Deep Neural Network Framework Research Articles

A novel fusion framework of deep bottleneck residual convolutional neural network for breast cancer classification from mammogram images.

With over 2.1 million new cases of breast cancer diagnosed annually, the incidence and mortality rate of this disease pose severe global health issues for women. Identifying the disease's influence is the only practical way to lessen it immediately. Numerous research works have developed automated methods using different medical imaging to identify BC. Still, the precision of each strategy differs based on the available resources, the issue's nature, and the dataset being used. We proposed a novel deep bottleneck convolutional neural network with a quantum optimization algorithm for breast cancer classification and diagnosis from mammogram images. Two novel deep architectures named three-residual blocks bottleneck and four-residual blocks bottle have been proposed with parallel and single paths. Bayesian Optimization (BO) has been employed to initialize hyperparameter values and train the architectures on the selected dataset. Deep features are extracted from the global average pool layer of both models. After that, a kernel-based canonical correlation analysis and entropy technique is proposed for the extracted deep features fusion. The fused feature set is further refined using an optimization technique named quantum generalized normal distribution optimization. The selected features are finally classified using several neural network classifiers, such as bi-layered and wide-neural networks. The experimental process was conducted on a publicly available mammogram imaging dataset named INbreast, and a maximum accuracy of 96.5% was obtained. Moreover, for the proposed method, the sensitivity rate is 96.45, the precision rate is 96.5, the F1 score value is 96.64, the MCC value is 92.97%, and the Kappa value is 92.97%, respectively. The proposed architectures are further utilized for the diagnosis process of infected regions. In addition, a detailed comparison has been conducted with a few recent techniques showing the proposed framework's higher accuracy and precision rate.

Polarization-based all-optical logic gates using diffractive neural networks

Optical logic operations are an essential part of optical computing. The inherent stability and low susceptibility of polarization to the external environment make it a suitable choice for acting as the logical state in computational tasks. Traditional polarization-based optical logic devices often rely on complex cascading structures to implement multiple logic gates. In this work, by leveraging the framework of deep diffractive neural networks (D2NN), we proposed a uniform approach to designing polarization-encoded all-optical logic devices with simpler and more flexible structures. We have implemented AND, OR, NOT, NAND, and NOR gates, as well as High-order Selector and Low-order Selector. These polarization-based all-optical logic devices using D2NN offer passive nature, stability, and high extinction ratio features, paving the way for a broader exploration of optical logic computing in the future.

Multi-sensor Data-driven Route Prediction in Instant Delivery with a 3-Conversion Network

Route prediction in instant delivery is still challenging due to the unique characteristics compared with conventional delivery services, such as strict deadlines, overlapped delivery time of multiple orders, and diverse individual preferences on delivery routes. Recently, development in the mobile Internet of Things (IoT) offers the opportunity to collect multi-sensor data with rich real-time information. Therefore, this study proposes a route prediction model called Roupid, which leverages multi-sensor data to improve the accuracy of route prediction in instant delivery. Specifically, we design a 3-Conversion Network-based route prediction framework to take full advantage of various information provided by multi-sensor data, including the encounter data sensed by Bluetooth low energy (BLE) beacons, active site data reported by smart handheld devices, and trajectory data detected by GPS. The 3-Conversion Network we propose is based on a deep neural network framework, which integrates an improved relational graph attention network with edge features (RGATE) to encode global information that couriers typically consider when planning routes. We evaluate our Roupid with real-world data collected from one of the largest instant delivery companies in the world, i.e., Eleme. Experimental results show that our Roupid outperforms other state-of-the-art baselines and offers up to 85.51% of the route prediction precision.

Simultaneously improving accuracy and computational cost under parametric constraints in materials property prediction tasks

Modern data mining techniques using machine learning (ML) and deep learning (DL) algorithms have been shown to excel in the regression-based task of materials property prediction using various materials representations. In an attempt to improve the predictive performance of the deep neural network model, researchers have tried to add more layers as well as develop new architectural components to create sophisticated and deep neural network models that can aid in the training process and improve the predictive ability of the final model. However, usually, these modifications require a lot of computational resources, thereby further increasing the already large model training time, which is often not feasible, thereby limiting usage for most researchers. In this paper, we study and propose a deep neural network framework for regression-based problems comprising of fully connected layers that can work with any numerical vector-based materials representations as model input. We present a novel deep regression neural network, iBRNet, with branched skip connections and multiple schedulers, which can reduce the number of parameters used to construct the model, improve the accuracy, and decrease the training time of the predictive model. We perform the model training using composition-based numerical vectors representing the elemental fractions of the respective materials and compare their performance against other traditional ML and several known DL architectures. Using multiple datasets with varying data sizes for training and testing, We show that the proposed iBRNet models outperform the state-of-the-art ML and DL models for all data sizes. We also show that the branched structure and usage of multiple schedulers lead to fewer parameters and faster model training time with better convergence than other neural networks. Scientific contribution: The combination of multiple callback functions in deep neural networks minimizes training time and maximizes accuracy in a controlled computational environment with parametric constraints for the task of materials property prediction.

