Foundation Task Research Articles

Metrological parameter planning method based on a multi-head sparse graph attention network for airborne products

The airborne system plays a crucial role in upholding the operational capabilities of an aircraft, and the caliber of its products significantly influences the overall reliability and safety of the aircraft. The foundational task in guaranteeing product quality is metrology, wherein metrological parameters play a pivotal role in shaping the ultimate product quality and determining metrological efficiency. Because the metrology documents are scattered in various departments and the data are highly isolated, manual planning methods are highly subjective, resulting in error-prone and inefficient results. The existing parameter planning methods based on deep learning suffer from the problems of strong training data dependency and poor model interpretability. For this reason, this paper proposes a metrological parameter planning method based on a multi-head sparse graph attention network (MHSGAT) for airborne products. First, an attribute graph structure is designed to represent product metrology information, and a metrology parameter graph dataset is constructed based on historical documents. Then, the proposed MHSGAT assigns sparse attention coefficients to the neighbors of nodes for feature aggregation. Finally, the metrology parameter planning results are output in the dense layer. The experimental results show that the proposed method outperforms the other four baseline methods. The significance of the proposed method in enhancing the metrological accuracy of airborne products and ensuring their quality is demonstrated by application cases.

AAPM task group report 135.B: Quality assurance for robotic radiosurgery.

AAPM Task Group Report 135.B covers new technology components that have been added to an established radiosurgery platform and updates the components that were not well covered in the previous report. Considering the current state of the platform, this task group (TG) is a combination of a foundational task group to establish the basis for new processes/technology and an educational task group updating guidelines on the established components of the platform. Because the technology discussed in this document has a relatively small user base compared to C-arm isocentric linacs, the authors chose to emphasize the educational components to assist medical physicists who are new to the technology and have not had the opportunity to receive in-depth vendor training at the time of reading this report. The TG has developed codes of practice, introduced QA, and developed guidelines which are generally expected to become enduring practice. This report makes prescriptive recommendations as there has not been enough longitudinal experience with some of the new technical components to develop a data-based risk analysis.

A Depth Awareness and Learnable Feature Fusion Network for Enhanced Geometric Perception in Semantic Correspondence.

Deep learning is becoming the most widely used technology for multi-sensor data fusion. Semantic correspondence has recently emerged as a foundational task, enabling a range of downstream applications, such as style or appearance transfer, robot manipulation, and pose estimation, through its ability to provide robust correspondence in RGB images with semantic information. However, current representations generated by self-supervised learning and generative models are often limited in their ability to capture and understand the geometric structure of objects, which is significant for matching the correct details in applications of semantic correspondence. Furthermore, efficiently fusing these two types of features presents an interesting challenge. Achieving harmonious integration of these features is crucial for improving the expressive power of models in various tasks. To tackle these issues, our key idea is to integrate depth information from depth estimation or depth sensors into feature maps and leverage learnable weights for feature fusion. First, depth information is used to model pixel-wise depth distributions, assigning relative depth weights to feature maps for perceiving an object's structural information. Then, based on a contrastive learning optimization objective, a series of weights are optimized to leverage feature maps from self-supervised learning and generative models. Depth features are naturally embedded into feature maps, guiding the network to learn geometric structure information about objects and alleviating depth ambiguity issues. Experiments on the SPair-71K and AP-10K datasets show that the proposed method achieves scores of 81.8 and 83.3 on the percentage of correct keypoints (PCK) at the 0.1 level, respectively. Our approach not only demonstrates significant advantages in experimental results but also introduces the depth awareness module and a learnable feature fusion module, which enhances the understanding of object structures through depth information and fully utilizes features from various pre-trained models, offering new possibilities for the application of deep learning in RGB and depth data fusion technologies. We will also continue to focus on accelerating model inference and optimizing model lightweighting, enabling our model to operate at a faster speed.

Subsampling Suffices for Adaptive Data Analysis

Ensuring that analyses performed on a dataset are representative of the entire population is one of the central problems in statistics. Most classical techniques assume that the dataset is independent of the analyst’s query and break down in the common setting where a dataset is reused for multiple, adaptively chosen, queries. This problem of adaptive data analysis was formalized in the seminal works of Dwork et al. (STOC, 2015) and Hardt and Ullman (FOCS, 2014). We identify a remarkably simple set of assumptions under which the queries will continue to be representative even when chosen adaptively: The only requirements are that each query takes as input a random subsample and outputs few bits. This result shows that the noise inherent in subsampling is sufficient to guarantee that query responses generalize. The simplicity of this subsampling-based framework allows it to model a variety of real-world scenarios not covered by prior work. In addition to its simplicity, we demonstrate the utility of this framework by designing mechanisms for two foundational tasks, statistical queries and median finding. In particular, our mechanism for answering the broadly applicable class of statistical queries is both extremely simple and state of the art in many parameter regimes.

