Target Classification Research Articles

Abstract C099: Genetic ancestry specific loss of heterozygosity of essential genes as therapeutic vulnerabilities in cancer

Abstract Barriers in access to clinical care among under-served populations in combination with a vast underrepresentation at a scientific research level perpetuate inequalities in cancer treatment. These disparities are further disseminated due to the lack of understanding underlying genomic differences contributing to cancer outcome among these patients. The contributions of germline and somatic variations in cancer often shape treatment regimens, and these variations are not well investigated in non-European populations. We have described a class of potential therapeutic targets (GEMINI genes) that reflect a combination of germline alterations with somatic loss of heterozygosity (LOH). GEMINI genes are essential genes with frequent germline polymorphisms, whose frequency varies by ancestry. Indeed, because genomes of African ancestry tend to be the most diverse, many GEMINI genes are the most polymorphic in this patient population. When normal cells are heterozygous, LOH leaves cancer cells reliant on the gene products encoded by a single allele, where as healthy cells retain both copies. LOH is a common genetic event, often affecting over 20% of the genome in certain cancers, making this approach widely useful. Targeting GEMINI vulnerabilities opens up possibilities for allele-specific inhibitors or degrader therapeutics that can be catered in an ancestry-dependent manner, to ameliorate inequities in cancer genomics research. One approach to readily target individual alleles is to utilize nucleic acid-based methods, as we’ve previously demonstrated using PRIM1. This therapeutic is expected to be the most specific when targeting polymorphisms that create the greatest differences between alleles. Here, we have prioritized a set of GEMINI targets based upon the extent of differences between alleles, and rates of heterozygosity and LOH across populations. Our initial analysis of TCGA data identified 5664 variants in 1278 essential genes that undergo LOH in cancer. We filtered GEMINI vulnerabilities by means of insertion-deletion (indel) difference of the alternate allele compared to the reference allele, followed by sorting based off maximum allele frequency by genetic ancestry. Here, we identify 31 variants across 23 genes enriched in African ancestry, 7 variants across 7 genes enriched in American ancestry, 12 variants across 12 genes enriched in East Asian ancestry, 4 variants enriched in 4 genes in South Asian ancestry, and 3 variants enriched in 3 genes in European ancestry. In addition, each of these ancestry-specific GEMINI genes exhibit a minimum 10% pan-cancer LOH rate, with the highest LOH frequencies observed in ovarian and kidney chromophobe cancers. As both these cancers exhibit disparities in mortality at a population level, targeting GEMINI genes as a therapeutic strategy may aid in lessening the disproportionate rate of outcome resultant of these cancers. Overall, we identify novel GEMINI genes to assess ancestry-specific genomic variations to serve as targets for precision treatment specific to underserved patient populations. Citation Format: Nicole Peiris, Catherine Hazard, Charlotte E. Farquhar, Andrei Loas, Bradley L. Pentelute, Rameen Beroukhim. Genetic ancestry specific loss of heterozygosity of essential genes as therapeutic vulnerabilities in cancer [abstract]. In: Proceedings of the 17th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2024 Sep 21-24; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2024;33(9 Suppl):Abstract nr C099.

Classification of Small Targets on Sea Surface Based on Improved Residual Fusion Network and Complex Time–Frequency Spectra

To address the problem that conventional neural networks trained on radar echo data cannot handle the phase of the echoes, resulting in insufficient information utilization and limited performance in detection and classification, we extend neural networks from the real-valued neural networks to the complex-valued neural networks, presenting a novel algorithm for classifying small sea surface targets. The proposed algorithm leverages an improved residual fusion network and complex time–frequency spectra. Specifically, we augment the Deep Residual Network-50 (ResNet50) with a spatial pyramid pooling (SPP) module to fuse feature maps from different receptive fields. Additionally, we enhance the feature extraction and fusion capabilities by replacing the conventional residual block layer with a multi-branch residual fusion (MBRF) module. Furthermore, we construct a complex time–frequency spectrum dataset based on radar echo data from four different types of sea surface targets. We employ a complex-valued improved residual fusion network for learning and training, ultimately yielding the result of small target classification. By incorporating both the real and imaginary parts of the echoes, the proposed complex-valued improved residual fusion network has the potential to extract more comprehensive features and enhance classification performance. Experimental results demonstrate that the proposed method achieves superior classification performance across various evaluation metrics.