Beyond images: an integrative multi-modal approach to chest x-ray report generation.

Image-to-text radiology report generation aims to automatically produce radiology reports that describe the findings in medical images. Most existing methods focus solely on the image data, disregarding the other patient information accessible to radiologists. In this paper, we present a novel multi-modal deep neural network framework for generating chest x-rays reports by integrating structured patient data, such as vital signs and symptoms, alongside unstructured clinical notes. We introduce a conditioned cross-multi-head attention module to fuse these heterogeneous data modalities, bridging the semantic gap between visual and textual data. Experiments demonstrate substantial improvements from using additional modalities compared to relying on images alone. Notably, our model achieves the highest reported performance on the ROUGE-L metric compared to relevant state-of-the-art models in the literature. Furthermore, we employed both human evaluation and clinical semantic similarity measurement alongside word-overlap metrics to improve the depth of quantitative analysis. A human evaluation, conducted by a board-certified radiologist, confirms the model's accuracy in identifying high-level findings, however, it also highlights that more improvement is needed to capture nuanced details and clinical context.

Detecting Multiple Mixed Bacteria Using Dual-Mode Hyperspectral Imaging and Deep Neural Networks

Identifying and analyzing mixed pathogenic bacteria is important for clinical diagnosis and antibiotic therapy of multiple bacterial infection. In this paper, a dual-mode hyperspectral microscopic detection technology with hybrid deep neural networks (DNNs) was proposed for simultaneous quantitative analysis of four kinds of pathogenic bacteria in mixed samples. To acquire both transmission and fluorescence spectra regarding the mixed pathogens, we developed a dual-mode hyperspectral detection system with fine spectral resolution and wide wavelength range, which can also generate spatial images that can be used to calculate the total amount of mixed bacteria. The dual-mode spectra were regarded as mixed proportion characteristics and the input of the neural network for predicting the proportion of each bacterium present in the mixture. To better analyze the dual-mode spectral data, we customized a mixed bacteria measurement network (MB-Net) with hybrid DNNs architectures based on spectral feature fusion. Using the fusion strategy, two DNNs frameworks applied for transmission/fluorescence spectral feature processing were stacked to form the MB-Net that processes these features simultaneously, and the achieved average coefficient of determination (R2) and RMSE of validation set are 0.96 and 0.03, respectively. To the best of our knowledge, it is the first time of simultaneously detecting four types of mixed pathogenic bacteria using spectral detection technology, showing excellent potential in clinical practice.

Developing a Deep Learning and Reliable Optimization Techniques for Solar Photovoltaic Power Prediction

Due to its significant solar energy generation, solar PV power facilities have recently been utilized. Although PV power stations are highly preferable, the state’s fundamental drawback is that its output current qualities are unpredictable. Consequently, to ensure a balance and complete functioning, it would be crucial to build systems that allow accurate future projections of solar PV production in the short or medium term. Research postulates a strategy for using deep learning to estimate the short-term electricity generated by solar photovoltaic facilities. This study offers a novel method for predicting photovoltaic systems output power utilizing a Hybrid Deep Neural Network framework, making significant advancements in the field of deep learning applications to transmission system prediction issues. CNN and LSTM are combined in the postulated HDNN paradigm. Traditional deep learning techniques are employed in the initial stage. The effectiveness assessments of these techniques are instead presented in greater detail after they have been trained using firefly optimization techniques. The method with the highest reliability is chosen out of all the techniques used in previous research. Deep learning and power efficiency create a combination that appears to have a successful future, predominantlyin improving sustainable management and the digitalization of the electrical sector.

