Modification Of Architecture Research Articles

Causal Inference for Modality Debiasing in Multimodal Emotion Recognition

Multimodal emotion recognition (MER) aims to enhance the understanding of human emotions by integrating visual, auditory, and textual modalities. However, previous MER approaches often depend on a dominant modality rather than considering all modalities, leading to poor generalization. To address this, we propose Causal Inference in Multimodal Emotion Recognition (CausalMER), which leverages counterfactual reasoning and causal graphs to capture relationships between modalities and reduce direct modality effects contributing to bias. This allows CausalMER to make unbiased predictions while being easily applied to existing MER methods in a model-agnostic manner, without requiring any architectural modifications. We evaluate CausalMER on the IEMOCAP and CMU-MOSEI datasets, widely used benchmarks in MER, and compare it with existing methods. On the IEMOCAP dataset with the MulT backbone, CausalMER achieves an average accuracy of 83.4%. On the CMU-MOSEI dataset, the average accuracies with MulT, PMR, and DMD backbones are 50.1%, 48.8%, and 48.8%, respectively. Experimental results demonstrate that CausalMER is robust in missing modality scenarios, as shown by its low standard deviation in performance drop gaps. Additionally, we evaluate modality contributions and show that CausalMER achieves balanced contributions from each modality, effectively mitigating direct biases from individual modalities.

YOLOv10-based Model for Player and Football Detection

This study presents an advanced YOLOv10n-based method for the automatic detection of football players and balls directly from match videos. We enhance the YOLOv10 architecture with several significant improvements, including additional detection heads, the integration of C2f_faster and C3_faster modules for enhanced processing speed and accuracy, and the inclusion of BotNet modules with self-attention mechanisms for managing complex visual scenes. Further, we incorporate GhostConv modules to reduce computational overhead while maintaining effective feature extraction. These architectural modifications ensure robust detection capabilities in real-time sports environments, addressing challenges such as high-speed movements, frequent occlusions, and variable lighting conditions typical of both indoor and outdoor stadiums. Validation on internet-sourced images from football matches demonstrates the practicality and effectiveness of our model.

Molecular Decoration and Unconventional Double Bond Migration in Irumamycin Biosynthesis

Irumamycin (Iru) is a complex polyketide with pronounced antifungal activity produced by a type I polyketide (PKS) synthase. Iru features a unique hemiketal ring and an epoxide group, making its biosynthesis and the structural diversity of related compounds particularly intriguing. In this study, we performed a detailed analysis of the iru biosynthetic gene cluster (BGC) to uncover the mechanisms underlying Iru formation. We examined the iru PKS, including the domain architecture of individual modules and the overall spatial structure of the PKS, and uncovered discrepancies in substrate specificity and iterative chain elongation. Two potential pathways for the formation of the hemiketal ring, involving either an olefin shift or electrocyclization, were proposed and assessed using 18O-labeling experiments and reaction activation energy calculations. Based on our findings, the hemiketal ring is likely formed by PKS-assisted double bond migration and TE domain-mediated cyclization. Furthermore, putative tailoring enzymes mediating epoxide formation specific to Iru were identified. The revealed Iru biosynthetic machinery provides insight into the complex enzymatic processes involved in Iru production, including macrocycle sculpting and decoration. These mechanistic details open new avenues for a targeted architecture of novel macrolide analogs through synthetic biology and biosynthetic engineering.

Advancing graph neural network architecture for fluid flow and heat transfer surrogate modeling: Variable boundary conditions and geometry

Graph neural networks (GNNs) represent a promising instrument for surrogate modeling, capable of handling unstructured computational meshes naturally. We address a typical issue of the accuracy degradation for larger computational domains due to the limited receptive field of GNN models and long-range global interactions between nodes of the mesh. We propose a modification of the GNN architecture allowing to improve the accuracy by a factor of 3 without significant increase in computational costs. The validation tests of the model concentrate on the two-dimensional stationary fluid flow around a bluff body in a channel and corresponding heat transfer. The problem formulation includes bluff bodies of randomly generated shapes and various boundary conditions. The model shows a robust performance for the out-of-domain data, i.e., the flow over an airfoil for different angles of attack.