Automatic target detection in vehicle images based on YOLOv10 deep learning algorithm

In this paper, we explore the effect of its application in vehicle image detection based on the latest YOLOv10 model.YOLOv10, as the latest version in the YOLO series, inherits and optimises the advantages of the previous generations of models, aiming to provide higher detection accuracy and faster inference speed. In our experiments, we firstly divided the dataset reasonably and trained the YOLOv10 model with data from the training set. By analysing the Precision-Recall curve, we found that the area below the PR curve is larger, which indicates that the AUC value of the model is close to 1, thus reflecting the superiority of the model in classification performance. In addition, from the F1-confidence curve, the model is still able to maintain good classification results at higher confidence levels and maintains high F1 scores in most confidence intervals, which further validates the reliability of the model. Test results show that the trained YOLOv10 model is able to accurately recognise vehicles and effectively predict vehicle types. This enables the model to monitor the vehicle conditions on the road in real time and achieve accurate detection and classification of road traffic targets. Therefore, this paper demonstrates the powerful ability of YOLOv10 in the field of vehicle detection, which provides an important reference for the development of future intelligent transport systems. Overall, this study not only verifies the effectiveness of YOLOv10 in practical applications, but also lays the foundation for further improvement of target detection technology.

Inflammasome components as new therapeutic targets in inflammatory disease.

Inflammation drives pathology in many human diseases for which there are no disease-modifying drugs. Inflammasomes are signalling platforms that can induce pathological inflammation and tissue damage, having potential as an exciting new class of drug targets. Small-molecule inhibitors of the NLRP3 inflammasome that are now in clinical trials have demonstrated proof of concept that inflammasomes are druggable, and so drug development programmes are now focusing on other key inflammasome molecules. In this Review, we describe the potential of inflammasome components as candidate drug targets and the novel inflammasome inhibitors that are being developed. We discuss how the signalling biology of inflammasomes offers mechanistic insights for therapeutic targeting. We also discuss the major scientific and technical challenges associated with drugging these molecules during preclinical development and clinical trials.

Bidirectional Decoupled Distillation for Heterogeneous Federated Learning.

Federated learning enables multiple devices to collaboratively train a high-performance model on the central server while keeping their data on the devices themselves. However, due to the significant variability in data distribution across devices, the aggregated global model's optimization direction may differ from that of the local models, making the clients lose their personality. To address this challenge, we propose a Bidirectional Decoupled Distillation For Heterogeneous Federated Learning (BDD-HFL) approach, which incorporates an additional private model within each local client. This design enables mutual knowledge exchange between the private and local models in a bidirectional manner. Specifically, previous one-way federated distillation methods mainly focused on learning features from the target class, which limits their ability to distill features from non-target classes and hinders the convergence of local models. To solve this limitation, we decompose the network output into target and non-target class logits and distill them separately using a joint optimization of cross-entropy and decoupled relative-entropy loss. We evaluate the effectiveness of BDD-HFL through extensive experiments on three benchmarks under IID, Non-IID, and unbalanced data distribution scenarios. Our results show that BDD-HFL outperforms state-of-the-art federated distillation methods across five baselines, achieving at most 3% improvement in average classification accuracy on the CIFAR-10, CIFAR-100, and MNIST datasets. The experiments demonstrate the superiority and generalization capability of BDD-HFL in addressing personalization challenges in federated learning.

