Developing computer-aided design (CAD) generation models has significantly enhanced design efficiency, facilitating innovation and transformation in the design industry. Existing methods typically require users to input prompts in a specific format, such as text descriptions or images, limiting their broader application in diverse scenarios. To address this limitation, we introduce FreeCAD, a user-friendly CAD generation framework that supports free-form inputs, including text descriptions and/or images, enabling users to express their design intentions more flexibly. Specifically, we propose a Large Language Models (LLMs)-based Text Translator, which effectively increases the success rate of generating CAD models by converting users' diversified requests for the same object into a unified expression. Additionally, the Multi-View Representation Fusion (MVRF) module enables the network to capture richer interaction information across views, facilitating the generation of more fine-grained CAD models. To support the training of FreeCAD, we construct a multimodal dataset RealCAD, comprising text, image, and CAD triplets, where the images are derived from the 3D printed products of CAD models. Extensive experiments demonstrate that FreeCAD consistently outperforms the existing state-of-the-art (SOTA) methods in multiple tasks.

Radar Echo Extrapolation (REE) can facilitate accurate and expeditious nowcasting of precipitation, reducing the reliance on complex Numerical Weather Prediction (NWP) models. Spatial-temporal forecasting methods dominate this task because they can fully exploit the spatio-temporal dependencies and complex dynamic patterns inherent in radar echo data. However, they struggle with handling uncertainty and incorporating domain-specific knowledge, often resulting in blurry or unrealistic predictions. We propose a Hypergraph-enhanced Latent Diffusion Model (HyDiff) to address these limitations. The EchoDiff has been utilized to aid in accurate extrapolation. To accurately describe precipitation microphysics while adhering to the hydro-microphysical and multiscale coupling principles, the method integrates additional semantic information into the model. Specifically, the Differential Reflectivity Factor (ZDR) and Differential Propagation Phase Shift (KDP) are incorporated into the model as additional semantic information. Furthermore, we introduce a Hypergraph Neural Network (HGNN) into the extrapolation method to capture correlation information across regions. Experiments show that HyDiff effectively handles uncertainty, incorporates domain-specific prior knowledge, and generates forecasts with high operational utility.

The emergence of new network attacks, especially multi-step attacks which exhibit complex patterns, presents challenges to network security. Intrusion detection in network traffic is one of the core means to ensure network operation security. Traditional intrusion detection methods identify abnormal traffic by modeling the patterns of normal traffic. This strategy focuses on the spatial characteristics of data, ignoring the temporal characteristics of traffic evolution, and may not fully capture the patterns of multi-step network attacks. To address this problem, we propose a Temporal-Gated Graph Neural Network (TGGNN) framework based on graph clustering sampling. Firstly, each network traffic data is viewed as a graph node. We use hierarchical graph clustering methods to sample representative points and further construct local evolutionary structures with their temporally adjacent nodes. Then, we design a novel way of constructing graph data that balances both local and global aspects of network traffic data. Based on the GGNN (Gated Graph Neural Network), we designed a Temporal-Gated Mechanism in the information propagation part, emphasizing the learning of the local evolutionary structure of the graph. Furthermore, a bidirectional LSTM network is used to further enhance the learning of dynamic patterns in network traffic data. Experimental results based on the UNSW-NB15 dataset demonstrate that our approach not only surpasses the performance of four recent baselines but also eliminates the need for feature engineering by adopting an end-to-end approach.

Abstract Iris segmentation is a critical step in the iris recognition system. Since the quality of iris database taken under different camera sensors varies greatly, thus most existing iris segmentation methods are designed for a particular collection device. Meanwhile, many iris segmentation methods based on convolutional neural networks (CNNs) require a lot of computational costs and hardware costs (storage space), which are not suitable for deploying on low‐performance devices. To address the above problems, an accurate and efficient heterogeneous iris segmentation network is proposed in this paper. First, the authors design an efficient feature extraction network, which combines depth‐wise separable convolution with traditional convolution to greatly reduce model parameters and computational cost while maintaining segmentation accuracy. Then, a Multi‐scale Context Information Extraction Module (MCIEM) is proposed to extract multi‐scale spatial information at a more granular level and enhance the discriminability of the iris region. Finally, a Multi‐layer Feature Information Fusion Module (MFIFM) is proposed to reduce the loss of information during the downsampling process. Experimental results on multi‐source heterogeneous iris database show that the proposed network can not only achieve state‐of‐the‐art performance but is also more efficient in terms of required parameters, calculated load, and storage space.