Learning Embedding for Signed Network in Social Media With Global Information

Signed networks contain nodes connected by positive and negative signed links. Signed network representation learning concentrates on learning the low-dimensional representations of these nodes, which facilitates downstream tasks such as link prediction using general data mining framework. However, most signed network embedding (SiNE) approaches neglect global information in the networks, limiting the capacity to learn genuine signed graph topology. To overcome this limitation, a deep graph neural network (GNN) framework SiG to learn SiNE with global information is proposed. To be more explicit, a hierarchical pooling mechanism is developed to encode the high-level features of the networks. Moreover, a graph convolution layer is introduced to aggregate both positive and negative information from neighbor nodes, and the concatenation of two parts generates the final embedding of nodes. Extensive experiments on four large real-world signed network datasets demonstrate the effectiveness and excellence of the proposed method.

Implementation of AI Based Model for Identification of Different Soil Types Using Tensorflow

AbstractThe agricultural industry is among the most important industry in the world and playing a critical part in human growth. Soil classification is becoming incredibly valuable, and recent study reveals that classification of images is proven as the preferred way of soil information for farmers. The testing of soil is the first step in construction planning to determine whether the plot of land is suitable for constructing any structure in order to prevent future disasters. In this work, the model has been proposed to address the visual classification of types of soil using the deep neural network (DNN) and tensor flow framework. Further, AI based sequential model has been created in python to classify the soil. This has been analyzed that with the use of proposed technique five types of soil viz. black, cinder, laterite, peat, and yellow soil has been classified while keeping the practical simplicity of the images. Far superior accuracy has been achieved using the CNN model on the above dataset. The model is useful for farmers to increase the crop production after identification of the appropriate soil type for respective crops.

EquiPNAS: improved protein-nucleic acid binding site prediction using protein-language-model-informed equivariant deep graph neural networks.

Protein language models (pLMs) trained on a large corpus of protein sequences have shown unprecedented scalability and broad generalizability in a wide range of predictive modeling tasks, but their power has not yet been harnessed for predicting protein-nucleic acid binding sites, critical for characterizing the interactions between proteins and nucleic acids. Here, we present EquiPNAS, a new pLM-informed E(3) equivariant deep graph neural network framework for improved protein-nucleic acid binding site prediction. By combining the strengths of pLM and symmetry-aware deep graph learning, EquiPNAS consistently outperforms the state-of-the-art methods for both protein-DNA and protein-RNA binding site prediction on multiple datasets across a diverse set of predictive modeling scenarios ranging from using experimental input to AlphaFold2 predictions. Our ablation study reveals that the pLM embeddings used in EquiPNAS are sufficiently powerful to dramatically reduce the dependence on the availability of evolutionary information without compromising on accuracy, and that the symmetry-aware nature of the E(3) equivariant graph-based neural architecture offers remarkable robustness and performance resilience. EquiPNAS is freely available at https://github.com/Bhattacharya-Lab/EquiPNAS.

An integrated deep neural network framework for predicting the wake flow in the wind field

Ultra-short-term wake flow prediction is crucial for wind resource assessment and wind farm operation control. To improve the power generation efficiency and stable operation level of wind farms, a kind of prediction method is proposed that integrates the physical model and mathematical model into a deep neural network, enabling the prediction of the precise wake flow with sparse measured data. The proposed method can predict the entire flow field in real-time, providing accurate and reliable predictions for wind farm operation and management. The results of evaluation and validation of the integrated method show that the proposed method can accurately achieve ultra-short-term prediction, with a small error in all directions of velocity. Compared with the widely used LSTM neural network model and Multilayer Perceptron, there are certain advantages in both spatial and temporal scales, with a significant reduction in the average absolute error, indicating better generalization performance and prediction accuracy in the prediction of the wake flow field.

High-speed and energy-efficient non-volatile silicon photonic memory based on heterogeneously integrated memresonator.

Recently, interest in programmable photonics integrated circuits has grown as a potential hardware framework for deep neural networks, quantum computing, and field programmable arrays (FPGAs). However, these circuits are constrained by the limited tuning speed and large power consumption of the phase shifters used. In this paper, we introduce the memresonator, a metal-oxide memristor heterogeneously integrated with a microring resonator, as a non-volatile silicon photonic phase shifter. These devices are capable of retention times of 12 hours, switching voltages lower than 5 V, and an endurance of 1000 switching cycles. Also, these memresonators have been switched using 300 ps long voltage pulses with a record low switching energy of 0.15 pJ. Furthermore, these memresonators are fabricated on a heterogeneous III-V-on-Si platform capable of integrating a rich family of active and passive optoelectronic devices directly on-chip to enable in-memory photonic computing and further advance the scalability of integrated photonic processors.