Exploring Reinforced Class Separability and Discriminative Representations for SAR Target Open Set Recognition

Current synthetic aperture radar (SAR) automatic target recognition (ATR) algorithms primarily operate under the closed-set assumption, implying that all target classes have been previously learned during the training phase. However, in open scenarios, they may encounter target classes absent from the training set, thereby necessitating an open set recognition (OSR) challenge for SAR-ATR. The crux of OSR lies in establishing distinct decision boundaries between known and unknown classes to mitigate confusion among different classes. To address this issue, we introduce a novel framework termed reinforced class separability for SAR target open set recognition (RCS-OSR), which focuses on optimizing prototype distribution and enhancing the discriminability of features. First, to capture discriminative features, a cross-modal causal features enhancement module (CMCFE) is proposed to strengthen the expression of causal regions. Subsequently, regularized intra-class compactness loss (RIC-Loss) and intra-class relationship aware consistency loss (IRC-Loss) are devised to optimize the embedding space. In conjunction with joint supervised training using cross-entropy loss, RCS-OSR can effectively reduce empirical classification risk and open space risk simultaneously. Moreover, a class-aware OSR classifier with adaptive thresholding is designed to leverage the differences between different classes. Consequently, our method can construct distinct decision boundaries between known and unknown classes to simultaneously classify known classes and identify unknown classes in open scenarios. Extensive experiments conducted on the MSTAR dataset demonstrate the effectiveness and superiority of our method in various OSR tasks.

Marginal debiased network for fair visual recognition

Deep neural networks (DNNs) are often prone to learn the spurious correlations between target classes and bias attributes, like gender and race, inherent in a major portion of training data (bias-aligned samples), thus showing unfair behavior and arising controversy in the modern pluralistic and egalitarian society. In this paper, we propose a novel marginal debiased network (MDN) to learn debiased representations. More specifically, a marginal softmax loss (MSL) is designed by introducing the idea of margin penalty into the fairness problem, which assigns a larger margin for bias-conflicting samples (data without spurious correlations) than for bias-aligned ones, so as to deemphasize the spurious correlations and improve generalization on unbiased test criteria. To determine the margins, our MDN is optimized through a meta learning framework. We propose a meta equalized loss (MEL) to perceive the model fairness, and adaptively update the margin parameters by meta-optimization which requires the trained model guided by the optimal margins should minimize MEL computed on an unbiased meta-validation set. Extensive experiments on BiasedMNIST, Corrupted CIFAR-10, CelebA and UTK-Face datasets demonstrate that our MDN can achieve a remarkable performance on under-represented samples and obtain superior debiased results against the previous approaches.

Full-length nanopore sequencing of circular RNA landscape in peripheral blood cells following sequential BNT162b2 mRNA vaccination

Circular RNAs (circRNA) lack 5′ or 3′ ends; their unique covalently closed structures prevent RNA degradation by exonucleases. These characteristics provide circRNAs with high pharmaceutical stability and biostability relative to current standard-of-care linear mRNAs. CircRNA levels are reportedly associated with certain human diseases, making them novel disease biomarkers and a noncanonical class of therapeutic targets. In this study, the endogenous circRNAs underlying the response to BNT162b2 mRNA vaccination were evaluated. To this end, peripheral blood samples were subjected to full-length sequencing of circRNAs via nanopore sequencing and transcriptome sequencing. Fifteen samples, comprising pre-, first, and second vaccination cohorts, were obtained from five healthcare workers with no history of SARS-CoV-2 infection or previous vaccination. A total of 4706 circRNAs were detected; following full-length sequencing, 4217 novel circRNAs were identified as being specifically expressed during vaccination. These circRNAs were enriched in the binding motifs of stress granule assemblies and SARS-CoV-2 RNA binding proteins, namely poly(A) binding protein cytoplasmic 1 (PABPC1), pumilio RNA binding family member 1 (PUM1), and Y box binding protein 1 (YBX1). Moreover, 489 circRNAs were identified as previously reported miRNA sponges. The differentially expressed circRNAs putatively originated from plasma B cells compared to circRNAs reported in human blood single-cell RNA sequencing datasets. The pre- and post-vaccination differences observed in the circRNA expression landscape in response to the SARS-CoV-2 BNT162b2 mRNA vaccine.