Potency–Magnitude Scaling Relations and a Unified Earthquake Catalog for the Western United States

Abstract Quantifying the size of earthquakes is a foundational task in seismology, and over the years several magnitude scales have been developed. Of these, only scales based on seismic moment or potency can properly characterize changes in event size without saturation. Here, we develop empirical potency–magnitude scaling relations for earthquakes in the western United States, allowing us to translate instrumental magnitude estimates into uniform measures of earthquake size. We use synthetic waveforms to validate the observed scaling relations and to provide additional insight into the differences between instrumental and physics-based magnitude scales. Each earthquake in our catalog is assigned a clustering designation distinguishing mainshocks from triggered seismicity, along with a potency-based magnitude estimate that is comparable to moment magnitude and that can be easily converted into other magnitude scales as needed. The developed catalog and associated scaling relations have broad applications for fundamental and applied studies of earthquake processes and hazards.

A Benchmark for Morphological Segmentation in Uyghur and Kazakh

Morphological segmentation and stemming are foundational tasks in natural language processing. They have become effective ways to alleviate data sparsity in agglutinative languages because of the nature of agglutinative language word formation. Uyghur and Kazakh, as typical agglutinative languages, have made significant progress in morphological segmentation and stemming in recent years. However, the evaluation metrics used in previous work are character-level based, which may not comprehensively reflect the performance of models in morphological segmentation or stemming. Moreover, existing methods avoid manual feature extraction, but the model’s ability to learn features is inadequate in complex scenarios, and the correlation between different features has not been considered. Consequently, these models lack representation in complex contexts, affecting their effective generalization in practical scenarios. To address these issues, this paper redefines the morphological-level evaluation metrics: F1-score and accuracy (ACC) for morphological segmentation and stemming tasks. In addition, two models are proposed for morpheme segmentation and stem extraction tasks: supervised model and unsupervised model. The supervised model learns character and contextual features simultaneously, then feature embeddings are input into a Transformer encoder to study the correlation between character and context embeddings. The last layer of the model uses a CRF or softmax layer to determine morphological boundaries. In unsupervised learning, an encoder–decoder structure introduces n-gram correlation assumptions and masked attention mechanisms, enhancing the correlation between characters within n-grams and reducing the impact of characters outside n-grams on boundaries. Finally, comprehensive comparative analyses of the performance of different models are conducted from various points of view. Experimental results demonstrate that: (1) The proposed evaluation method effectively reflects the differences in morphological segmentation and stemming for Uyghur and Kazakh; (2) Learning different features and their correlation can enhance the model’s generalization ability in complex contexts. The proposed models achieve state-of-the-art performance on Uyghur and Kazakh datasets.

Snow Disaster Risk Assessment Based on Long-Term Remote Sensing Data: A Case Study of the Qinghai–Tibet Plateau Region in Xizang

The risk analysis and assessment of snow disasters are essential foundational tasks in natural disaster management and profoundly impact the scientific and precise formulation of disaster prevention, preparedness, and mitigation strategies. Employing the theory and methodology of snow disaster assessment, this research focuses on historical and potential snow disasters in the Qinghai–Tibetan Plateau (QTP) Region. Utilizing a long-time-series snow depth remote sensing dataset, we extracted six assessment indicators for historical snow disaster risk factors and potential snow disaster risk factors. We determined the weights of these six assessment indicators using the entropy weight method. Subsequently, we established a snow disaster assessment model to evaluate the grade distribution of snow disasters in the study area. This method can effectively solve the problem of the sparse data distribution of meteorological stations and reflect degrees of snow disaster risk on a large spatial scale. The findings reveal that areas with a relatively high snow disaster risk are primarily concentrated in the western part of the Ali Region, the central part of Chamdo, and near the border in Southern Xizang. Additionally, regions with a high frequency of snow disasters are predominantly located at the junction of Nagchu, Chamdo, and Nyingchi in the eastern part of Xizang. These results contribute valuable insights into the risk assessment of snow disasters and facilitate the development of effective strategies for disaster management in the region.