Prediction method of longitudinal surface settlement caused by double shield tunnelling based on deep learning

The deep learning method faces the challenges of small sample data and high dimensional shield operational parameters in predicting the longitudinal surface settlement caused by shield excavation. In this study, various optimization algorithms were compared, and the slime mould algorithm (SMA) was optimally chosen to optimize the hyperparameters of random forest (RF), and SMA-RF was used for dimensionality reduction and feature contribution analysis. A double-input deep neural network (D-DNN) framework was proposed for the prediction of surface settlement, which considers the influence of twin tunnels and effectively increases the high-fidelity data in the database. The results show that SMA performs best among various optimization algorithms; employing features that have a cumulative contribution value exceeding 90% as input can result in high prediction accuracy; there is significant uncertainty in the feature contribution analysis for small sample data; the reduced shield running parameters show a strong nonlinear relationship with surface settlement; compared with S-DNN, D-DNN takes into account the excavation of twin tunnels and expands the database capacity by more than 1.5 times, with an average increase of 27.85% in the R2 and an average decrease of 53.2% in the MAE.

Enhancing facial recognition accuracy through multi-scale feature fusion and spatial attention mechanisms

<abstract><p>Nowadays, advancements in facial recognition technology necessitate robust solutions to address challenges in real-world scenarios, including lighting variations and facial position discrepancies. We introduce a novel deep neural network framework that significantly enhances facial recognition accuracy through multi-scale feature fusion and spatial attention mechanisms. Leveraging techniques from FaceNet and incorporating atrous spatial pyramid pooling and squeeze-excitation modules, our approach achieves superior accuracy, surpassing 99% even under challenging conditions. Through meticulous experimentation and ablation studies, we demonstrate the efficacy of each component, highlighting notable improvements in noise resilience and recall rates. Moreover, the introduction of the Feature Generative Spatial Attention Adversarial Network (FFSSA-GAN) model further advances the field, exhibiting exceptional performance across various domains and datasets. Looking forward, our research emphasizes the importance of ethical considerations and transparent methodologies in facial recognition technology, paving the way for responsible deployment and widespread adoption in the security, healthcare, and retail industries.</p></abstract>

Cross-domain prediction approach of human lower limb voluntary movement intention for exoskeleton robot based on EEG signals.

Exoskeleton robot control should ideally be based on human voluntary movement intention. The readiness potential (RP) component of the motion-related cortical potential is observed before movement in the electroencephalogram and can be used for intention prediction. However, its single-trial features are weak and highly variable, and existing methods cannot achieve high cross-temporal and cross-subject accuracies in practical online applications. Therefore, this work aimed to combine a deep convolutional neural network (CNN) framework with a transfer learning (TL) strategy to predict the lower limb voluntary movement intention, thereby improving the accuracy while enhancing the model generalization capability; this would also provide sufficient processing time for the response of the exoskeleton robotic system and help realize robot control based on the intention of the human body. The signal characteristics of the RP for lower limb movement were analyzed, and a parameter TL strategy based on CNN was proposed to predict the intention of voluntary lower limb movements. We recruited 10 subjects for offline and online experiments. Multivariate empirical-mode decomposition was used to remove the artifacts, and the moment of onset of voluntary movement was labeled using lower limb electromyography signals during network training. The RP features can be observed from multiple data overlays before the onset of voluntary lower limb movements, and these features have long latency periods. The offline experimental results showed that the average movement intention prediction accuracy was 95.23% ± 1.25% for the right leg and 91.21% ± 1.48% for the left leg, which showed good cross-temporal and cross-subject generalization while greatly reducing the training time. Online movement intention prediction can predict results about 483.9 ± 11.9ms before movement onset with an average accuracy of 82.75%. The proposed method has a higher prediction accuracy with a lower training time, has good generalization performance for cross-temporal and cross-subject aspects, and is well-prioritized in terms of the temporal responses; these features are expected to lay the foundation for further investigations on exoskeleton robot control.

Sustainable Drone Detection using Deep Learning Paradigms

Drones are becoming more and more common, which has advantages but also concerns. They can be used to support illegal operations like drug trafficking and endanger places that are important to security. While advances in sensor technology haven't produced reliable answers in the literature, current drone detection and neutralization methods frequently require previous detection and categorization. Using radio frequency (RF) signals and a frequency signature-based deep learning model, this work promotes an environmentally friendly and multidisciplinary method of drone identification and categorization. Using commercial drones, we generated a novel drone RF dataset and thoroughly evaluated two-stage and combined detection-classification frameworks, assessing their efficacy in simultaneous multi-signal and single-signal scenarios. According to our investigation, the Deep Residual Neural Network (DRNN) framework achieves classification performance that is equivalent to that of the YOLOV5 framework, but it offers improved detection performance in multi-signal settings when compared to traditional Goodness-of-Fit (GoF) spectrum sensing. This study opens the door for multidisciplinary efforts to address new security concerns by showcasing the possibility of long-term, data-driven drone detection methods.