Abstract MP16: Expression of an AT 1 R Receptor Mutant which Genetically Biases Gαq Specifically in Vascular Smooth Muscle Cell Causes Hypertension and Enhanced Vascular Contraction

Angiotensin II (ANG) type 1 receptor (AT 1 R) belongs to the 7-membrane superfamily of G protein-coupled receptors (GPCR) which have the largest class of drug targets. AT 1 Rs can adopt different conformations that can bias G-protein coupling (particularly Gaq) or b-arrestin signaling. The b-arrestin pathway is reported to be cardioprotective while the canonical Gaq-dependent pathway is typically associated with the detrimental effects of ANG. We hypothesize that selective and inducible activation of the AT 1 R Gaq pathway in the absence of b-arrestin signaling in vascular smooth muscle (SMC) results in hypertension and vascular dysfunction via an AT 1 R-dependent mechanism. We generated an AT 1 R mutant (AT 1 R TSTS-AAAA ) which ablates the phosphorylation sites for GPCR kinases (GRK) in the C-terminal to impair b-arrestin recruitment and to induce a Gaq bias. Transgenic mice were made with a construct that induces both AT 1 R TSTS-AAAA and tdTomato expression in response to Cre-recombinase. Primary fibroblasts cultured from AT 1 R TSTS-AAAA mice exhibited a Cre-induced expression of tdTomato and marked increase in phosphorylated-extracellular signal-regulated kinase (P-ERK) in response to ANG when compared to mice cells expressing wildtype AT 1 R. Mice carrying inducible AT 1 R TSTS-AAAA were bred with tamoxifen-inducible SMC-specific CRE (SMC CreERT2 , S-TSTS) mice. Transgene expression in S-TSTS mice was successfully activated by tamoxifen as measured by the expression of the embedded tdTomato reporter in the aorta SMC. S-TSTS mice exhibited a trend of increased systolic blood pressure (SBP) 1 week after tamoxifen injection (147.76 ± 13.21 vs. 108.85 ± 4.58 mmHg). Candesartan (10 mg/kg/day in drinking water) was then administered to S-TSTS mice once hypertension was established. Candesartan treatment for 1 week reversed the hypertension in S-TSTS mice (100.00 ± 2.88 mmHg). Interestingly, 2 weeks after ending candesartan treatment, SBP was once again elevated in S-TSTS mice (142.01 ± 7.65 vs. 90.35 ± 1.14 mmHg, p<0.05). In addition, mesenteric arteries from transgenic mice exhibited a robust increase in ANG-induced constriction when compared to control mice (ANG, max constriction 60.25 ± 4.78% vs. 14.12 ± 4.72%, p<0.05). Our findings support the hypothesis that signaling through the Gaq-dependent pathway in SMC leads to increased vasoconstriction and hypertension through an AT 1 R-dependent mechanism.

Abstract P343: Cerebral injury and cognitive performances are correlated with immunoinflammatory markers in hypertensive patients

