Multilayer Perception Network Research Articles

Design and Simulation-Based Validation of an AI Model for Predicting Grab-Type Ship Unloader Operation Data

Along with seaports automation, there is growing interest in the automation of Grab-Type Ship Unloader (GTSU) that unloads coal and iron ore from bulk carriers. Autonomous unloading operations of GTSU offer the potential for significant productivity improvement and cost savings. In this paper, an AI model trained with manual operation data was designed for GTSU automation operation, and the AI model was verified through the equation-of-motion-based GTSU operation simulator. The operation data of hoist, grab, and trolley were predicted by training the designed AI model with the manual operation data of GTSU. Before applying the predicted data to the actual equipment, the predicted driving data was verified using the equation-of-motion-based GTSU operation simulator. The AI prediction model was designed using the Multi-Layer Perception network, a type of artificial neural network. The AI prediction model was evaluated with the Mean-Squared Error indicator, and the validation loss was found to be less than 0.02. In addition, verification of the prediction data was performed using the GTSU dynamics-based simulator. The Mean Relative Error was up to 0.50, and the R2 score value exceeded 0.92, indicating that the model is effective in predicting operation data. The proposed AI prediction model will be effectively utilized to implement a fully automated unloading system.

Predictive modeling of specific fuel consumption in compression ignition engines using neural networks: a comparative analysis across diesel and polymer-based fuels

The study utilized neural network modeling to forecast the fuel consumption in compression ignition engines fueled by diesel, HDPE PO, PP PO, and LDPE PO. Using empirical data, a neural network model was constructed and used to estimate specific fuel consumption (SFC). Employing orthogonal arrays and parameter adjustments ensured accurate prediction of SFC, which was validated through experimentation. The multilayer perception network, coupled with traditional backpropagation, facilitates the nonlinear mapping of inputs to outcomes. In the LM10TP architecture, the key metrics from the training set included an impressive R-squared value of 1, indicating a perfect fit with a root mean square error (RMSE) of 0.0012 and a mean square error (MSE) of 1.5143E-06. Similarly, the validation set exhibited robust performance metrics with an R-squared value of 0.9999, an RMSE of 0.0011, and an MSE of 1.2185E-06. These metrics underscore the efficacy of neural networks in both the training and validation phases, affirming their utility as reliable predictive tools for SFC. Overall, this study highlights the effectiveness of neural network modeling for accurately predicting fuel consumption in compression ignition engines across diesel and polymer-based fuels. By leveraging empirical data and sophisticated modeling techniques, this study contributes to advancing predictive capabilities in the field, offering valuable insights for optimizing engine performance and fuel efficiency.

Springback prediction using point series and deep learning

One of the main challenges that prevent wide adoption of Single Point Incremental Forming (SPIF) is the geometric accuracy of the process resulting primarily from the effect of springback. There are various expedients that can be adopted to address this, but one of the most common is tool path correction. The challenge is then how best to predict springback so as to implement tool path correction. It is established that springback, to a large extent, is related to the geometry of the part to be manufactured. The proposed mechanism uses a novel point series representation to capture local geometries that then form a global bank of geometries for general use. Each point series can then be associated with a predicted springback value generated using deep or machine learning. Experiments are reported using a Long Short Term Memory (LSTM) model coupled with a Multilayer Perception Network (MLP), and a Support Vector Machine (SVM) regression model. A best R2, “Coefficient of Determination”, of 0.9181 was obtained indicating that the proposed approach provided a realistic solution to the current limitations of SPIF.