Evaluating the Effectiveness of Deep Learning Models for Foundational Program Analysis Tasks

While deep neural networks provide state-of-the-art solutions to a wide range of programming language tasks, their effectiveness in dealing with foundational program analysis tasks remains under explored. In this paper, we present an empirical study that evaluates four prominent models of code (i.e., CuBERT, CodeBERT, GGNN, and Graph Sandwiches) in two such foundational tasks: (1) alias prediction, in which models predict whether two pointers must alias, may alias or must not alias; and (2) equivalence prediction, in which models predict whether or not two programs are semantically equivalent. At the core of this study is CodeSem, a dataset built upon the source code of real-world flagship software (e.g., Linux Kernel, GCC, MySQL) and manually validated for the two prediction tasks. Results show that all models are accurate in both prediction tasks, especially CuBERT with an accuracy of 89% and 84% in alias prediction and equivalence prediction, respectively. We also conduct a comprehensive, in-depth analysis of the results of all models in both tasks, concluding that deep learning models are generally capable of performing foundational tasks in program analysis even though in specific cases their weaknesses are also evident. Our code and evaluation data are publicly available at https://github.com/CodeSemDataset/CodeSem.

Detecting oracle bone inscriptions via pseudo-category labels

Oracle bone inscriptions (OBIs) are the earliest Chinese characters and reserve abundant historical information. OBIs are detected by locating their positions in digital images. This has been a foundational task in modern archeological studies. Due to the development of deep neural networks(DNNs) in computer vision, detecting OBIs can be implemented by a more concise method instead of designing complicated hand-crafted features. However, existing models cannot perform well when noise areas are similar to some inscriptions. In this work, we present a simple but effective pseudo-label-based architecture for OBIs detection. Different from previous approaches, our method performs OBIs detection with the employment of information from multilabel annotations rather than single location information. We append a plug-and-play module that predicts the pseudo-label of an inscription after the backbone network for learning the particular structure prior to each inscription and brings this information to the backbone network by means of feature fusion. We make remarkable improvements on different backbone networks when using the proposed method on an OBIs detection dataset. The quantitative and qualitative results show that the proposed model can detect OBIs well and is an effective tool for assisting in the discovery and recognition of ancient writing.

Coupling speckle noise suppression with image classification for deep-learning-aided ultrasound diagnosis

Objective. During deep-learning-aided (DL-aided) ultrasound (US) diagnosis, US image classification is a foundational task. Due to the existence of serious speckle noise in US images, the performance of DL models may be degraded. Pre-denoising US images before their use in DL models is usually a logical choice. However, our investigation suggests that pre-speckle-denoising is not consistently advantageous. Furthermore, due to the decoupling of speckle denoising from the subsequent DL classification, investing intensive time in parameter tuning is inevitable to attain the optimal denoising parameters for various datasets and DL models. Pre-denoising will also add extra complexity to the classification task and make it no longer end-to-end. Approach. In this work, we propose a multi-scale high-frequency-based feature augmentation (MSHFFA) module that couples feature augmentation and speckle noise suppression with specific DL models, preserving an end-to-end fashion. In MSHFFA, the input US image is first decomposed to multi-scale low-frequency and high-frequency components (LFC and HFC) with discrete wavelet transform. Then, multi-scale augmentation maps are obtained by computing the correlation between LFC and HFC. Last, the original DL model features are augmented with multi-scale augmentation maps. Main results. On two public US datasets, all six renowned DL models exhibited enhanced F1-scores compared with their original versions (by 1.31%–8.17% on the POCUS dataset and 0.46%–3.89% on the BLU dataset) after using the MSHFFA module, with only approximately 1% increase in model parameter count. Significance. The proposed MSHFFA has broad applicability and commendable efficiency and thus can be used to enhance the performance of DL-aided US diagnosis. The codes are available at https://github.com/ResonWang/MSHFFA.

Part-Aware Point Cloud Completion through Multi-Modal Part Segmentation.