Text-independent speaker identification using modified SincNet with robust features from suitable acoustic region and appropriate optimizer for raw audio analysis

Speaker identification is a method of identifying an individual from a set of speakers, and text-independent speaker identification systems allow speakers to utter any phrase without any constraints. This study is focused on raw audio analysis as phase, fine-grained frequency patterns, timing cues, and other minute characteristics are preserved when raw waveforms are processed as compared to handcrafted features like Mel-Frequency Cepstral Coefficients (MFCC) and visual representation of audio-like spectrogram. Due to the depth of information, which includes variations in speech rhythm, pitch, and vocal tract shape, it is beneficial for identifying speakers. The deep learning architecture known as SincNet has gained popularity in speaker identification because of its parametric Sinc functions that allow it to operate directly on the raw audio input. In this paper, we have considered SincNet as the baseline model for speaker identification. The effect of proper speech boundary detection, including high-level features and effective optimizer selection are analysed. The precise identification of the signal start and terminus point is important for eliminating the redundant non-speech regions. We have included endpoint detection module as a pre-processing step in the system. Proper feature extraction and selection are crucial to the model's success. To extract more abstract features from the data, we have added more convolution layers to the original SincNet model. Further, we investigated the hyperparameter tuning protocol's sensitivity to the optimizer and selected the suitable optimizer for raw audio analysis. With all the modifications in the system architecture, we are able to archive improvements of 12.76 %, 13.33 %, and 13.39 % respectively for training, validation, and testing over the original SincNet model. In terms of validation loss, our proposed approach attains 0.35 in comparison to the original SincNet loss of 1.02. With this significant improvement, the total training time is marginally increased by 20 minutes for our proposed model. We have performed our investigation on the LibriSpeech dataset to check the effectiveness of our proposed system in comparison to the other model..

Comparative Assessment of Different Architectures of Convolutional Neural Network for Semantic Segmentation of Forest Disturbances from Multi-Temporal Satellite Images

Algorithms based on convolutional neural networks are the most efficient for semantic segmentation of images, including segmentation of forest cover disturbances from satellite images. In this study, we consider the applicability of various modifications of the U-net architecture of convolutional neural network for recognizing logged, burnt and windthrow areas in forests from multi-temporal and multi-seasonal Sentinel-2 satellite images. The assessment was carried out on three test sites that differ significantly in the characteristics of forest stands and forest management. The highest accuracy (average F-measure of 0.59) was obtained from the U-net model, while the models that showed the best results during training (Attention U-Net and MobilNetv2 U-Net) did not improve segmentation of independent data. The resulting accuracy estimates are close to those previously published for forests with a substantial proportion of selective logged areas. Characteristics of logged areas and windthrows, namely their area and type are the main factor determining the accuracy of semantic segmentation. Substantial differences were also revealed between the images taken in different seasons of the year, with the maximum segmentation accuracy based on winter pairs of images. According to summertime and different-season pairs of images, the area of forest disturbances is substantially underestimated. Forest species composition has a less significant effect, although for two of the three test sites, the maximum accuracy was observed in dark coniferous forests, and the minimum in deciduous forests. There was no statistically significant effect of slope lighting factor calculated from digital elevation model on segmentation accuracy based for winter pairs of images. The accuracy of segmentation of burnt areas, which was assessed using the example of 14 large forest fires in 2021-2022, is unsatisfactory, which is probably due to the varying degrees of damage to the forest cover in the burnt areas.

Water-Mediated Surface Engineering Enhances High-Voltage Stability of Fast-Charge LiCoO2 Cathodes.

Maintaining the surface structure stability of LiCoO2 (LCO) during rapid charge-discharge processes (>5C) and under high-voltage conditions (>4.2 V) is challenging due to interfacial side reactions, cobalt dissolution, and oxygen redox activity at deeply delithiated states, all of which contribute to performance degradation. Herein, different from traditional surface coating methods, we report a water-mediated strategy that modifies the surface architecture of LCO, creating a passivating layer to inhibit surface degradation and enhance cycling stability under fast charging conditions. The surface etching of LCO by H2O is accompanied by a concurrent Li+/H+ cation exchange, which passivates surface oxygen with H+ ions, thereby enhancing both the hydrophobicity and structural stability. Consequently, the modified LCO exhibits superior capacity retention, which is 2.5 times that of the pristine LCO, after 100 cycles at a current density of 1000 mA g-1 (∼6C at 4.5 V). Even at an elevated temperature of 45 °C, it maintains impressive cycling stability at a current density of 500 mA g-1 (∼3C), as demonstrated in practical full-cell configurations. Investigation with multiple samples confirmed that the water-mediated strategy demonstrated broad applicability. We emphasize that the water-mediated modification of the surface architecture on cathode materials offers significant insights into enhancing the stability of high-energy-density lithium-ion batteries (LIBs).

