of integrating sensory information perceived from the body adjusting fine movement and deficiencies of language semantic conceptual formation.

The Role of Metacognition (Metacomprehension) and Inferential Ability on Reading Comprehension Ability

In this study, the researchers explored the role of the metacognition component on reading comprehension ability and the role of reading comprehension ability in predicting the level of accuracy of metacomprehension. The first stage of research used experimental research to see the effect of inferential ability and metacognition on the reading comprehension ability. The second stage of research using the correlational method is used to study the relationship between the independent and dependent variables. This study involved 300 primary school students from 10 schools with a composition of 200 students involved in the first phase of research and 100 students involved in the second phase of research. Samples were taken randomly. The results show that the individual's reading comprehension ability can predict the level of metacomprehension accuracy. Metacognitive strategies that are carried out through planning, monitoring and evaluation have a significant impact on students' reading comprehension skills. Inferential-based questions have a significant impact on the accuracy of metacomprehension. In addition, the metacognitive component (metacomprehension), and inferential ability can improve students' reading comprehension skills, especially increasing inferential understanding. The implication of this research is that teachers can consider these aspects as well as optimise the role of these variables to enhance students' reading comprehension skills.

Data and knowledge co-driving for cancer subtype classification on multi-scale histopathological slides

Artificial intelligence-enabled histopathological data analysis has become a valuable assistant to the pathologist. However, existing models lack representation and inference abilities compared with those of pathologists, especially in cancer subtype diagnosis, which is unconvincing in clinical practice. For instance, pathologists typically observe the lesions of a slide from global to local, and then can give a diagnosis based on their knowledge and experience. In this paper, we propose a Data and Knowledge Co-driving (D&K) model to replicate the process of cancer subtype classification on a histopathological slide like a pathologist. Specifically, in the data-driven module, the bagging mechanism in ensemble learning is leveraged to integrate the histological features from various bags extracted by the embedding representation unit. Furthermore, a knowledge-driven module is established based on the Gestalt principle in psychology to build the three-dimensional (3D) expert knowledge space and map histological features into this space for metric. Then, the diagnosis can be made according to the Euclidean distance between them. Extensive experimental results on both public and in-house datasets demonstrate that the D&K model has a high performance and credible results compared with the state-of-the-art methods for diagnosing histopathological subtypes. Code: https://github.com/Dennis-YB/Data-and-Knowledge-Co-driving-for-Cancer-Subtypes-Classification.

Contributions of metalinguistic awareness and lexical inference to reading comprehension in upper‐elementary Chinese students

BackgroundThe present study aimed to examine and compare the relative contributions of three facets of metalinguistic awareness (phonological awareness, orthographic awareness and morphological awareness) and lexical inferencing ability to Chinese reading comprehension across upper elementary grades.MethodFifty‐three third graders, 79 fourth graders and 108 fifth graders were administered a battery of tests including non‐verbal intelligence, phonological awareness, morphological awareness, orthographic awareness, lexical inference and reading comprehension.ResultsThe multivariate analyses across grade groups revealed the continued progress in children's performances on morphological awareness but not on phonological and orthographic awareness. More important, hierarchical regressions exhibited the same pattern across grades, suggesting that morphological awareness was the sole consistent indicator of reading comprehension in all grades. Finally, lexical inferencing ability contributed to reading comprehension only in fifth‐grade students after accounting for the effects of metalinguistic awareness.ConclusionOur findings highlight the importance of morphological awareness as the most potent meaning‐building metalinguistic skill that can consistently predict Chinese reading comprehension in upper elementary children and indicate that the contribution of lexical inferencing ability differs as a function of children’s reading skills. Results were discussed in relation to the Reading Systems Framework and implications for educational instruction were provided.

Real-Time Phaseless Microwave Frequency-Diverse Imaging with Deep Prior Generative Neural Network

The millimeter-wave frequency-diverse imaging regime has recently received considerable attention in both the security screening and synthetic aperture radar imaging literature. Considering that the minor systematic errors and alignment errors could still produce heavily corrupted images, these complex-based imaging reconstructions rely heavily on the precise measurement of both phase and amplitude of radiation field patterns and echo signals. In the literature, it is shown that by leveraging phase-retrieval techniques, salient reconstruction images can still be acquired, even in the presence of significant phase errors, which could ease the phase error calibration pressure to a large extent in practical imaging applications. In this paper, in the regime of phaseless frequency-diverse imaging, with the powerful feature inference and generation power of unsupervised generative models, an end-to-end deep prior generative neural network is designed to achieve near real-time imaging. The harsh imaging reconstruction with both the high radiation mode correlations and extremely low scene compression sampling ratio, which are extremely troublesome to tackle for generally applied matched-filter and compressed sensing approach in the current frequency-diverse imaging literature, can still be preferably handled with our reconstruction network. The well-trained reconstruction network is constituted by prior inference and deep generative modules with excellent generative capabilities and significant prior inference abilities. Using simulation experiments with radiation field data, we verify that the integration of phase-free frequency-change imaging with deep learning networks can effectively improve reconstruction capabilities and improve robustness to systematic phase errors. Compared with existing imaging methods, our imaging method has high imaging performance and can even reconstruct targets under low compression ratio conditions, which is somewhat competitive with current state-of-the-art algorithms. Moreover, we find that the proposed method has good anti-noise and stability.