An artificial neural network based approach for predicting the proton beam spot dosimetric characteristics of a pencil beam scanning technique

Utilising Machine Learning (ML) models to predict dosimetric parameters in pencil beam scanning proton therapy presents a promising and practical approach. The study developed Artificial Neural Network (ANN) models to predict proton beam spot size and relative positional errors using 9000 proton spot data. The irradiation log files as input variables and corresponding scintillation detector measurements as the label values. The ANN models were developed to predict six variables: spot size in the x-axis, y-axis, major axis, minor axis, and relative positional errors in the x-axis and y-axis. All ANN models used a Multi-layer perception (MLP) network using one input layer, three hidden layers, and one output layer. Model performance was validated using various statistical tools. The log file recorded spot size and relative positional errors, which were compared with scintillator-measured data. The Root Mean Squared Error (RMSE) values for the x-spot and y-spot sizes were 0.356 mm and 0.362 mm, respectively. Additionally, the maximum variation for the x-spot relative positional error was 0.910 mm, while for the y-spot, it was 1.610 mm. The ANN models exhibit lower prediction errors. Specifically, the RMSE values for spot size prediction in the x, y, major, and minor axes are 0.053 mm, 0.049 mm, 0.053 mm, and 0.052 mm, respectively. Additionally, the relative spot positional error prediction model for the x and y axes yielded maximum errors of 0.160 mm and 0.170 mm, respectively. The normality of models was validated using the residual histogram and Q-Q plot. The data over fit, and bias were tested using K (k = 5) fold cross-validation, and the maximum RMSE value of the K fold cross-validation among all the six ML models was less than 0.150 mm (R-Square 0.960). All the models showed excellent prediction accuracy. Accurately predicting beam spot size and positional errors enhances efficiency in routine dosimetric checks.

Long-term monitoring chlorophyll-a concentration using HJ-1 A/B imagery and machine learning algorithms in typical lakes, a cold semi-arid region.

Chlorophyll a (Chl-a) in lakes serves as an effective marker for assessing algal biomass and the nutritional level of lakes, and its observation is feasible through remote sensing methods. HJ-1 (Huanjing-1) satellite, deployed in 2008, incorporates a CCD capable of a 30 m resolution and has a revisit interval of 2 days, rendering it a superb choice or supplemental sensor for monitoring trophic state of lakes. For effective long-term and regional-scale mapping, both the imagery and the evaluation of machine learning algorithms are essential. The several typical machine learning algorithms, i.e., Support Vector Regression (SVR), Gradient Boosting Decision Trees (GBDT), XGBoost (XGB), Random Forest (RF), K-Nearest Neighbor (KNN), Kernel Ridge Regression (KRR), and Multi-Layer Perception Network (MLP), were developed using our in-situ measured Chl-a. A cross-validation grid to identify the most effective hyperparameter combinations for each algorithm was used, as well as the selected optimal superparameter combinations. In Chl-a mapping of three typical lakes, the R2 of GBDT, XGB, RF, and KRR all reached 0.90, while XGB algorithm also exhibited stable performance with the smallest error (RMSE = 3.11 μg/L). Adjustments were made to align the Chl-a spatial-temporal patterns with past data, utilizing HJ1-A/B CCD images mapping through XGB algorithm, which demonstrates its stability. Our results highlight the considerable effectiveness and utility of HJ-1 A/B CCD imagery for evaluation and monitoring trophic state of lakes in a cold arid region, providing the application cases contribute to the ongoing efforts to monitor water qualities.

MM-NeRF: Large-Scale Scene Representation with Multi-Resolution Hash Grid and Multi-View Priors Features

Reconstructing large-scale scenes using Neural Radiance Fields (NeRFs) is a research hotspot in 3D computer vision. Existing MLP (multi-layer perception)-based methods often suffer from issues of underfitting and a lack of fine details in rendering large-scale scenes. Popular solutions are to divide the scene into small areas for separate modeling or to increase the layer scale of the MLP network. However, the subsequent problem is that the training cost increases. Moreover, reconstructing large scenes, unlike object-scale reconstruction, involves a geometrically considerable increase in the quantity of view data if the prior information of the scene is not effectively utilized. In this paper, we propose an innovative method named MM-NeRF, which integrates efficient hybrid features into the NeRF framework to enhance the reconstruction of large-scale scenes. We propose employing a dual-branch feature capture structure, comprising a multi-resolution 3D hash grid feature branch and a multi-view 2D prior feature branch. The 3D hash grid feature models geometric details, while the 2D prior feature supplements local texture information. Our experimental results show that such integration is sufficient to render realistic novel views with fine details, forming a more accurate geometric representation. Compared with representative methods in the field, our method significantly improves the PSNR (Peak Signal-to-Noise Ratio) by approximately 5%. This remarkable progress underscores the outstanding contribution of our method in the field of large-scene radiance field reconstruction.