Point cloud completion aims to generate high-resolution point clouds using incomplete point clouds as input and is the foundational task for many 3D visual applications. However, most existing methods suffer from issues related to rough localized structures. In this paper, we attribute these problems to the lack of attention to local details in the global optimization methods used for the task. Thus, we propose a new model, called PA-NET, to guide the network to pay more attention to local structures. Specifically, we first use textual embedding to assist in training a robust point assignment network, enabling the transformation of global optimization into the co-optimization of local and global aspects. Then, we design a novel plug-in module using the assignment network and introduce a new loss function to guide the network's attention towards local structures. Numerous experiments were conducted, and the quantitative results demonstrate that our method achieves novel performance on different datasets. Additionally, the visualization results show that our method efficiently resolves the issue of poor local structures in the generated point cloud.

Enhancing Predictive Expert Method for Link Prediction in Heterogeneous Information Social Networks

Link prediction is a critical prerequisite and foundation task for social network security that involves predicting the potential relationship between nodes within a network or graph. Although the existing methods show promising performance, they often ignore the unique attributes of each link type and the impact of diverse node differences on network topology when dealing with heterogeneous information networks (HINs), resulting in inaccurate predictions of unobserved links. To overcome this hurdle, we propose the Enhancing Predictive Expert Method (EPEM), a comprehensive framework that includes an individual feature projector, a predictive expert constructor, and a trustworthiness investor. The individual feature projector extracts the distinct characteristics associated with each link type, eliminating shared attributes that are common across all links. The predictive expert constructor then creates enhancing predictive experts, which improve predictive precision by incorporating the individual feature representations unique to each node category. Finally, the trustworthiness investor evaluates the reliability of each enhancing predictive expert and adjusts their contributions to the prediction outcomes accordingly. Our empirical evaluations on three diverse heterogeneous social network datasets demonstrate the effectiveness of EPEM in forecasting unobserved links, outperforming the state-of-the-art methods.

CFSE: a Chinese short text classification method based on character frequency sub-word enhancement

As a foundation task of natural language processing, text classification is widely used in information retrieval, public opinion analysis, and other related tasks. Facing the problem of sparse features of Chinese short texts, which affects the classification accuracy of Chinese short texts, this paper proposes a Chinese short text classification method based on the Character Frequency Sub-word Enhancement (CFSE), which can effectively improve the classification accuracy of Chinese short texts. First, the initial Chinese-character sequence is mapped to the corresponding Character Frequency Sub-word (CFS) sequence based on the global character1 frequency information. Second, the relationship features among data are extracted based on BiLSTM-Att processing CFS sequence, and the semantic features of the initial Chinese-character sequence are obtained through ERNIE. Finally, these two kinds of features are fused and input into the text classifier to obtain the classification results. Experimental results show that the proposed method can improve the classification accuracy of Chinese short texts.

Prompt-based for Low-Resource Tibetan Text Classification

Text classification is a critical and foundational task in Tibetan natural language processing, it plays a crucial role in various applications, such as sentiment analysis and information extraction. However, the limited availability of annotated data poses a significant challenge to Tibetan natural language processing. This paper proposes a prompt learning-based method for low-resource Tibetan text classification to overcome this challenge. This method utilizes pre-trained language models to learn text representation and generation capabilities on a large-scale unsupervised Tibetan corpus, enabling few-shot Tibetan text classification. Experimental results demonstrate that the proposed method significantly improves the performance of Tibetan text classification in low-resource scenarios. This work provides a new research idea and method for low-resource language processing, such as Tibetan natural language processing. Hopefully, it will inspire subsequent work on low-resource language processing.

Family Dysfunction, Parental Role Abdication and High Rate of Kidnapping among Young People in Enugu State, Nigeria. (Psychological Implications)

The family is one of the most fundamental social unit of the society. It consists of group of individuals united by blood, marriage, adoption and other intimate ties. Being regarded as the foundation of every human being, any form of negative treatment, abnormality and confusion emanating from it poses a great threat and catastrophe to the young people and society at large. Therefore, this study sought to investigate family dysfunction, role abdication and high rate of kidnapping in Enugu State, Nigeria. Two purpose of study and hypotheses guided the study accordingly. The study was delimited to Enugu South local government area of Enugu State. Correlational survey design was used as well as a sample study of 300 young people purposively selected using accidental sampling. The instrument was researcher structured questionnaire adapted from The Questionnaire of Family Functioning (2008) and Dyadic Parent-Adolescent-Role confusion scale (DPARCS, 2018) face validated by three experts from Educational Psychology, Sociology and Measurement and Evaluation from Enugu State University of Science and Technology. Reliability co-efficient of 0.76 was determined using Cronbach Alpha. Data was collected with the help of two research assistant and hypotheses analyzed using Pearson Product Moment Correlation. Based on the findings, it was recommended amongst others, that families should rise up to its foundational task of unconditional love as well as assumption of its responsibilities to family members in order to groom youths that are well equipped emotionally and psychologically.