The Deep Promotion Time Cure Model.

We propose a novel method for predicting time-to-event data in the presence of cure fractions based on flexible survival models integrated into a deep neural network (DNN) framework. Our approach allows for nonlinear relationships and high-dimensional interactions between covariates and survival and is suitable for large-scale applications. To ensure the identifiability of the overall predictor formed of an additive decomposition of interpretable linear and nonlinear effects and potential higher-dimensional interactions captured through a DNN, we employ an orthogonalization layer. We demonstrate the usefulness and computational efficiency of our method via simulations and apply it to a large portfolio of U.S. mortgage loans. Here, we find not only a better predictive performance of our framework but also a more realistic picture of covariate effects.

Deep Learning for Fast and Reliable Initial Access in AI-Driven 6G mm Wave Networks

We present DeepIA, a deep neural network (DNN) framework for fast and reliable initial access (IA) for artificial intelligence (AI)-driven 6G millimeter wave (mmWave) networks. DeepIA reduces the beam sweep time compared to a conventional exhaustive search-based IA process by utilizing only a subset of the available beams. DeepIA maps received signal strengths (RSSs) obtained from a subset of beams to the beam that is best oriented to the receiver. In both line of sight (LoS) and non-line of sight (NLoS) conditions, DeepIA reduces the IA time and outperforms the conventional IA&#x0027;s beam prediction accuracy. We show that DeepIA&#x0027;s accuracy saturates with the number of beams used, and also depends on the particular choice of the beams used. The choice of beams that are selected is consequential and improves accuracy by upto upto <inline-formula><tex-math notation="LaTeX">$35\%$</tex-math></inline-formula> and <inline-formula><tex-math notation="LaTeX">$70\%$</tex-math></inline-formula> in NLoS and LoS channels. We find that, averaging multiple RSS snapshots further reduces the number of beams needed and achieves more than <inline-formula><tex-math notation="LaTeX">$95\%$</tex-math></inline-formula> accuracy in both LoS and NLoS conditions. We introduce interference into our models to understand impact on performance. Finally, we evaluate the beam prediction time of DeepIA through embedded hardware implementation and show the improvement over the baseline approach.

Deep Reinforcement Learning-Based Scheduling for NR-U/WiGig Coexistence in Unlicensed mmWave Bands

This paper investigates the coexistence of the New Radio-based access to Unlicensed spectrum (NR-U) network and the Wireless Gigabit (WiGig) network in unlicensed millimeter-wave (mmWave) bands. To enable the NR-U network to achieve equitable and harmonious spectrum sharing with WiGig systems, we develop two new classes of user equipment (UE) scheduling schemes by exploiting the deep reinforcement learning (DRL) technique. Specifically, we first propose the distributed deep reinforcement learning scheduling (DeepDS) scheme, wherein multiple deep neural networks (DNNs) are used to make decisions for different panels in an independent fashion. Thereafter, to reduce the computational cost of adopting multiple DNNs, we design the centralized deep reinforcement learning scheduling (DeepCS) scheme that introduces the shared DNN framework to perform decisions for all panels in parallel at one time. The objective of both DeepDS and DeepCS is to maximize the total data rate of the NR-U network with as little interference to WiGig systems as possible, while satisfying the quality of service (QoS) requirement for each UE. We first formulate this problem into the constrained Markov decision process framework. To address the multi-constraint issue, we put forth a new DRL algorithm that incorporates the Lagrangian primal-dual optimization into the deep Q-network framework, referred to as adaptive multi-constraint deep Q-network (AMC-DQN). With AMC-DQN, both DeepDS and DeepCS can achieve their goals even without acquiring prior operations about the WiGig network. Simulation results show that compared with the state-of-the-art omniLBT and dirLBT, both DeepDS and DeepCS yield significant performance benefits in terms of the total network data rate. We also demonstrate the ability of DeepDS and DeepCS to satisfy the QoS requirements of different UEs and their robustness against various simulation setups. Furthermore, compared with DeepDS, DeepCS can save a large amount of computational cost although at the expense of a slightly lower data rate.

A Deep Neural Network Framework for Landslide Susceptibility Mapping by Considering Time-Series Rainfall

A Deep Neural Network Framework for Landslide Susceptibility Mapping by Considering Time-Series Rainfall