Design of Industrial IoT Intrusion Security Detection System Based on LightGBM Feature Algorithm and Multi-layer Perception Network

Continuous improvement of machine learning technology has provided more support for intrusion security detection in the industrial Internet of Things (IoT). The intrusion security detection system based on LightGBM feature algorithm and multi-layer perception network fusion provides more options for improving security detection, further enhancing the effectiveness of industrial IoT intrusion security detection. By comparing the applications of different models in the security detection process, the model constructed based on the LightGBM feature algorithm can achieve higher accuracy and precision in industrial IoT intrusion security detection, as well as higher F1-score and AUC values; A more reasonable detection time also lays the foundation for improving the overall efficiency of industrial IoT intrusion security detection. Therefore, in the field of industrial IoT intrusion security detection, the detection model constructed in this article can provide more support for further improvement and improvement of IoT intrusion security detection performance.

Towards estimating the thermal properties of carbon allotropes and their derivatives: Hybridization between the artificial neural network method and the experimental design approach

Carbon allotropes and their derivatives have attracted the interest of many researchers owing to their ability to conduct heat. In this study, a multi-layer perception network (MLP) was developed to predict its thermal transport properties. Usually, trial and error are the appropriate solutions to find the number of hidden layers, the number of neurons, the activation, and the transfer function. To overcome this problem, we used the Surface Response Methodology (RSM), which identifies the influence of each parameter on the response and ends up finding a robust model of Artificial Neural Networks (ANN). The study demonstrated that hybridization between (ANN) and (RSM) methods is an efficient and powerful method for modeling the thermal conductivity of carbon allotropes; it allows to shorten the time to test different combinations. Notably, the study achieved exceptional results with a correlation coefficient surpassing 99.89%, a coefficient of determination exceeding 99.78%, and minimal RMSE (Root Mean Square Error).

Artificial Neural Network-Based Modeling of Performance Spark Ignition Engine Fuelled with Bioethanol and Gasoline

Machine learning technology can distinguish the relationship between engine characteristics and performances. Therefore, the goal of the present work is to predict the performance parameters of a single-cylinder 4-stroke gasoline engine at different ignition timings using a blended mixture of gasoline and bioethanol by an artificial neural network (ANN). Experimental data for training and testing in the proposed ANN was obtained at a dynamic speed and full load condition. An ANN model was developed based on standard Back-Propagation algorithm for the spark ignition engine. Multi-layer perception network (MLP) was used for non-linear mapping between the input and output parameters. An optimizer in the family of quasi-Newton methods (lbfgs) and the rectified linear unit function were used to assess the percentage error between the desired and the predicted values. The network input parameters are engine speed, fuel, and ignition timing. Furthermore, torque, power, specific fuel consumption (SFC), thermal efficiency (ηth), and energy consumption (EC) are taken as output parameters. The results show that ANN is the proper method for predicting SIE performance because it has accurate prediction results that are very similar to experimental results. Moreover, from the observation results, the ANN model can predict the engine performance quite well with correlation coefficient (R)=0.962139 and MSE=0.003967 for data testing.