The dimensionality of morphological awareness in reading comprehension among Chinese early adolescent readers.

Recent research has included multiple measures of morphological awareness to address the dimensionality of morphological construct in the context of modeling skilled reading. However, a majority of studies have an Anglocentric focus. The current study aims to extend the previous studies to logographic learners by evaluating the dimensionality of morphological awareness in higher-order reading comprehension among Chinese adolescent readers. A total of 686 early adolescent students (339 fifth-grade students and 347 sixth-grade students) participated in the study. They completed a series of morphological awareness measurements (morpheme recognition, morpheme discrimination, and compound structure awareness), vocabulary knowledge, lexical inference and reading comprehension. By testing three alternative path models, the study showed that morphological awareness and vocabulary knowledge were best represented as parallel covariates in predicting Chinese reading comprehension. More important, the study highlighted the mediator of lexical inference in associating morphological awareness, vocabulary knowledge and reading comprehension among Chinese readers. Empirical findings suggest that morphological awareness and vocabulary knowledge seem to be under a unitary construct in logographic reading acquisition and that word-meaning inference ability connects the path between morphological awareness and reading comprehension. These findings contribute to the complexity in the conceptualization of Chinese morphological awareness and reading instruction by examining the ways in which the morphological construct supports higher-order reading development.

Analysis of Citizen Science-Based Flowering Plant Diversity Worksheet Development to Improve Students’ Critical Thinking Ability

The purpose of this study was to assess the feasibility of the developed Citizen Science-Based Flowering Plant Diversity Worksheet as well as the enhancement of students' critical thinking abilities. This was a pre-experiment with a pretest-posttest design with one group. Validator employed a four-scale validation sheet to assess the feasibility of the Citizen Science-Based Flowering Plant Diversity Worksheet based on competency, practical, knowledge construction, and the appearance of critical thinking ability. Five critical thinking ability indicators were tested: interpretation, analysis, explanation, evaluation, and inference. These indications were tested in 104 10th grade students through a critical thinking ability test before and after studying with the Citizen Science-Based Flowering Plant Diversity worksheet. Descriptive statistics were used to analyze the data. The Citizen Science-Based Flowering Plant Diversity Worksheet was found feasible (95.12%) and students' critical thinking abilities increased moderately (N-Gain = 0.56). The most notable development in critical thinking ability was in interpretation ability, while inference ability improved the least. As a result, the Citizen Science-Based Flowering Plant Diversity Worksheet may be inferred to be quite effective in improving students' critical thinking abilities.

Contributions of morphological awareness and lexical inferencing ability to L2 vocabulary knowledge among Chinese EFL learners: a structural equation modeling analysis

Abstract The present study investigated the direct and indirect contributions of morphological awareness to vocabulary knowledge among a group of Chinese English as a foreign language (EFL) students in a university in China. Using multiple regression analysis and structural equation modeling analysis, the study found that morphological awareness made significant contributions to vocabulary knowledge after lexical inferencing was controlled for. To be more specific, the prediction of morphological awareness on vocabulary depth was stronger than its prediction on vocabulary breadth. The direct effects of morphological awareness were stronger than its indirect effects on vocabulary. More importantly, the study revealed that the relationships between morphological awareness (morpheme recognition and morpheme discrimination) and vocabulary knowledge were mediated by discourse clues rather than grammar clues.

Spatiotemporal Feature Enhancement Aids the Driving Intention Inference of Intelligent Vehicles.

In order that fully self-driving vehicles can be realized, it is believed that systems where the driver shares control and authority with the intelligent vehicle offer the most effective solution. An understanding of driving intention is the key to building a collaborative autonomous driving system. In this study, the proposed method incorporates the spatiotemporal features of driver behavior and forward-facing traffic scenes through a feature extraction module; the joint representation was input into an inference module for obtaining driver intentions. The feature extraction module was a two-stream structure that was designed based on a deep three-dimensional convolutional neural network. To accommodate the differences in video data inside and outside the cab, the two-stream network consists of a slow pathway that processes the driver behavior data with low frame rates, along with a fast pathway that processes traffic scene data with high frame rates. Then, a gated recurrent unit, based on a recurrent neural network, and a fully connected layer constitute an intent inference module to estimate the driver’s lane-change and turning intentions. A public dataset, Brain4Cars, was used to validate the proposed method. The results showed that compared with modeling using the data related to driver behaviors, the ability of intention inference is significantly improved after integrating traffic scene information. The overall accuracy of the intention inference of five intents was 84.92% at a time of 1 s prior to the maneuver, indicating that making full use of traffic scene information was an effective way to improve inference performance.