Using articulatory feature detectors in progressive networks for multilingual low-resource phone recognitiona).

Systems inspired by progressive neural networks, transferring information from end-to-end articulatory feature detectors to similarly structured phone recognizers, are described. These networks, connecting the corresponding recurrent layers of pre-trained feature detector stacks and newly introduced phone recognizer stacks, were trained on data from four Asian languages, with experiments testing the system on those languages and four African languages. Later adjustments of these networks include the use of contrastive predictive coding layers at the inputs to those networks' recurrent portions. Such adjustments allow for performance differences to be attributed to the presence or absence of individual feature detectors (for consonant place/manner and vowel height/backness). Some of these differences manifest after feature-level comparisons of recognizer outputs, as well as through considering variations and ablations in architecture and training setup. These differences encourage further exploration of methods to reduce errors with phones having specific articulatory features as well as further architectural modifications.

SCC-NET: segmentation of clinical cancer image for head and neck squamous cell carcinoma.

Squamous cell carcinoma (SCC) accounts for 90% of head and neck cancer. The majority of cases can be diagnosed and even treated with endoscopic examination and surgery. Deep learning models have been adopted for various medical endoscopy exams. However, few reports have been on deep learning algorithms for segmenting head and neck SCC. Head and neck SCC pre-treatment endoscopic images during 2016-2020 were collected from the Kaohsiung Veterans General Hospital Department of Otolaryngology-Head and Neck Surgery. We present a new modification of the neural architecture search-U-Net-based model called SCC-Net for segmenting our enrolled endoscopic photos. The modification included a new technique called "Learnable Discrete Wavelet Pooling" to design a new formulation that combines the outputs of different layers using a channel attention module and assigns weights based on their importance in the information flow. We also incorporated the cross-stage-partial design from CSPnet. The performance was compared with other eight state-of-the-art image segmentation models. We collected a total of 556 pathologically confirmed SCC photos. The new SCC-Net algorithm achieves a high mean intersection over union (mIOU) of 87.2%, accuracy of 97.17%, and recall of 97.15%. When comparing the performance of our proposed model with eight different state-of-the-art image segmentation artificial neural network models, our model performed best in mIOU, Dice similarity coefficient, accuracy, and recall. Our proposed SCC-Net architecture was able to successfully segment lesions from white light endoscopic images with promising accuracy, with a single model performing well in all upper aerodigestive tracts.

Improving Low-Latency Predictions in Multi-Exit Neural Networks via Block-Dependent Losses.

As the size of a model increases, making predictions using deep neural networks (DNNs) is becoming more computationally expensive. Multi-exit neural network is one promising solution that can flexibly make anytime predictions via early exits, depending on the current test-time budget which may vary over time in practice (e.g., self-driving cars with dynamically changing speeds). However, the prediction performance at the earlier exits is generally much lower than the final exit, which becomes a critical issue in low-latency applications having a tight test-time budget. Compared to the previous works where each block is optimized to minimize the losses of all exits simultaneously, in this work, we propose a new method for training multi-exit neural networks by strategically imposing different objectives on individual blocks. The proposed idea based on grouping and overlapping strategies improves the prediction performance at the earlier exits while not degrading the performance of later ones, making our scheme to be more suitable for low-latency applications. Extensive experimental results on both image classification and semantic segmentation confirm the advantage of our approach. The proposed idea does not require any modifications in the model architecture and can be easily combined with existing strategies aiming to improve the performance of multi-exit neural networks.

Hybrid Space Calibrated 3D Network of Diffractive Hyperspectral Optical Imaging Sensor.