Technical debt describes a trade-off between short-term goals and long-term code quality during software development. Self-admitted technical debt (SATD), a type of technical debt, is intentionally introduced by developers. The existence of SATD is likely to leave hidden dangers for future changes in software systems, so identifying SATD is an essential task. Before this, many methods for recognizing SATD (such as pattern matching-based, natural language processing-based, text mining-based, etc.) have been proposed. This paper will present a pre-trained deep learning model to complete the SATD recognition task. An efficient deep learning model interpretation tool Captum can be used to understand the experimental results. At the same time, a new interpretation view is proposed for the matching-based model. Finally, combined with the research in this paper, reasonable suggestions are put forward for future SATD recognition tasks.

This book caught my eye by its title. While it is unusual to relate ‘next generation’ to microarrays, ‘next generation’ is widely used to refer to a variety of rapidly advancing DNA sequencing technologies. Since 2004, the Next Generation Sequencing (NGS) technologies have increased throughput more than a million fold over the first-generation sequencing (called Sanger sequencing) at the time of this review. The availability of massive digital sequence information at base pair resolution not only provides higher sensitivity and specificity on numeric measurements than what microarrays can offer but also allows inferences to be made regarding genetic variations within the same set of sequences, which are added benefits that are not available in microarrays. The important questions are as follows: Will NGS completely replace microarrays? Where will microarray technologies go? How can we apply what we learned from microarray Bioinformatics to NGS Bioinformatics? I started reading with curiosity about how this book can help address these questions. The volume does not cover all of the latest microarray and NGS applications. Instead, the authors have assembled a set of popular techniques and have focused on explaining how they work. Although the topics are highly technical and specific, most chapters are well written and provide sufficient technical and mathematic details as well as references for a well-trained bioinformatician to implement the algorithms, use the software tools and follow the protocols therein. The book is a good resource for someone who wants to get a general idea of, or seek guidance for, a particular microarray application. Each chapter is written in a format consisting of four sections after an abstract to make it different from a journal publication. First, ‘Introduction’ gives a general overview of the topic. Then ‘Materials and Methods’ sections describe the details of a particular application or a protocol and use one or more examples to illustrate how algorithms or software tools work. At the end, there is a unique ‘Notes’ section to provide tips, highlight pitfalls or clarify key concepts. Each chapter is written to be read independently, so there are nominal overlaps among some chapters. The book is a collection of 26 technological reports and reviews at intermediate to advanced levels. They are partitioned into five sections: (i) Introduction and Resources for Microarray Bioinformatics; (ii) Microarray Data Analysis (Top–Down Approach), which details the analysis of systematic characterizations of the cell to understand the functions and associations of components; (iii) Microarray Bioinformatics in Systems Biology (Bottom–Up Approach), which describes modeling individual components, their interactions based on microarray results and the inference of regulatory relationships at the system level; (iv) Next Generation Sequencing Data Analysis, a bonus section that is not implied in the book title and (v) Emerging Applications of Microarray and Next Generation Sequencing. Because each chapter has a good abstract, instead of repeating this information, a brief highlight is given below to provide a quick overview of the contents of the book. Number codes are given to acknowledge the source of the highlights. Chapter reference is a C plus a number. Study subjects are as follows: SNP for single-nucleotide polymorphism, DEG for differential gene expression, ChIP-Seq for chromatin immunoprecipitation sequencing, DB for database and GRN for gene regulatory network. Section 1 provides general information about microarrays. It starts with a primer on the history of microarray technology, the workflow for using it and how it has become a standardized tool (C1). This is followed by a demonstration of how to use the KEGG database to analyze Omics data (C2, DB). An overview is given on the structure and usage of the Gene Expression Omnibus (GEO) database, which is one of the two major archives that have collected almost a million microarray experiments (C3, DB). Section 2 is the heart of the book that introduces a number of advanced tools in microarray data analysis. This includes an actively maintained R package named ASCAT for allele-specific copy number analysis for tumors to study genomic aberrations in cancers, which is particularly useful to deal with aneuploidy, cell admixture and tumor heterogeneity (C4, SNP); a clustering algorithm called Kernel-imbedded Gaussian processes, which is able to account for both linear and non-linear relationships in gene expressions and outperforms popular classification methods such as support vector machine (SVM) on test data sets (C5, DES); a modified iterative signature algorithm (ISA), called enrichment-constrained time-dependent ISA (ECTDISA), which takes advantage of the dependencies of adjacent time points and known biological associations within biological modules to classify difficult gene clusters (C6, DES); the Bioconductor package ‘GeneAnswers’ to enhance functional interpretations by integrating and visualizing at once all the gene ontology, disease ontology, and protein–protein interactions in which top differentially expressed genes are involved (C7, DES); an example of