Artificial Neural Network based prediction of Specific Fuel Consumption in various Compression Ratio CI Engine Fueled with Various Plastic Composition Pyrolysis oil and Diesel

This study looks at the usage of Artificial Neural Network Modeling for predicting Specific Fuel Consumption for compression ignition engines running on Diesel, Low-Density Polyethylene Pyrolysis Oil (LDPE PO), High-Density Polyethylene Pyrolysis Oil (HDPE PO), and Polypropylene Pyrolysis Oil (PP PO). Using preliminary practical findings, The Model of ANN was built to estimate SFC by adjusting the parameters of the engine. SFC is anticipated by adjusting the parameters and utilizing the orthogonal array. The experiment is then carried out using an orthogonal array. The outcomes of experimental work are used to create an ANN model. In the case of a non-linear mapping between input & output, the classic back-propagation method and a multi-layer perception network are utilized. The values of Mean Square Error (MSE), Root Mean Square Error (RMSE), and Regression Coefficient (R2) for LM10TP architecture are 1.5143×10–06, 0.0012 & 1 in training & the validation values 1.2185×10–06, 0.0011 and 0.9999 respectively. It was discovered that neural networks are useful tools for prediction since they performed well in specific fuel consumption prediction in both training and validation.

Deep Neural Network for Blind Visual Quality Assessment of 4K Content

The 4K content can deliver a more immersive visual experience to consumers due to the huge improvement in spatial resolution. However, the high spatial resolution brings a great challenge for video transmission and storage. Therefore, it is necessary to compress or downscale the 4K content before transmitting it to end-users. Existing blind image quality assessment (BIQA) methods are not suitable for 4K contents due to the high spatial resolution and specific distortions caused by upscaling methods. In this paper, we propose a deep learning-based BIQA model for 4K content, aiming to recognize true and pseudo 4K content and meanwhile evaluate their visual quality. Considering the characteristic that high spatial resolution can represent more abundant high-frequency information, we first propose a Grey-level Co-occurrence Matrix (GLCM) based texture complexity measure to select three representative image patches from a 4K image, which can reduce the computational complexity and is proven to be very effective for the overall quality prediction. Then, we extract various visual features from the intermediate layers of the convolutional neural network (CNN) and integrate them into the quality-aware feature representation. Finally, two multilayer perception (MLP) networks are utilized to map the quality-aware features into the class probability and the quality score of each patch respectively. The overall quality index is obtained through averaging the results of all patches. The proposed model is trained via the multi-task learning manner and the uncertainty principle is introduced to balance the losses of the classification and regression tasks. The experimental results show that the proposed model outperforms all compared BIQA metrics on four 4K content quality assessment databases.

Development and validation of machine learning models for postoperative venous thromboembolism prediction in colorectal cancer inpatients: a retrospective study.

Colorectal cancer (CRC) is a heterogeneous group of malignancies distinguished by distinct clinical features. The association of these features with venous thromboembolism (VTE) is yet to be clarified. Machine learning (ML) models are well suited to improve VTE prediction in CRC due to their ability to receive the characteristics of a large number of features and understand the dataset to obtain implicit correlations. Data were extracted from 4,914 patients with colorectal cancer between August 2019 and August 2022, and 1,191 patients who underwent surgery on the primary tumor site with curative intent were included. The variables analyzed included patient-level factors, cancer-level factors, and laboratory test results. Model training was conducted on 30% of the dataset using a ten-fold cross-validation method and model validation was performed using the total dataset. The primary outcome was VTE occurrence in postoperative 30 days. Six ML algorithms, including logistic regression (LR), random forest (RF), extreme gradient boosting (XGBoost), weighted support vector machine (SVM), a multilayer perception (MLP) network, and a long short-term memory (LSTM) network, were applied for model fitting. The model evaluation was based on six indicators, including receiver operating characteristic curve-area under the curve (ROC-AUC), sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV), and Brier score. Two previous VTE models (Caprini and Khorana) were used as the benchmarks. The incidence of postoperative VTE was 10.8%. The top ten significant predictors included lymph node metastasis, C-reactive protein, tumor grade, anemia, primary tumor location, sex, age, D-dimer level, thrombin time, and tumor stage. In our results, the XGBoost model showed the best performance, with a ROC-AUC of 0.990, a SEN of 96.9%, a SPE of 96.1% in training dataset and a ROC-AUC of 0.908, a SEN of 77.5%, a SPE of 93.7% in validation dataset. All ML models outperformed the previously developed models (Caprini and Khorana). This study developed postoperative VTE predictive models using six ML algorithms. The XGBoost VTE model might supply a complementary tool for clinical VTE prophylaxis decision-making and the proposed risk factors could shed some light on VTE risk stratification in CRC patients.