Diffractive multispectral optical imaging plays an essential role in optical sensing, which typically suffers from the image blurring problem caused by the spatially variant point spread function. Here, we propose a novel high-quality and efficient hybrid space calibrated 3D network "HSC3D" for spatially variant diffractive multispectral imaging that utilizes the 3D U-Net structure combined with space calibration modules of magnification and rotation effects to achieve high-accuracy eight-channel multispectral restoration. The algorithm combines the advantages of the space calibrated module and U-Net architecture with 3D convolutional layers to improve the image quality of diffractive multispectral imaging without the requirements of complex equipment modifications and large amounts of data. A diffractive multispectral imaging system is established by designing and manufacturing one diffractive lens and four refractive lenses, whose monochromatic aberration is carefully corrected to improve imaging quality. The mean peak signal-to-noise ratio and mean structural similarity index of the reconstructed multispectral images are improved by 3.33 dB and 0.08, respectively, presenting obviously improved image quality compared with a typical Unrolled Network algorithm. The new algorithm with high space calibrated ability and imaging quality has great application potential in diffraction lens spectroscopy and paves a new method for complex practical diffractive multispectral image sensing.

Computer-aided Detection and Diagnosis of Cancer Regions in Mammogram Images Using Resource-Efficient CNN Architecture.

The automatic computer-assisted mammogram classification system is important for women patients to detect and diagnose the cancer regions. In this work, the mammogram images are classified into three cases: healthy, benign and cancer, using the proposed Resource Efficient Convolutional Neural Network (RECNN architecture). The proposed mammogram image classification system consists of Data Augmentation (DA) module and Spatial transformation module and CNN architecture with a segmentation module. The DA methods are used to increase the mammogram image count and Spatial Gabor Transform is used as the spatial transformation module for transforming the spatial pixels into spatial-frequency pixels. Then, the proposed RECNN architecture is used to perform the classification of mammogram images into healthy, benign and cancer cases. Further, the cancer mammogram images are diagnosed as either 'Early' or 'Severe' using the proposed RECNN architecture in this work. The proposed MCDS obtains 98.65% SeDR, 98.93% SpDR and 98.84% ADR for benign case mammogram images on DDSM dataset and also obtains 98.84% SeDR, 98.7% SpDR and 98.92% ADR for benign case mammogram images on DDSM dataset. The proposed MCDS obtains 98.94% SeDR, 98.86% SpDR and 98.96% ADR for benign case mammogram images on DDSM dataset and also obtains 98.89% SeDR, 98.88% SpDR and 99.03% ADR for benign case mammogram images on MIAS dataset. This proposed method is tested on the mammogram images from DDSM and MIAS datasets and the experimental results are compared with other similar mammogram classification works in this paper. Based on several performance evaluation measures, the experimental results show that MCDS outperforms the state-of-the-art methods currently used for the diagnosis and detection of mammography cancer.

Hybrid Reconstruction Approach for Polychromatic Computed Tomography in Highly Limited-Data Scenarios.

Conventional strategies aimed at mitigating beam-hardening artifacts in computed tomography (CT) can be categorized into two main approaches: (1) postprocessing following conventional reconstruction and (2) iterative reconstruction incorporating a beam-hardening model. While the former fails in low-dose and/or limited-data cases, the latter substantially increases computational cost. Although deep learning-based methods have been proposed for several cases of limited-data CT, few works in the literature have dealt with beam-hardening artifacts, and none have addressed the problems caused by randomly selected projections and a highly limited span. We propose the deep learning-based prior image constrained (PICDL) framework, a hybrid method used to yield CT images free from beam-hardening artifacts in different limited-data scenarios based on the combination of a modified version of the Prior Image Constrained Compressed Sensing (PICCS) algorithm that incorporates the L2 norm (L2-PICCS) with a prior image generated from a preliminary FDK reconstruction with a deep learning (DL) algorithm. The model is based on a modification of the U-Net architecture, incorporating ResNet-34 as a replacement of the original encoder. Evaluation with rodent head studies in a small-animal CT scanner showed that the proposed method was able to correct beam-hardening artifacts, recover patient contours, and compensate streak and deformation artifacts in scenarios with a limited span and a limited number of projections randomly selected. Hallucinations present in the prior image caused by the deep learning model were eliminated, while the target information was effectively recovered by the L2-PICCS algorithm.

Long-read transcriptomics of Ostreid herpesvirus 1 uncovers a conserved expression strategy for the capsid maturation module and pinpoints a mechanism for evasion of the ADAR-based antiviral defence.