using the multiple comparison R package ‘multcomp’ to analyze splicing variants across cell lines and conditions (C8,Isoform); technology comparisons across multiple microarray platforms using overlaps of highly expressed genes (C9) and enriched functional categories (C10), which found that different platforms produced consistent outcomes; and two information integration tools to synthesize data across multiple microarray experiments—COALESCE uses both expression data and upstream sequences to discover motifs of candidate transcription factor binding sites, while MEFIT uses a Bayesian algorithm and known functional groups to predict co-expression of gene pairs (C11). Section 3 focuses on numerical modeling for gene regulatory networks based on time series gene expression data. Topics include the following: an ordinary differential equations approach to model regulation dynamics of a three-gene feed forward loop based on time series data (C12, GRN); an improved dynamic Bayesian networks algorithm that takes into account non-homogeneity in temporal processes to improve network structure discovery performance (C13, GRN); the R package qpgraph that infers regulatory networks coherent with annotation databases by considering gene interactions as a multivariate event instead of a bivariate one (C14, GRN) and a non-linear model to study time series data (C15, GRN, P53). Section 4 is about microarray applications that are being replaced by NGS technologies. The section has an overview of NGS technologies and their major applications (C16); a review of mRNA-sequencing (RNA-Seq) experiments and data analysis using human as an example to introduce basic concepts, formats and their conversions, data visualization and statistical analysis (C17, RNA-Seq); an introduction to ChIP-Seq, a technology that uses antibody-enriched DNA sequences to discover DNA-binding sites (C18, ChIP-Seq); the software tool ChIPDiff to study differential histone modification sites (C19, ChIP-Seq); a computational tool, SISSRs that discovers DNA-binding sites by identifying non-random high sequence coverage areas (C20, ChIP-Seq); a de novo sequence motif discovery and visualization tool called ChIPMotifs that takes outputs from ChIP-Chip, ChIP-Seq and ChIP-PET experiments and discovers motifs by integrating results of three popular tools, namely, MEME, MaMF and Weeder. Section 5 focuses on meta-data analysis across experimental studies. It includes the following: a Hidden Markov Model-based pooled local index of significance (PLIS) procedure to increase the sensitivity of identifying susceptibility loci in genome-wide association analysis by explicitly considering high linkage disequilibrium of SNPs (C22, SNP); an algorithm called gene set top scoring pairs to find core signature genes in functional modules using ranks of gene expression levels (C23, DEG); the R package JAIME for jointly analyzing multiple ChIP-Chip experiments that applies a hierarchical mixture model and expectation–maximization algorithm to integrate ChIP-Seq data sets (C24, ChIP-Seq); an overview of ChIP-Chip and ChIP-Seq analysis technologies (C25, ChIP-Seq) and an analysis workflow called BiNGS!SL-seq for short hairpin RNA (shRNA) discovery by synthetic lethal screening (C26, shRNA). It is evident in this book that although some microarray applications have been replacing by the advanced capabilities of NGS (as described in Section 4), the impacts of microarrays to biology are profound and far-reaching. The established and emerging uses of microarray technology, infrastructure, domain expertise, knowledge base and commercial products will continue to benefit research and clinical applications in the foreseeable future. After several iterations, many microarray technologies have converged to a few commercial platforms including Affymetrix, Illumina, Agilent, Roche Nimblegen, Fluidigm and NanoString. The consistent performance and low cost of those products have accelerated the adoption of microarrays by both research and clinical labs. This convergence has stimulated a plethora of developments in standardizing and optimizing the experiments as well as in-depth studies to understand errors and to decipher signals from complex data sets. The impacts have been reflected in the large portfolio of algorithms and tools introduced by the book. The maturity of microarray technology also led to the FDA-approved diagnostics of the Roche AmpliChip® CYP450 Test in 2005, 10 years after the first microarray article was published and Pathwork® Diagnostics’ Tissue of Origin Test in 2010. Those accomplishments are clear advantages of microarray technology over NGS in large-scale genomics studies. NGS, on the other hand, is still rapid evolving and has yet to mature in both technological sophistication and methodology. The book also points out another fact that can not be overlooked—the accumulation of nearly one million experimental data sets that are compliant to community data-sharing standards like MIAME. Such resources have inspired many data mining efforts to discover biological relevance that are unable to be done in any one single experiment. Many authors in this book are developers of original tools used for such purposes. Most of these tools are available in public resources such as Bioconductor. Fortunately, many of same statistical considerations and data mining techniques used for microarray data are transferable to the analysis of NGS data. Enabling technologies tend to evolve and be absorbed into other emerging applications to push limits. Microarray technologies are following the same principle. As an example, microarrays have been used as an orthogonal validation tool for whole genome sequencing. The low cost and reliable readouts are an ideal test to make sure the sequencing is done for the intended individual. Targeted sequencing by capturing through hybridization is another new application area for microarrays. Although the book has not discussed about them, in the future, we should expect to see more synergies of the microarray technologies and data analysis methods outlined in this book with other technologies such as NGS.