A Neural Network Based Approach to Inverse Kinematics Problem for General Six-Axis Robots.

Inverse kinematics problems (IKP) are ubiquitous in robotics for improved robot control in widespread applications. However, the high non-linearity, complexity, and equation coupling of a general six-axis robotic manipulator pose substantial challenges in solving the IKP precisely and efficiently. To address this issue, we propose a novel approach based on neural network (NN) with numerical error minimization in this paper. Within our framework, the complexity of IKP is first simplified by a strategy called joint space segmentation, with respective training data generated by forward kinematics. Afterwards, a set of multilayer perception networks (MLP) are established to learn from the foregoing data in order to fit the goal function piecewise. To reduce the computational cost of the inference process, a set of classification models is trained to determine the appropriate forgoing MLPs for predictions given a specific input. After the initial solution is sought, being improved with a prediction error minimized, the refined solution is finally achieved. The proposed methodology is validated via simulations on Xarm6-a general 6 degrees of freedom manipulator. Results further verify the feasibility of NN for IKP in general cases, even with a high-precision requirement. The proposed algorithm has showcased enhanced efficiency and accuracy compared to NN-based approaches reported in the literature.

Cross-Domain Explicit–Implicit-Mixed Collaborative Filtering Neural Network

Collaborative filtering (CF) is a classical model for recommender systems. Though neural network-based CF models have been shown to perform well in some cases, they still suffer from the sparsity and cold-start issues. To address these issues, we propose a novel neural network-based CF model, termed the cross-domain explicit–implicit-mixed CF neural network (CEICFNet). The proposed model utilizes deep neural networks to learn latent factors not only from the explicit ratings and the implicit interactions but also in a cross-domain manner. In particular, domain-shared multilayer perception (MLP) networks are designed to learn the user rating latent factors and the user interaction latent factors from the explicit ratings and the implicit interactions, respectively, which act as bridges for transferring knowledge across domains. On the other hand, domain-specific MLP networks are designed to learn the item rating latent factors and the item interaction latent factors from the explicit ratings and the implicit interactions, respectively. Then, in each domain, based on the user rating latent factors and the item rating latent factors, the rating predictive representation for each user–item pair can be learned by an MLP. Similarly, the interaction predictive representation for each user–item pair can be learned. For integrating the explicit ratings and the implicit interactions, in each domain, a fully connected layer is used to automatically assign different weights to the rating predictive representations and the interaction predictive representations, based on which the final interaction probability can be generated. Extensive experiments are conducted on five datasets and the results have confirmed the effectiveness of our model.