Ostreid herpesvirus 1 (OsHV-1), a member of the family Malacoherpesviridae (order Herpesvirales), is a major pathogen of bivalves. However, the molecular details of the malacoherpesvirus infection cycle and its overall similarity to the replication of mammalian herpesviruses (family Orthoherpesviridae) remain obscure. Here, to gain insights into the OsHV-1 biology, we performed long-read sequencing of infected blood clams, Anadara broughtonii, which yielded over one million OsHV-1 long reads. These data enabled the annotation of the viral genome with 78 gene units and 274 transcripts, of which 67 were polycistronic mRNAs, 35 ncRNAs, and 20 natural antisense transcripts (NATs). Transcriptomics and proteomics data indicate preferential transcription and independent translation of the capsid scaffold protein as an OsHV-1 capsid maturation protease isoform. The conservation of this transcriptional architecture across Herpesvirales likely indicates its functional importance and ancient origin. Moreover, we traced RNA editing events using short-read sequencing and supported the presence of inosine nucleotides in native OsHV-1 RNA, consistent with the activity of adenosine deaminase acting on dsRNA 1 (ADAR1). Our data suggest that, whereas RNA hyper-editing is concentrated in specific regions of the OsHV-1 genome, single-nucleotide editing is more dispersed along the OsHV-1 transcripts. In conclusion, we reveal the existence of conserved pan-Herpesvirales transcriptomic architecture of the capsid maturation module and uncover a transcription-based viral counter defence mechanism, which presumably facilitates the evasion of the host ADAR antiviral system.

The Architectural Intersection of Museums and Disability Policy

The way in which people with disabilities devise methods to move through the world and faced barriers before architectural modifications evolved based on legislation intersects with the architectural preservation of culturally significant sites, specifically museums. By excluding the disabled community in museum spaces, it perpetuates the ableist mindset of society further by limiting information, creating a disparity in cultural climate and community belonging. How do museums preserve historical significance, while also creating a space for all? In order to devise an accepting space within museums and go beyond the idea of inclusion, people within public spaces need to strive to understand the way that individuals interact with the institution, while adjusting the space in an inherent way to include access for all beyond the legal policy.

Multibarrier Collaborative Modulation Devices with Ultra-High Logic Operation Density.

The demands for highly miniaturized and multifunctional electronics are rapidly increasing. As scaling-down processes of transistors are restricted by physical limits, reconfigurable electronics with switchable operation functions for different tasks are developed for higher function integration based on split- or vertical-dual-gate structures. To promote the present reconfigurable electronics and exceed the function integration limit, the critical issue is to integrate complex operations into simple circuit forms by establishing more control dimensions. This work proposes a multibarrier collaborative (MBC) modulation architecture to increase the control dimension by multiple forms of potential barriers and achieves combinational and reconfigurable logic operations by a single MBC device. The MBC architecture exhibits ultrahigh logic operation density, including 58.8% area reduction for multiplexer operations and 71.4% area reduction for 4-logic reconfigurable operations. Besides, a hardware security module composed of 4 MBC devices implementing 8 types of logic operations is demonstrated. This work reveals an effective design of function integration for next-generation electronics.

Versatile Spiral Donor‐Based Thermally Activated Delayed Fluorescence Materials for Highly Efficient TADF‐OLED and TSF‐OLED Applications

AbstractHerein, a design strategy is explored for thermally activated delayed fluorescence (TADF) materials by employing the meta‐linkage of the spiral‐donors 10H‐spiro[acridine‐9,9'‐thioxanthene] (DspiroS) and 10',10'‐dimethyl‐10H,10'H‐spiro[acridine‐9,9'‐anthracene] (DspiroAc) to the robust acceptor 2,4,6‐triphenyl‐1,3,5‐triazine (TRZ). Two distinct TADF materials, m‐DspiroS‐TRZ and m‐DspiroAc‐TRZ, exhibiting unique photophysical properties and performance characteristics were synthesized. Interestingly, even subtle modifications in the molecular architecture can significantly impact the organization of materials in their aggregated state, thereby governing photophysical properties and inducing corresponding alterations in photoelectric characteristics. Notably, m‐DspiroS‐TRZ exhibits superior photophysical properties and exciton dynamics data, achieving a high photoluminescence quantum yield (PLQY) value of up to 95.9% and a rapid reverse intersystem crossing (RISC) rate (𝒌𝑹𝑰𝑺𝑪) of 1.0 × 106 s−1. This positions m‐DspiroS‐TRZ as a potentially excellent terminal emissive and sensitizing host material, inspiring further exploration of its applications in electroluminescence. Consequently, TADF organic light‐emitting device (TADF‐OLED) and TADF‐sensitized fluorescence (TSF‐OLED) based on m‐DspiroS‐TRZ have achieved maximum external quantum efficiencies (EQEs) of 31.8% and 34.5%, respectively, demonstrating the significant versatile potential of m‐DspiroS‐TRZ.

