Project Homepage Research Articles

APMG: 3D Molecule Generation Driven by Atomic Chemical Properties.

Recently, mask-fill-based 3D Molecular Generation (MG) methods have become very popular in virtual drug design. However, the existing MG methods ignore the chemical properties of atoms and contain inappropriate atomic position training data, which limits their generation capability. To mitigate the above issues, this paper presents a novel mask-fill-based 3D molecule generation model driven by atomic chemical properties (APMG). Specifically, we construct a new attention-MPNN-based encoder and introduce the electronic information into atom representations to enrich chemical properties. Also, a multi-functional classifier is designed to predict the electronic information of each generated atom, guiding the type prediction of elements and bonds. By design, the proposed method uses the chemical properties of atoms and their correlations for high-quality molecule generation. Second, to optimize the atomic position training data, we propose a novel atomic training position generation approach using the Chi-Square distribution. We evaluate our APMG method on the CrossDocked dataset and visualize the docking states of the pockets and generated molecules. The obtained results demonstrate the superiority and merits of APMG over the state-of-the-art approaches. The dataset and codes will be available on the project homepage: https://github.com/JU-HuaY/APMG.

MMDG-DTI: Drug–target interaction prediction via multimodal feature fusion and domain generalization

Recently, deep learning has become the essential methodology for Drug–Target Interaction (DTI) prediction. However, the existing learning-based representation methods embed the prior knowledge encapsulated by the training data and, as such, acquire redundant domain information irrelevant to the DTI prediction, compromising the generalization capability of a deep network to unfamiliar instances. To tackle this limitation, we propose a novel DTI prediction method using Multi-Modal feature fusion and Domain Generalization (MMDG-DTI). Specifically, we first use pre-trained Large Language Models (LLMs) to obtain generalized textual features covering nearly the whole biological text vocabulary, with the capacity to handle unseen samples and extract robust and discriminative features. In parallel, we propose a unified structural feature extractor based on a hybrid Graph Neural Network (GNN). The extractor improves the performance of MMDG-DTI by extracting features from another modality and discarding the representation of domain-specific prior substructures. Then, we integrate the complementary information of LLM and GNN features using an innovative classifier, while enhancing the generalization ability with the help of Domain Adversarial Training (DAT) and contrastive learning. Last, we propose a novel cross-domain evaluation protocol to enhance the predication fidelity in practical applications. The quantitative and visualization results obtained on several benchmarking datasets demonstrate that MMDG-DTI achieves state-of-the-art performance under both single-domain and cross-domain settings, confirming the superiority of the proposed method. The datasets and codes will be available on the project homepage: https://github.com/JU-HuaY/MMDG-DTI.

EUKARYOME: the rRNA gene reference database for identification of all eukaryotes.

Molecular identification of micro- and macroorganisms based on nuclear markers has revolutionized our understanding of their taxonomy, phylogeny and ecology. Today, research on the diversity of eukaryotes in global ecosystems heavily relies on nuclear ribosomal RNA (rRNA) markers. Here, we present the research community-curated reference database EUKARYOME for nuclear ribosomal 18S rRNA, internal transcribed spacer (ITS) and 28S rRNA markers for all eukaryotes, including metazoans (animals), protists, fungi and plants. It is particularly useful for the identification of arbuscular mycorrhizal fungi as it bridges the four commonly used molecular markers-ITS1, ITS2, 18S V4-V5 and 28S D1-D2 subregions. The key benefits of this database over other annotated reference sequence databases are that it is not restricted to certain taxonomic groups and it includes all rRNA markers. EUKARYOME also offers a number of reference long-read sequences that are derived from (meta)genomic and (meta)barcoding-a unique feature that can be used for taxonomic identification and chimera control of third-generation, long-read, high-throughput sequencing data. Taxonomic assignments of rRNA genes in the database are verified based on phylogenetic approaches. The reference datasets are available in multiple formats from the project homepage, http://www.eukaryome.org.