Immunity and inflammation play a pivotal role in hypertension onset and cardiovascular organ damage. This role has been thoroughly characterized for classical hypertension targets as the kidneys, the heart and the vasculature. Many studies have characterized this in the experimental context, however immunoinflammatory challenges links with brain injury in the clinical context of hypertension are still lacking. In this study we will characterize how immunoinflammatory mediators can be predictors of cerebral injury characterized by advanced neuroimaging in hypertensive patients. Hypertensive patients underwent cerebral MRI and by advanced neuroimaging we characterized the microstructural injury by DTI fiber tracking. We administered the Montreal Cognitive Assessment to evaluate cognitive function, measured circulating C-reactive protein (hs-CRP) and counted circulating white blood cells (WBC). We found a large cluster of white matter tracts whose integrity parameters were associated with immunoinflammatory biomarkers (Figure A). In particular, a cluster of tracts of the limbic system and encompassing the corpus callosum show a correlation between their loss of integrity and hs-CRP levels. Another cluster of associative tracts spanning the temporal lobe show association between their loss of integrity and the WBC. Finally, the Frontal Aslant and the Uncinate Fasciculus show association with both immunoinflammatory markers (Figure B). In this study we performed advanced neuroimaging analyses to characterize the interplay between immunoinflammatory risk and cerebral alterations in hypertensives. Furthermore, our previous work identified that white matter damage in the Forceps Minor and Superior Longitudinal Fasciculus was a typical trait of hypertensive patients and associated to loss of cognitive functions. Our data further show the inflammatory risk association with worse cognitive functions, suggesting a clinical relevance in terms of damage evidenced by advanced MRI in hypertensive patients.

Open-Set Recognition Model for SAR Target Based on Capsule Network with the KLD

Synthetic aperture radar (SAR) automatic target recognition (ATR) technology has seen significant advancements. Despite these advancements, the majority of research still operates under the closed-set assumption, wherein all test samples belong to classes seen during the training phase. In real-world applications, however, it is common to encounter targets not previously seen during training, posing a significant challenge to the existing methods. Ideally, an ATR system should not only accurately identify known target classes but also effectively reject those belonging to unknown classes, giving rise to the concept of open set recognition (OSR). To address this challenge, we propose a novel approach that leverages the unique capabilities of the Capsule Network and the Kullback-Leibler divergence (KLD) to distinguish unknown classes. This method begins by deeply mining the features of SAR targets using the Capsule Network and enhancing the separability between different features through a specially designed loss function. Subsequently, the KLD of features between a testing sample and the center of each known class is calculated. If the testing sample exhibits a significantly larger KLD compared to all known classes, it is classified as an unknown target. The experimental results of the SAR-ACD dataset demonstrate that our method can maintain a correct identification rate of over 95% for known classes while effectively recognizing unknown classes. Compared to existing techniques, our method exhibits significant improvements.

Hierarchical cavitation intensity recognition using Sub-Master Transition Network-based acoustic signals in pipeline systems

Deep learning (DL) has emerged as the most effective approach for cavitation intensity recognition. However, current DL-based models face limitations due to the inherent assumption that all target classes are treated equally. In particular, when certain classes are more difficult to distinguish than others, or when classes can be structured into a hierarchy of categories, DL-based cavitation recognition models struggle to achieve optimal performance. In this study, we propose a hierarchical cavitation intensity recognition framework using the Sub-Master Transition Network (SMTNet) to improve the performance of cavitation intensity recognition of acoustic signal valves. The SMTNet model achieves cavitation intensity recognition from coarse to fine according to the hierarchical properties of cavitation states, overcoming the limitations of DL methods. We also present a global filter layer to address for the long-term dependence of the acoustic signals. In addition, we introduce a 1D double hierarchical residual block to capture sensitive features of acoustic signals. Moreover, a hierarchical multi-label tree is embedded in the SMTNet to map the semantic structure of cavitation states. The proposed model has been evaluated on three real-world cavitation datasets from SAMSON AG, all showing superior results and outperforming SOTA DL methods.

Prototype-wise self-knowledge distillation for few-shot segmentation

Few-shot segmentation was proposed to obtain segmentation results for a image with an unseen class by referring to a few labeled samples. However, due to the limited number of samples, many few-shot segmentation models suffer from poor generalization. Prototypical network-based few-shot segmentation still has issues with spatial inconsistency and prototype bias. Since the target class has different appearance in each image, some specific features in the prototypes generated from the support image and its mask do not accurately reflect the generalized features of the target class. To address the support prototype consistency issue, we put forward two modules: Data Augmentation Self-knowledge Distillation (DASKD) and Prototype-wise Regularization (PWR). The DASKD module focuses on enhancing spatial consistency by using data augmentation and self-knowledge distillation. Self-knowledge distillation helps the model acquire generalized features of the target class and learn hidden knowledge from the support images. The PWR module focuses on obtaining a more representative support prototype by conducting prototype-level loss to obtain support prototypes closer to the category center. Broad evaluation experiments on PASCAL-5i and COCO-20i demonstrate that our model outperforms the prior works on few-shot segmentation. Our approach surpasses the state of the art by 7.5% in PASCAL-5i and 4.2% in COCO-20i.