Multi-modal emotion recognition using EEG and speech signals

Automatic Emotion Recognition (AER) is critical for naturalistic Human–Machine Interactions (HMI). Emotions can be detected through both external behaviors, e.g., tone of voice and internal physiological signals, e.g., electroencephalogram (EEG). In this paper, we first constructed a multi-modal emotion database, named Multi-modal Emotion Database with four modalities (MED4). MED4 consists of synchronously recorded signals of participants’ EEG, photoplethysmography, speech and facial images when they were influenced by video stimuli designed to induce happy, sad, angry and neutral emotions. The experiment was performed with 32 participants in two environment conditions, a research lab with natural noises and an anechoic chamber. Four baseline algorithms were developed to verify the database and the performances of AER methods, Identification-vector + Probabilistic Linear Discriminant Analysis (I-vector + PLDA), Temporal Convolutional Network (TCN), Extreme Learning Machine (ELM) and Multi-Layer Perception Network (MLP). Furthermore, two fusion strategies on feature-level and decision-level respectively were designed to utilize both external and internal information of human status. The results showed that EEG signals generate higher accuracy in emotion recognition than that of speech signals (achieving 88.92% in anechoic room and 89.70% in natural noisy room vs 64.67% and 58.92% respectively). Fusion strategies that combine speech and EEG signals can improve overall accuracy of emotion recognition by 25.92% when compared to speech and 1.67% when compared to EEG in anechoic room and 31.74% and 0.96% in natural noisy room. Fusion methods also enhance the robustness of AER in the noisy environment. The MED4 database will be made publicly available, in order to encourage researchers all over the world to develop and validate various advanced methods for AER.

Real-time dense-view imaging for three-dimensional light-field display based on image color calibration and self-supervised view synthesis.

Three-Dimensional (3D) light-field display has achieved promising improvement in recent years. However, since the dense-view images cannot be collected fast in real-world 3D scenes, the real-time 3D light-field display is still challenging to achieve in real scenes, especially at the high-resolution 3D display. Here, a real-time 3D light-field display method with dense-view is proposed based on image color correction and self-supervised optical flow estimation, and a high-quality and high frame rate of 3D light-field display can be realized simultaneously. A sparse camera array is firstly used to capture sparse-view images in the proposed method. To eliminate the color deviation of the sparse views, the imaging process of the camera is analyzed, and a practical multi-layer perception (MLP) network is proposed to perform color calibration. Given sparse views with consistent color, the optical flow can be estimated by a lightweight convolutional neural network (CNN) at high speed, which uses the input image pairs to learn the optical flow in a self-supervised manner. With inverse warp operation, dense-view images can be synthesized in the end. Quantitative and qualitative experiments are performed to evaluate the feasibility of the proposed method. Experimental results show that over 60 dense-view images at a resolution of 1024 × 512 can be generated with 11 input views at a frame rate over 20 fps, which is 4× faster than previous optical flow estimation methods PWC-Net and LiteFlowNet3. Finally, large viewing angles and high-quality 3D light-field display at 3840 × 2160 resolution can be achieved in real-time.

Intrusion Detection System for Industrial Internet of Things Based on Deep Reinforcement Learning

The Industrial Internet of Things has grown significantly in recent years. While implementing industrial digitalization, automation, and intelligence introduced a slew of cyber risks, the complex and varied industrial Internet of Things environment provided a new attack surface for network attackers. As a result, conventional intrusion detection technology cannot satisfy the network threat discovery requirements in today’s Industrial Internet of Things environment. In this research, the authors have used reinforcement learning rather than supervised and unsupervised learning, because it could very well improve the decision-making ability of the learning process by integrating abstract thinking of complete understanding, using deep knowledge to perform simple and nonlinear transformations of large-scale original input data into higher-level abstract expressions, and using learning algorithm or learning based on feedback signals, in the lack of guiding knowledge, which is based on the trial-and-error learning model, from the interaction with the environment to find the best good solution. In this respect, this article presents a near-end strategy optimization method for the Industrial Internet of Things intrusion detection system based on a deep reinforcement learning algorithm. This method combines deep learning’s observation capability with reinforcement learning’s decision-making capability to enable efficient detection of different kinds of cyberassaults on the Industrial Internet of Things. In this manuscript, the DRL-IDS intrusion detection system is built on a feature selection method based on LightGBM, which efficiently selects the most attractive feature set from industrial Internet of Things data; when paired with deep learning algorithms, it effectively detects intrusions. To begin, the application is based on GBM’s feature selection algorithm, which extracts the most compelling feature set from Industrial Internet of Things data; then, in conjunction with the deep learning algorithm, the hidden layer of the multilayer perception network is used as the shared network structure for the value network and strategic network in the PPO2 algorithm; and finally, the intrusion detection model is constructed using the PPO2 algorithm and ReLU (R). Numerous tests conducted on a publicly available data set of the Industrial Internet of Things demonstrate that the suggested intrusion detection system detects 99 percent of different kinds of network assaults on the Industrial Internet of Things. Additionally, the accuracy rate is 0.9%. The accuracy, precision, recall rate, F1 score, and other performance indicators are superior to those of the existing intrusion detection system, which is based on deep learning models such as LSTM, CNN, and RNN, as well as deep reinforcement learning models such as DDQN and DQN.