SIMIL: SIMple Issue Logic for GPUs

GPU architectures have become popular for executing general-purpose programs. In particular, they are some of the most efficient architectures for machine learning applications which are among the most trendy and demanding applications nowadays.This paper presents SIMIL (SIMple Issue Logic for GPUs), an architectural modification to the issue stage that replaces scoreboards with a Dependence Matrix to track dependencies among instructions and avoid data hazards. We show that a Dependence Matrix is more effective in the presence of repetitive use of source operands, which is common in many applications. Besides, a Dependence Matrix with minor extensions can also support a simplistic out-of-order issue. Evaluations on an NVIDIA Tesla V100-like GPU show that SIMIL provides a speed-up of up to 2.39 in some machine learning programs and 1.31 on average for various benchmarks, while it reduces energy consumption by 12.81%, with only 1.5% area overhead. We also show that SIMIL outperforms a recently proposed approach for out-of-order issue that uses register renaming.

A novel lightweight deep learning model based on SqueezeNet architecture for viral lung disease classification in X-ray and CT images

COVID-19 has affected hundreds of millions of individuals, seriously harming the global population’s health, welfare, and economy. Furthermore, health facilities are severely overburdened due to the record number of COVID-19 cases, which makes prompt and accurate diagnosis difficult. Automatically identifying infected individuals and promptly placing them under special care is a critical step in reducing the burden of such issues. Convolutional Neural Networks (CNN) and other machine learning techniques can be utilized to address this demand. Many existing Deep learning models, albeit producing the intended outcomes, were developed using millions of parameters, making them unsuitable for use on devices with constrained resources. Motivated by this fact, a novel lightweight deep learning model based on Efficient Channel Attention (ECA) module and SqueezeNet architecture, is developed in this work to identify COVID-19 patients from chest X-ray and CT images in the initial phases of the disease. After the proposed lightweight model was tested on different datasets with two, three and four classes, the results show its better performance over existing models. The outcomes shown that, in comparison to the current heavyweight models, our models reduced the cost and memory requirements for computing resources dramatically, while still achieving comparable performance. These results support the notion that proposed model can help diagnose Covid-19 in patients by being easily implemented on low-resource and low-processing devices.

Rice homolog of Arabidopsis Xylem NAC domain 1 (OsXND1), a NAC transcription factor regulates drought stress responsive root system architecture in indica rice.

Rice yield is greatly constrained by drought stress. In Arabidopsis, XYLEM NAC DOMAIN 1 (XND1) gene regulates the xylem formation, efficiency of water transport, and the delicate equilibrium between drought tolerance and resistance to pathogens. However, diversity and the role of rice homologs of OsXND1 is not reported so far. This study hypothesized that the rice homolog of OsXND1 also regulates drought stress tolerance through modulation of root architecture. Initially, phylogenetic analysis identified two OsXND1 homologs (Os02g0555300 and Os04g0437000) in rice. Further, 14 haplotypes were identified in the OsXND1 of which Hap1 and Hap3 were major haplotypes. The association analysis of OsXND1 with 16 different traits, including 10 root traits, showed three SNPs (Chr02:20972728-Promoter variant; Chr02:20972791-5' UTR variant, and Chr02:20973745-3' UTR variant) were significantly associated with root area, root surface area, total root length, and convex hull area only under drought stress in indica rice. Besides, the superior haplotype of OsXND1 increased the root area, root surface area, total root length, and convex hull area by 46%, 40%, 38%, and 42%, respectively, under drought stress conditions. Therefore, the identified superior haplotype of OsXND1 can be utilized in haplotype breeding programs for the improvement of drought tolerance in rice.