Identifying and Filling the Chemobiological Gaps of Gut Microbial Metabolites.

Human gut microbial metabolites are currently undergoing much research due to their involvement in multiple biological processes that are important for health, including immunity, metabolism, nutrition, and the nervous system. Metabolites exert their effect through interaction with host and bacterial proteins, suggesting the use of "metabolite-mimetic" molecules as drugs and nutraceutics. In the present work, we retrieve and analyze the full set of published interactions of these compounds with human and microbiome-relevant proteins and find patterns in their structure, chemical class, target class, and biological origins. In addition, we use virtual screening to expand (more than 4-fold) the interactions, validate them with retrospective analyses, and use bioinformatic tools to prioritize them based on biological relevance. In this way, we fill many of the chemobiological gaps observed in the published data. By providing these interactions, we expect to speed up the full clarification of the chemobiological space of these compounds by suggesting many reliable predictions for fast, focused experimental testing.

A Novel Confocal Scanning Protein-Protein Interaction Assay (PPI-CONA) Reveals Exceptional Selectivity and Specificity of CC0651, a Small Molecule Binding Enhancer of the Weak Interaction between the E2 Ubiquitin-Conjugating Enzyme CDC34A and Ubiquitin.

Protein-protein interactions (PPIs) are some of the most challenging target classes in drug discovery. Highly sensitive detection techniques are required for the identification of chemical modulators of PPIs. Here, we introduce PPI confocal nanoscanning (PPI-CONA), a miniaturized, microbead based high-resolution fluorescence imaging assay. We demonstrate the capabilities of PPI-CONA by detecting low affinity ternary complex formation between the human CDC34A ubiquitin-conjugating (E2) enzyme, ubiquitin, and CC0651, a small molecule enhancer of the CDC34A-ubiquitin interaction. We further exemplify PPI-CONA with an E2 enzyme binding study on CC0651 and a CDC34A binding specificity study of a series of CC0651 analogues. Our results indicate that CC0651 is highly selective toward CDC34A. We further demonstrate how PPI-CONA can be applied to screening very low affinity interactions. PPI-CONA holds potential for high-throughput screening for modulators of PPI targets and characterization of their affinity, specificity, and selectivity.

Enhancing User Localization with an Integrated Sensing and Communication (ISAC) System: An Experimental UAV Search-and-Rescue Use Case

This paper explores the potential of an Integrated Sensing and Communication (ISAC) system to enhance search-and-rescue operations. While prior research has explored ISAC capabilities in Unmanned Aerial Vehicles (UAVs), our study focuses on addressing the specific challenges posed by modern communication standards (e.g., power, frequency, and bandwidth limitations) in the context of search-and-rescue missions. The paper details effective methods for processing echoed signals generated by downlink transmissions and evaluates key performance indicators, including Noise Equivalent Sigma Zero (NESZ) and channel capacity. Additionally, we utilize synchronization uplink signals transmitted by User Equipment (UE) to improve target detection and classification of possible victims by fusing SAR imagery with triangulation results from uplink signals. An experimental campaign validates the proposed setup by integrating SAR images of the environment with active localization results, both produced by a UAV equipped with a Software Defined Radio (SDR) payload. Our results demonstrate the system’s capability to detect and localize buried targets in avalanche scenarios, with localization errors ranging from centimeters to 10 m depending on environmental conditions. This successful integration highlights the practical applicability of our approach in challenging search-and-rescue missions.