Particle Swarm Optimization Algorithm Using Velocity Pausing and Adaptive Strategy

Particle swarm optimization (PSO) as a swarm intelligence-based optimization algorithm has been widely applied to solve various real-world optimization problems. However, traditional PSO algorithms encounter issues such as premature convergence and an imbalance between global exploration and local exploitation capabilities when dealing with complex optimization tasks. To address these shortcomings, an enhanced PSO algorithm incorporating velocity pausing and adaptive strategies is proposed. By leveraging the search characteristics of velocity pausing and the terminal replacement mechanism, the problem of premature convergence inherent in standard PSO algorithms is mitigated. The algorithm further refines and controls the search space of the particle swarm through time-varying inertia coefficients, symmetric cooperative swarms concepts, and adaptive strategies, balancing global search and local exploitation. The performance of VASPSO was validated on 29 standard functions from Cec2017, comparing it against five PSO variants and seven swarm intelligence algorithms. Experimental results demonstrate that VASPSO exhibits considerable competitiveness when compared with 12 algorithms. The relevant code can be found on our project homepage.

Segmenting medical images with limited data

While computer vision has proven valuable for medical image segmentation, its application faces challenges such as limited dataset sizes and the complexity of effectively leveraging unlabeled images. To address these challenges, we present a novel semi-supervised, consistency-based approach termed the data-efficient medical segmenter (DEMS). The DEMS features an encoder–decoder architecture and incorporates the developed online automatic augmenter (OAA) and residual robustness enhancement (RRE) blocks. The OAA augments input data with various image transformations, thereby diversifying the dataset to improve the generalization ability. The RRE enriches feature diversity and introduces perturbations to create varied inputs for different decoders, thereby providing enhanced variability. Moreover, we introduce a sensitive loss to further enhance consistency across different decoders and stabilize the training process. Extensive experimental results on both our own and three public datasets affirm the effectiveness of DEMS. Under extreme data shortage scenarios, our DEMS achieves 16.85% and 10.37% improvement in dice score compared with the U-Net and top-performed state-of-the-art method, respectively. Given its superior data efficiency, DEMS could present significant advancements in medical segmentation under small data regimes. The project homepage can be accessed at https://github.com/NUS-Tim/DEMS.

Implementation of an antimicrobial stewardship program for urinary tract infections in long-term care facilities: a cluster-controlled intervention study

BackgroundWidespread inappropriate use of antimicrobial substances drives resistance development worldwide. In long-term care facilities (LTCF), antibiotics are among the most frequently prescribed medications. More than one third of antimicrobial agents prescribed in LTCFs are for urinary tract infections (UTI). We aimed to increase the number of appropriate antimicrobial treatments for UTIs in LTCFs using a multi-faceted antimicrobial stewardship intervention.MethodsWe performed a non-randomized cluster-controlled intervention study. Four LTCFs of the Geriatric Health Centers Graz were the intervention group, four LTCFs served as control group. The main components of the intervention were: voluntary continuing medical education for primary care physicians, distribution of a written guideline, implementation of the project homepage to distribute guidelines and videos and onsite training for nursing staff. Local nursing staff recorded data on UTI episodes in an online case report platform. Two blinded reviewers assessed whether treatments were adequate.Results326 UTI episodes were recorded, 161 in the intervention group and 165 in the control group. During the intervention period, risk ratio for inadequate indication for treatment was 0.41 (95% CI 0.19–0.90), p = 0.025. In theintervention group, the proportion of adequate antibiotic choices increased from 42.1% in the pre-intervention period, to 45.9% during the intervention and to 51% in the post-intervention period (absolute increase of 8.9%). In the control group, the proportion was 36.4%, 33.3% and 33.3%, respectively. The numerical difference between intervention group and control group in the post-intervention period was 17.7% (difference did not reach statistical significance). There were no significant differences between the control group and intervention group in the safety outcomes (proportion of clinical failure, number of hospital admissions due to UTI and adverse events due to antimicrobial treatment).ConclusionsAn antimicrobial stewardship program consisting of practice guidelines, local and web-based education for nursing staff and general practitioners resulted in a significant increase in adequate treatments (in terms of decision to treat the UTI) during the intervention period. However, this difference was not maintained in the post-intervention phase. Continued efforts to improve the quality of prescriptions further are necessary.Trial registrationThe trial was registered at ClinicalTrials.gov NCT04798365.

Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination

Cross-view geo-localization holds significant potential for various applications, but drastic differences in viewpoints and visual appearances between cross-view images make this task extremely challenging. Recent works have made notable progress in cross-view geo-localization. However, existing methods either ignore the correspondence between geometric spatial layout in cross-view images or require high costs or strict constraints to achieve such alignment. In response to these challenges, we propose a Feature Recombination Module (FRM) that explicitly establishes the geometric spatial layout correspondences between two views. Unlike existing methods, FRM aligns geometric spatial layout by directly recombining features, avoiding image preprocessing, and introducing no additional computational and parameter costs. This effectively reduces ambiguities caused by geometric misalignments between ground-level and aerial-level images. Furthermore, it is not sensitive to frameworks and applies to both CNN-based and Transformer-based architectures. Additionally, as part of the training procedure, we also introduce a novel weighted (B+1)-tuple loss (WBL) as optimization objective. Compared to the widely used weighted soft margin ranking loss, this innovative loss enhances convergence speed and final performance. Based on the two core components (FRM and WBL), we develop an end-to-end network architecture (FRGeo) to address these limitations from a different perspective. Extensive experiments show that our proposed FRGeo not only achieves state-of-the-art performance on cross-view geo-localization benchmarks, including CVUSA, CVACT, and VIGOR, but also is significantly superior or competitive in terms of computational complexity and trainable parameters. Our project homepage is at https://zqwlearning.github.io/FRGeo.

H4MER: Human 4D Modeling by Learning Neural Compositional Representation with Transformer.

Despite the impressive results achieved by deep learning based 3D reconstruction, the techniques of directly learning to model 4D human captures with detailed geometry have been less studied. This work presents a novel neural compositional representation for Human 4D Modeling with transformER (H4MER). Specifically, our H4MER is a compact and compositional representation for dynamic human by exploiting the human body prior from the widely used SMPL parametric model. Thus, H4MER can represent a dynamic 3D human over a temporal span with the codes of shape, initial pose, motion and auxiliaries. A simple yet effective linear motion model is proposed to provide a rough and regularized motion estimation, followed by per-frame compensation for pose and geometry details with the residual encoded in the auxiliary codes. We present a novel Transformer-based feature extractor and conditional GRU decoder to facilitate learning and improve the representation capability. Extensive experiments demonstrate our method is not only effective in recovering dynamic human with accurate motion and detailed geometry, but also amenable to various 4D human related tasks, including monocular video fitting, motion retargeting, 4D completion, and future prediction. The video demos are in the project homepage: https://boyanjiang.github.io/H4MER/.

An Adaptive Ant Colony Optimization for Solving Large-Scale Traveling Salesman Problem

The ant colony algorithm faces dimensional catastrophe problems when solving the large-scale traveling salesman problem, which leads to unsatisfactory solution quality and convergence speed. To solve this problem, an adaptive ant colony optimization for large-scale traveling salesman problem (AACO-LST) is proposed. First, AACO-LST improves the state transfer rule to make it adaptively adjust with the population evolution, thus accelerating its convergence speed; then, the 2-opt operator is used to locally optimize the part of better ant paths to further optimize the solution quality of the proposed algorithm. Finally, the constructed adaptive pheromone update rules can significantly improve the search efficiency and prevent the algorithm from falling into local optimal solutions or premature stagnation. The simulation based on 45 traveling salesman problem instances shows that AACO-LST improves the solution quality by 79% compared to the ant colony system (ACS), and in comparison with other algorithms, the PE of AACO-LST is not more than 1% and the Err is not more than 2%, which indicates that AACO-LST can find high-quality solutions with high stability. Finally, the convergence speed of the proposed algorithm was tested. The data shows that the average convergence speed of AACO-LST is more than twice that of the comparison algorithm. The relevant code can be found on our project homepage.

Personas-based Student Grouping using reinforcement learning and linear programming

Group discussions and assignments play a pivotal role in the classroom and online study. Existing research has mainly focused on exploring the educational impact of group learning, while the study on automated grouping still remains under-explored. This paper proposes a principled method that aims to achieve personalized, accurate, and efficient grouping outcomes. Dubbed as Personas-based Student Grouping (PSG), our method first applies unsupervised clustering techniques to assign personas to students based on their behavioral characteristics. Based on their personas, we then utilize deep reinforcement learning to search for appropriate grouping rules and perform linear programming to obtain a suitable grouping scheme. Finally, the teaching effectiveness is fed back as the rewards to the reinforcement learning model to optimize future grouping scheme selections. Extensive experiments conducted on MOOCs datasets show that PSG can achieve more advantageous performance in both efficiency and effectiveness compared to the manual or random grouping mechanism. We hope PSG can provide students with a more enhanced learning experience and contribute to the future development of education. Our project homepage is available at https://PSG-project.pages.dev.