Cross-Well Lithology Identification Based on Wavelet Transform and Adversarial Learning

For geological analysis tasks such as reservoir characterization and petroleum exploration, lithology identification is a crucial and foundational task. The logging lithology identification tasks at this stage generally build a lithology identification model, assuming that the logging data share an independent and identical distribution. This assumption, however, does not hold among various wells due to the variations in depositional conditions, logging apparatus, etc. In addition, the current lithology identification model does not fully integrate the geological knowledge, meaning that the model is not geologically reliable and easy to interpret. Therefore, we propose a cross-domain lithology identification method that incorporates geological information and domain adaptation. This method consists of designing a named UAFN structure to better extract the semantic (depth) features of logging curves, introducing geological information via wavelet transform to improve the model’s interpretability, and using dynamic adversarial domain adaptation to solve the data-drift issue cross-wells. The experimental results show that, by combining the geological information in wavelet coefficients with semantic information, more lithological features can be extracted in the logging curve. Moreover, the model performance is further improved by dynamic domain adaptation and wavelet transform. The addition of wavelet transform improved the model performance by an average of 6.25%, indicating the value of the stratigraphic information contained in the wavelet coefficients for lithology prediction.

The Fire “Once Kindled on Moriah”: Wojtyła and Newman on the Conditions for the Proposal of Christ

The Fire “Once Kindled on Moriah”: Wojtyła and Newman on the Conditions for the Proposal of Christ1 Stephen Morgan (bio) The turn to the subject makes demands upon the transmission of the Christian faith that prior, largely objectivist approaches do not meet. In the personalism encountered in the work of John Henry Newman (1801–1890) and Karol Wojtyła (1920–2005) a balance between objective truth and personal experience can be found, which answers present need. This article considers two short texts from these eminent thinkers that indicate their views of this balance; each man would work out and develop more fully these views in later works. In particular, this article identifies within these shorter works a basis for the church’s foundational task of evangelization. First, this article will briefly consider the effects of the subjective turn in Western thought to the task of proposing Jesus Christ in the present age, before suggesting that some of the difficulties this turn has presented can be resolved by looking at the work of John Henry Newman and Karol Wojtyła/Pope John Paul II. It will consider two shorter works, one by each theologian: first, John Henry Newman’s fifth Oxford University Sermon, “Personal Influence, the Means of Propagating the Faith,”2 and, secondly, Karol Wojytła’s final published essay before his election to the papacy as Pope John Paul II, “Subjectivity and the Irreducible in Man.”3 The limitations of dealing with these two works—and their suitability to [End Page 23] stand as representative of the more developed treatment given by both men as to the question in other works—will be addressed before moving on to a detailed consideration of approach proposed by each theologian in turn. Finally, in concluding remarks, this article will consider the extent to which what each proposes may be seen as complimentary to the perspectives of the other. I. MORIAH, CATHOLIC FAITH, AND THE TURN TO THE SUBJECT The biblical notion of Mount Moriah is one on which both Newman and Wojtyła reflected. The Old Testament text makes mention of Mount Moriah in two places: the location where Abraham is commanded to go to sacrifice Isaac (Gen. 22:2) and as the site where Solomon constructed his Temple (2 Chron. 3:1). Modern biblical scholars disagree not only about the location, but even about the etymology of the name.4 What is not in doubt, however, is the symbolic meaning attached to the use of the name by John Henry Newman in his fifth University Sermon.5 Newman, clearly referencing the account in Genesis, sees Mount Moriah as the location of the first act of faith in the history of salvation. That is, as the place where Abraham, having been prepared to sacrifice Isaac and having kindled a fire to complete that offering, is provided by God with the ram caught in the thicket of thorns as the sacrificial victim in place of Isaac. It is this ram that becomes the sacrifice that ratifies the Covenant of Abraham’s faith, and it is this faith—faith in the living God—that Newman sees as the object of propagation, for which he posits personal influence as the appropriate means of its transmission. It should be noted that the significance of Moriah clearly occurred to Wojtyła, who treats it in poetic form (though no less theologically nor convincingly), in his Roman Trip-tych: Meditations.6 In a powerful series of short sections of the poem, entitled “III. A Hill in the Land of Moriah,” he contemplated the significance of the episode in Genesis, ending the third sub-section, “A conversation between a father and son in the land of Moriah,” with the recognition that Abraham’s act of faith in that place anticipates and prophesies its own consummation in the Paschal Mystery: [End Page 24] Henceforth the hill in the land of Moriah will be expectation— for on its heights the mystery will find fulfillment.7 In the ages when Christianity flourished largely within the context of what came to be called Christendom, where the vast majority of the faithful were baptized as infants and brought up in societies...