Engineered transcription-associated Cas9 targeting in eukaryotic cells.

DNA targeting Class 2 CRISPR-Cas effector nucleases, including the well-studied Cas9 proteins, evolved protospacer-adjacent motif (PAM) and guide RNA interactions that sequentially license their binding and cleavage activities at protospacer target sites. Both interactions are nucleic acid sequence specific but function constitutively; thus, they provide intrinsic spatial control over DNA targeting activities but naturally lack temporal control. Here we show that engineered Cas9 fusion proteins which bind to nascent RNAs near a protospacer can facilitate spatiotemporal coupling between transcription and DNA targeting at that protospacer: Transcription-associated Cas9 Targeting (TraCT). Engineered TraCT is enabled in eukaryotic yeast or human cells when suboptimal PAM interactions limit basal activity and when one or more nascent RNA substrates are still tethered to the actively transcribed target DNA in cis. Using yeast, we further show that this phenomenon can be applied for selective editing at one of two identical targets in distinct gene loci, or, in diploid allelic loci that are differentially transcribed. Our work demonstrates that temporal control over Cas9's targeting activity at specific DNA sites may be engineered without modifying Cas9's core domains and guide RNA components or their expression levels. More broadly, it establishes co-transcriptional RNA binding as a cis-acting mechanism that can conditionally stimulate CRISPR-Cas DNA targeting in eukaryotic cells.

Image-free Hu invariant moment measurement by single-pixel detection

The global feature invariant moment plays a crucial role in capturing the feature information of a target. At present, the extraction of the target moment depends on the processing of the two-dimensional image, which leads to the high computational resource occupation due to the large amount of inherent data and the high information redundancy of the image. This reduces the real-time performance of target moment extraction to a certain extent. In order to solve this problem, we introduce a novel strategy for extracting Hu invariant moments, which is based on the combination of single-pixel detection technology and the Hu invariant moments theory. To the best of our knowledge, this is the first proposal of an image-free extraction strategy employing single-pixel detection. The extraction of moments is seamlessly integrated into the single-pixel detection process, facilitating efficient and rapid acquisition, followed by the construction of Hu invariant moments. Theoretical analysis and experimental validation of our proposed method are carried out, demonstrating the effective acquisition of Hu invariant moments. This technology holds substantial application value and prospects in various fields, including target analysis, understanding, classification, and recognition.

Underwater Acoustic Target Classification Using Scattering Transform With Small Sample Size

Underwater Acoustic Target Classification Using Scattering Transform With Small Sample Size

Top-Down Priors Disambiguate Target and Distractor Features in Simulated Covert Visual Search.

Several models of visual search consider visual attention as part of a perceptual inference process, in which top-down priors disambiguate bottom-up sensory information. Many of these models have focused on gaze behavior, but there are relatively fewer models of covert spatial attention, in which attention is directed to a peripheral location in visual space without a shift in gaze direction. Here, we propose a biologically plausible model of covert attention during visual search that helps to bridge the gap between Bayesian modeling and neurophysiological modeling by using (1) top-down priors over target features that are acquired through Hebbian learning, and (2) spatial resampling of modeled cortical receptive fields to enhance local spatial resolution of image representations for downstream target classification. By training a simple generative model using a Hebbian update rule, top-down priors for target features naturally emerge without the need for hand-tuned or predetermined priors. Furthermore, the implementation of covert spatial attention in our model is based on a known neurobiological mechanism, providing a plausible process through which Bayesian priors could locally enhance the spatial resolution of image representations. We validate this model during simulated visual search for handwritten digits among nondigit distractors, demonstrating that top-down priors improve accuracy for estimation of target location and classification, relative to bottom-up signals alone. Our results support previous reports in the literature that demonstrated beneficial effects of top-down priors on visual search performance, while extending this literature to incorporate known neural mechanisms of covert spatial attention.