Weakly Supervised 3D Multi-Person Pose Estimation for Large-Scale Scenes Based on Monocular Camera and Single LiDAR

Depth estimation is usually ill-posed and ambiguous for monocular camera-based 3D multi-person pose estimation. Since LiDAR can capture accurate depth information in long-range scenes, it can benefit both the global localization of individuals and the 3D pose estimation by providing rich geometry features. Motivated by this, we propose a monocular camera and single LiDAR-based method for 3D multi-person pose estimation in large-scale scenes, which is easy to deploy and insensitive to light. Specifically, we design an effective fusion strategy to take advantage of multi-modal input data, including images and point cloud, and make full use of temporal information to guide the network to learn natural and coherent human motions. Without relying on any 3D pose annotations, our method exploits the inherent geometry constraints of point cloud for self-supervision and utilizes 2D keypoints on images for weak supervision. Extensive experiments on public datasets and our newly collected dataset demonstrate the superiority and generalization capability of our proposed method. Project homepage is at \url{https://github.com/4DVLab/FusionPose.git}.

RGBD1K: A Large-Scale Dataset and Benchmark for RGB-D Object Tracking

RGB-D object tracking has attracted considerable attention recently, achieving promising performance thanks to the symbiosis between visual and depth channels. However, given a limited amount of annotated RGB-D tracking data, most state-of-the-art RGB-D trackers are simple extensions of high-performance RGB-only trackers, without fully exploiting the underlying potential of the depth channel in the offline training stage. To address the dataset deficiency issue, a new RGB-D dataset named RGBD1K is released in this paper. The RGBD1K contains 1,050 sequences with about 2.5M frames in total. To demonstrate the benefits of training on a larger RGB-D data set in general, and RGBD1K in particular, we develop a transformer-based RGB-D tracker, named SPT, as a baseline for future visual object tracking studies using the new dataset. The results, of extensive experiments using the SPT tracker demonstrate the potential of the RGBD1K dataset to improve the performance of RGB-D tracking, inspiring future developments of effective tracker designs. The dataset and codes will be available on the project homepage: https://github.com/xuefeng-zhu5/RGBD1K.

CL3D: Unsupervised Domain Adaptation for Cross-LiDAR 3D Detection

Domain adaptation for Cross-LiDAR 3D detection is challenging due to the large gap on the raw data representation with disparate point densities and point arrangements. By exploring domain-invariant 3D geometric characteristics and motion patterns, we present an unsupervised domain adaptation method that overcomes above difficulties. First, we propose the Spatial Geometry Alignment module to extract similar 3D shape geometric features of the same object class to align two domains, while eliminating the effect of distinct point distributions. Second, we present Temporal Motion Alignment module to utilize motion features in sequential frames of data to match two domains. Prototypes generated from two modules are incorporated into the pseudo-label reweighting procedure and contribute to our effective self-training framework for the target domain. Extensive experiments show that our method achieves state-of-the-art performance on cross-device datasets, especially for the datasets with large gaps captured by mechanical scanning LiDARs and solid-state LiDARs in various scenes. Project homepage is at https://github.com/4DVLab/CL3D.git.

A control system of rail-guided vehicle assisted by transdifferentiation strategy of lower organisms

Rail-guided vehicle is a logistics management device widely used to perform various material handling operations instead of manual labor. In processing scenarios, the dimensions of the material transfer path of a rail-guided vehicle are typically very large, which makes the optimization of the material transfer path very difficult. The transdifferentiation behavior of lower organisms was introduced into the evolutionary algorithm, and a large-scale differential evolution algorithm based on the transdifferentiation strategy was proposed, for achieving high-efficiency processing. This strategy makes it possible for some individuals with poor fitness to reach maturity again and be selected for the next iteration after losing some information and returning to their juvenile stage, which helps maintain the diversity of the population. Simulation results show that the proposed algorithm not only achieves an average 25.68% higher output rate than the comparison algorithms on the test cases but also has an excellent and stable effect distribution level on the extended problem space, which shows that the superiority of the proposed algorithm is not affected by the processing parameters. This research is expected to provide technical guidance for the processing of key components in the ship and aviation manufacturing industries. The code with a 31-page manual is available on our project homepage https://github.com/MLNST-JUST/DE-TS.