CAT-CPI: Combining CNN and transformer to learn compound image features for predicting compound-protein interactions.

Compound-protein interaction (CPI) prediction is a foundational task for drug discovery, which process is time-consuming and costly. The effectiveness of CPI prediction can be greatly improved using deep learning methods to accelerate drug development. Large number of recent research results in the field of computer vision, especially in deep learning, have proved that the position, geometry, spatial structure and other features of objects in an image can be well characterized. We propose a novel molecular image-based model named CAT-CPI (combining CNN and transformer to predict CPI) for CPI task. We use Convolution Neural Network (CNN) to learn local features of molecular images and then use transformer encoder to capture the semantic relationships of these features. To extract protein sequence feature, we propose to use a k-gram based method and obtain the semantic relationships of sub-sequences by transformer encoder. In addition, we build a Feature Relearning (FR) module to learn interaction features of compounds and proteins. We evaluated CAT-CPI on three benchmark datasets—Human, Celegans, and Davis—and the experimental results demonstrate that CAT-CPI presents competitive performance against state-of-the-art predictors. In addition, we carry out Drug-Drug Interaction (DDI) experiments to verify the strong potential of the methods based on molecular images and FR module.

IDA-Net: Intensity-distribution aware networks for semantic segmentation of 3D MLS point clouds in indoor corridor environments

• An intensity-distribution aware (IDA) descriptor to mine the intensity information. • A prior mask to describe the intensity-distribution pattern. • A two-stage embedding strategy to improve the performance of fusing different information. Semantic segmentation of 3D mobile laser scanning point clouds is the foundational task for scene understanding in several fields. Most existing segmentation methods tend to simply stack the common point attributes, such as the coordinates and intensity, but ignore their heterogeneous. This paper presents IDA-Net, an intensity-distribution aware network that mines the uniqueness and discrepancy of these two modalities in a separate way for point cloud segmentation under indoor corridor environments. Specifically, IDA-Net consists of two key components. Firstly, an intensity-distribution aware (IDA) descriptor is proposed to mine the intensity distribution pattern. It outputs a multi-channel mask for each point to represent the intensity distribution information. Secondly, a two-stage embedding network is designed to fuse the coordinates and intensity information efficiently. It includes a guiding operation in training stage and a refining operation in testing stage. IDA-Net was evaluated on two indoor corridor areas. Experimental results show that the proposed method significantly improves the performance of segmentation. Specifically, with backbone of KPConv, IDA-Net achieves high mIoU of 90.58% and 88.94% on the above two testing areas respectively, which demonstrates the superiority of the designed IDA descriptor and two-stage embedding network.

4G/5G Cell-level multi-indicator forecasting based on dense-MLP

With the development of 4G/5G, the rapid growth of traffic has caused a large number of cell indicators to exceed the warning threshold, and network quality has deteriorated. It is necessary for operators to solve the congestion in advance and effectively to guarantee the quality of user experience. Cell-level multi-indicator forecasting is the foundation task for proactive complex network optimization. In this paper, we propose the 4G/5G Cell-level multi-indicator forecasting method based on the dense-Multi-Layer Perceptron (MLP) neural network, which adds additional fully-connected layers between non-adjacent layers in an MLP network. The model forecasted the following week's traffic indicators of 13000 cells according to the six-month historical indicators of 65000 cells in the 4G&5G network, which got the highest weighted MAPE score (0.2484) in the China Mobile problem statement in the ITU-T AI/ML in 5G Challenge 2021. Furthermore, the proposed model has been integrated into the AsiaInfo 4G/5G energy-saving system and deployed in Jiangsu Province of China.