Objective Recognition of Tinnitus Location Using Electroencephalography Connectivity Features.

Purpose: Tinnitus is a common but obscure auditory disease to be studied. This study will determine whether the connectivity features in electroencephalography (EEG) signals can be used as the biomarkers for an efficient and fast diagnosis method for chronic tinnitus.Methods: In this study, the resting-state EEG signals of tinnitus patients with different tinnitus locations were recorded. Four connectivity features [including the Phase-locking value (PLV), Phase lag index (PLI), Pearson correlation coefficient (PCC), and Transfer entropy (TE)] and two time-frequency domain features in the EEG signals were extracted, and four machine learning algorithms, included two support vector machine models (SVM), a multi-layer perception network (MLP) and a convolutional neural network (CNN), were used based on the selected features to classify different possible tinnitus sources.Results: Classification accuracy was highest when the SVM algorithm or the MLP algorithm was applied to the PCC feature sets, achieving final average classification accuracies of 99.42 or 99.1%, respectively. And based on the PLV feature, the classification result was also particularly good. And MLP ran the fastest, with an average computing time of only 4.2 s, which was more suitable than other methods when a real-time diagnosis was required.Conclusion: Connectivity features of the resting-state EEG signals could characterize the differentiation of tinnitus location. The connectivity features (PCC and PLV) were more suitable as the biomarkers for the objective diagnosing of tinnitus. And the results were helpful for clinicians in the initial diagnosis of tinnitus.

DNN‐aided read‐voltage threshold optimization for MLC flash memory with finite block length

The error-correcting performance of multi-level-cell (MLC) NAND flash memory is closely related to the block length of error-correcting codes (ECCs) and log-likelihood-ratios of the read-voltage thresholds. Driven by this issue, this paper optimizes the read-voltage thresholds for MLC flash memory to improve the decoding performance of ECCs with finite block length. First, through the analysis of channel coding rate and decoding error probability under finite block length, the optimization problem of read-voltage thresholds to minimize the maximum decoding error probability is formulated. Second, a cross-iterative search algorithm to optimize read-voltage thresholds under the perfect knowledge of flash memory channel is developed. However, it is challenging to analytically characterize the voltage distribution under the effect of data retention noise. To address this problem, a deep neural network (DNN)-aided optimization strategy to optimize the read-voltage thresholds is developed, where a multi-layer perception network is employed to learn the relationship between voltage distribution and read-voltage thresholds. Simulation results show that, compared with the existing schemes, the proposed DNN-aided read-voltage threshold optimization strategy with a well-designed Low Density Parity Check (LDPC) code can not only improve the program-and-erase endurance but also reduce the read latency.

A CNN Architecture for Learning Device Activity From MMV

Device activity detection has been extensively investigated for grant-free massive machine-type communications. Instead of using deep Multi-Layer Perception (MLP) networks, this letter proposes a novel convolutional neural network (CNN) architecture for learning device activity from multiple-measurement vectors (MMV). With the use of 1×1 convolutional layers, the proposed CNN could exploit the full potential of shared sparsity among multiple measurements. Extensive simulations show that the proposed CNN outperforms the existing deep MLP network in both performance and computational complexity, especially when the number of measurements increases.