Structured lactation support and human donor milk for German NICUs-Protocol on an intervention design based on a multidimensional status quo and needs assessment (Neo-MILK).

Mother's own milk is the best nutrition for every newborn and especially for vulnerable infants such as preterm infants with a very low birth weight below 1,500 grams (VLBW). If no MOM is available, human donor milk is the alternative of choice. Mothers of preterm born infants face challenging conditions that impair sufficient milk production. For this reason, it is particularly important to provide structural lactation support and, at the same time, to promote the establishment of human donor milk banks. Via a multidisciplinary approach the Neo-MILK study will develop an intervention for structured breastfeeding and lactation support. This will be based on a comprehensive status quo and needs assessment. In addition, the implementation of human donor milk banks (HDMB) will be supported by the development of standards. Intervention development is participatory, involving different disciplines and stakeholders. All surveys are subject to approval by the ethics committee. During the course of the project, the results will be communicated to the scientific community and the general public via publications, the project homepage and social media. DRKS00024799 (German Clinical Trials Register).

Mirror Segmentation via Semantic-aware Contextual Contrasted Feature Learning

Mirrors are everywhere in our daily lives. Existing computer vision systems do not consider mirrors, and hence may get confused by the reflected content inside a mirror, resulting in a severe performance degradation. However, separating the real content outside a mirror from the reflected content inside it is non-trivial. The key challenge is that mirrors typically reflect contents similar to their surroundings, making it very difficult to differentiate the two. In this article, we present a novel method to segment mirrors from a single RGB image. To the best of our knowledge, this is the first work to address the mirror segmentation problem with a computational approach. We make the following contributions: First, we propose a novel network, called MirrorNet+, for mirror segmentation, by modeling both contextual contrasts and semantic associations. Second, we construct the first large-scale mirror segmentation dataset, which consists of 4,018 pairs of images containing mirrors and their corresponding manually annotated mirror masks, covering a variety of daily-life scenes. Third, we conduct extensive experiments to evaluate the proposed method and show that it outperforms the related state-of-the-art detection and segmentation methods. Fourth, we further validate the effectiveness and generalization capability of the proposed semantic awareness contextual contrasted feature learning by applying MirrorNet+ to other vision tasks, i.e., salient object detection and shadow detection. Finally, we provide some applications of mirror segmentation and analyze possible future research directions. Project homepage: https://mhaiyang.github.io/TOMM2022-MirrorNet+/index.html .

FeynCalc goes multiloop

We report on the new functionality of the open-source MATHEMATICA package FEYNCALC relevant for multiloop calculations. In particular, we focus on such tasks as topology identification by means of the Pak algorithm, search for equivalent master integrals and their graph representations as well as automatic derivations of Feynman parametric representations for a wide class of multiloop integrals. The functions described in this report are expected to be finalized with the official release of FEYNCALC 10. The current development snapshot of the package including the documentation is publicly available on the project homepage. User feedback is highly encouraged.

Internet-Based Self-Assessment for Symptoms of Internet Use Disorder—Impact of Gender, Social Aspects, and Symptom Severity: German Cross-sectional Study

Internet use disorder (IUD) is a new type of behavioral addiction in the digital age. At the same time, internet applications and eHealth can also provide useful support in medical treatment. The purpose of this study is to examine if an internet-based eHealth service can reach individuals with IUD. In particular, it should be investigated whether both male and female individuals with more severe IUDs can be reached. Data were retrieved from the OMPRIS (online-based motivational intervention to reduce problematic internet use and promote treatment motivation in internet gaming disorder and internet use disorder) project (DRKS00019925), an internet-based motivational intervention to reduce problematic internet use and promote treatment motivation in internet gaming disorder and IUD. During the recruitment process (August 2020-March 2022), a total of 3007 individuals filled out the standardized scale for the assessment of internet and computer game addiction (AICA-S). The assessment was accessible via the project homepage. There was no preselection of participants at this stage of the study; however, the offer was addressed to people with hazardous internet use and IUDs. The web-based assessment was free and could be found via search engines, but attention was also drawn to the service via newspaper articles, radio reports, and podcasts. Out of 3007 who participated in the web-based self-assessment, 1033 (34.4%) are female, 1740 (57.9%) are male, 67 (2.2%) are diverse individuals, and 167 (5.5%) did not disclose their gender. The IUD symptom severity score showed a wide range between the AICA-S extreme values of 0 and 27 points. On average, the total sample (mean 8.19, SD 5.47) was in the range of hazardous IUD behavior (AICA-S cutoff>7.0). Furthermore, 561 individuals (18.7% of the total sample; mean 17.42, SD 3.38) presented severe IUD (AICA-S cutoff>13.5). Focusing on female and male participants, 20.9% (363/1740) of the men and 14.9% (151/1033) of the women scored above 13.5 points, which can be considered pathological IUD behavior (χ22,2773=16.73, P<.001, effect size: Cramér V=0.078). Unemployment, being in vocational training or studying at a university, and being male were significantly associated with high IUD symptoms. Using a large sample, the study showed that both mildly and severely IUD-affected individuals can be reached via the internet. An internet-based eHealth offer can thus be a good way to reach patients with IUD where they are addicted-on the internet. In addition, eHealth services increase the likelihood of reaching female patients, who hardly ever come to specialized outpatient clinics and hospitals. Since social problems, especially unemployment, have a strong association with disease severity, the integration of social counseling into treatment seems advisable in terms of a multidisciplinary approach. German Clinical Trials Register (DRKS) DRKS00019925; https://drks.de/search/de/trial/DRKS00019925.

Deep Cross-Attention Network for Crowdfunding Success Prediction

Crowdfunding creates opportunities for entrepre- neurs. It allows startup companies to reach a large audience for fundraising and bring their creative ideas to life. In this work, we are concerned with crowdfunding project success prediction problem, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , to predict whether a project will successfully reach its funding goal by using its project profiles. This is important for startup companies to refine their project profiles and achieve their goals. Crowdfunding project success prediction is a typical classification problem but with a few critical challenges. On the one hand, with only coarse-grained project status as weak supervision, it is hard for a deep learning network to learn the relationship between project profiles and explain why it makes this prediction. On the other hand, on the project homepage, there are various modalities of description, including metadata, textual description, images, and videos. Among those, videos play an important role in the success of a crowdfunding project, however, were ignored in previous works, due to the difficulty in extracting useful semantic and authentic information from videos, especially for the crowdfunding project where information in different modalities are unaligned. To this end, we propose a novel framework called Deep Cross-Attention Network to learn and fuse information from introduction videos and textual descriptions of project profiles. More specifically, we develop a cross-attention block to align and represent mismatched textual description and untrimmed introduction videos and fuse the information from these two modalities, which effectively remedies the lack of supervised information caused by project status as weak supervision. More importantly, with our cross-attention mechanism, the model is able to interpret how it makes such predictions and show which keywords and keyframes it depends on. We conduct extensive experiments on two crowdfunding datasets (collected from Kickstarter and Indiegogo) and show that our method achieves superior performance over existing state-of-the-art baselines.

HPC AI500 V3.0: A scalable HPC AI benchmarking framework

In recent years, the convergence of High Performance Computing (HPC) and artificial intelligence (AI) makes the community desperately need a benchmark to guide the design of next-generation scalable HPC AI systems. The success of the HPL benchmarks and the affiliated TOP500 ranking indicates that scalability is the fundamental requirement to evaluate HPC systems. However, being scalable in terms of these emerging AI workloads like deep learning (DL) raises nontrivial challenges. This paper formally and systematically analyzes the factor that limits scalability in DL workloads and presents HPC AI500 v3.0, a scalable HPC AI benchmarking framework. The HPC AI500 V3.0 methodology is inspired by bagging, which utilizes the collective wisdom of an ensemble of base models and enables the benchmarks to be adaptively scalable to different scales of HPC systems. We implement HPC AI500 V3.0 in a highly customizable manner, maintaining the space of various optimization from both system and algorithm levels. By reusing the representative workloads in HPC AI500 V2.0, we evaluate HPC AI500 V3.0 on typical HPC systems, and the results show it has near-linear scalability. Furthermore, based on the customizable design, we present a case study to perform a trade-off between AI model quality and its training speed. The source code of HPC AI500 V3.0 is publicly available from the HPC AI500 project homepage https://www.benchcouncil.org/aibench/hpcai500/.