Deep Learning Model for the Automatic Classification of White Blood Cells

Blood cell count is highly useful in identifying the occurrence of a particular disease or ailment. To successfully measure the blood cell count, sophisticated equipment that makes use of invasive methods to acquire the blood cell slides or images is utilized. These blood cell images are subjected to various data analyzing techniques that count and classify the different types of blood cells. Nowadays, deep learning-based methods are in practice to analyze the data. These methods are less time-consuming and require less sophisticated equipment. This paper implements a deep learning (D.L) model that uses the DenseNet121 model to classify the different types of white blood cells (WBC). The DenseNet121 model is optimized with the preprocessing techniques of normalization and data augmentation. This model yielded an accuracy of 98.84%, a precision of 99.33%, a sensitivity of 98.85%, and a specificity of 99.61%. The proposed model is simulated with four batch sizes (BS) along with the Adam optimizer and 10 epochs. It is concluded from the results that the DenseNet121 model has outperformed with batch size 8 as compared to other batch sizes. The dataset has been taken from the Kaggle having 12,444 images with the images of 3120 eosinophils, 3103 lymphocytes, 3098 monocytes, and 3123 neutrophils. With such results, these models could be utilized for developing clinically useful solutions that are able to detect WBC in blood cell images.

Piston-Based Material Extrusion of Ti-6Al-4V Feedstock for Complementary Use in Metal Injection Molding

Piston-based material extrusion enables cost savings for metal injection molding users when it is utilized as a complementary shaping process for green parts in small batch sizes. This, however, requires the use of series feedstock and the production of sufficiently dense green parts in order to ensure metal injection molding-like material properties. In this paper, a methodological approach is presented to identify material-specific process parameters for an industrially used Ti-6Al-4V metal injection molding feedstock based on the extrusion force. It was found that for an optimum extrusion temperature of 95 °C and printing speed of 8 mm/s an extrusion force of 1300 N ensures high-density green parts without under-extrusion. The resulting sintered part properties exhibit values comparable to metal injection molding in terms of part density (max. 99.1%) and tensile properties (max. yield strength: 933 MPa, max. ultimate tensile strength: 1000 MPa, max. elongation at break: 18.5%) depending on the selected build orientation. Thus, a complementary use could be demonstrated in principle for the Ti-6Al-4V feedstock.

A model for measuring distruption risk in the prefabrication supply chain

Use of prefabrication in construction projects is increasing due to the benefits in cost, time, quality, and safety. However, utilizing prefabrication introduces uncertainties inherent with the supply chain of the process. These uncertainties, if not managed, can disrupt the prefabrication process and result in schedule delays and cost overruns. This study proposes a model to measure disruption risks in the prefabrication process. The model was used in measuring the disruption risks of prefabrication of headwalls in patients’ rooms for a healthcare project as a pilot study. The risk model could successfully identify the disruption risks originating anywhere in the supply chain based on input information such as required material quantity, batch sizes of material deliveries, production rates, and batch sizes of transporting the headwall units. Using the model, the project team identified two uncertainties that could lead to possible disruptions: the start of the prefabrication processes and the required production rate to meet the on-site schedule. This is a first step to developing a risk exposure model that can prove valuable to the risk managers to analyse and manage the impact of disruptions. This will help the risk managers in making informed decisions about where to focus their limited resources.

An Analysis of the Use of Feed-Forward Sub-Modules for Transformer-Based Image Captioning Tasks

In the quest to make deep learning systems more capable, a number of more complex, more computationally expensive and memory intensive algorithms have been proposed. This switchover glosses over the capabilities of many of the simpler systems or modules within them to adequately address current and future problems. This has led to some of the deep learning research being inaccessible to researchers who don’t possess top-of-the-line hardware. The use of simple feed forward networks has not been explicitly explored in the current transformer-based vision-language field. In this paper, we use a series of feed-forward layers to encode image features, and caption embeddings, alleviating some of the effects of the computational complexities that accompany the use of the self-attention mechanism and limit its application in long sequence task scenarios. We demonstrate that a decoder does not require masking for conditional short sequence generation where the task is not only dependent on the previously generated sequence, but another input such as image features. We perform an empirical and qualitative analysis of the use of linear transforms in place of self-attention layers in vision-language models, and obtain competitive results on the MSCOCO dataset. Our best feed-forward model obtains average scores of over 90% of the current state-of-the-art pre-trained Oscar model in the conventional image captioning metrics. We also demonstrate that the proposed models take less time training and use less memory at larger batch sizes and longer sequence lengths.

Traffic Flow Online Prediction Based on a Generative Adversarial Network with Multi-Source Data

Traffic prediction is essential for advanced traffic planning, design, management, and network sustainability. Current prediction methods are mostly offline, which fail to capture the real-time variation of traffic flows. This paper establishes a sustainable online generative adversarial network (GAN) by combining bidirectional long short-term memory (BiLSTM) and a convolutional neural network (CNN) as the generative model and discriminative model, respectively, to keep learning with continuous feedback. BiLSTM constantly generates temporal candidate flows based on valuable memory units, and CNN screens out the best spatial prediction by returning the feedback gradient to BiLSTM. Multi-dimensional indicators are selected to map the multi-view fusion local trend for accurate prediction. To balance computing efficiency and accuracy, different batch sizes are pre-tested and allocated to different lanes. The models are trained with rectified adaptive moment estimation (RAdam) by dividing the dataset into the training and testing sets with a rolling time-domain scheme. In comparison with the autoregressive integrated moving average (ARIMA), BiLSTM, generating adversarial network for traffic flow (GAN-TF), and generating adversarial network for non-signal traffic (GAN-NST), the proposed improved generating adversarial network for traffic flow (IGAN-TF) successfully generates more accurate and stable flows and performs better.

Optimization of Multi Two-Level Batch Testing

As the COVID-19 pandemic continues to spread, rapid testing could help curb asymptomatic transmission. Thus, it is paramount that the most efficient testing method be identified and implemented, as to reduce the strain on the medical community. This project introduces a novel batch testing method called multi two-level batch testing, which was hypothesized to increase the efficiency of batch testing in terms of minimizing the number of tests performed for a given population. While Dorfman’s two-level and Li’s multi-level batch testing methods already exist, this method offers a novel strategy distinct from existing methods. A Java simulation was created to iteratively compute the number of tests required for each testing method at various percentages of population infection rate, batch sizes, and other parameters specific to each method. Based on this simulation, it can be shown that the multi two-level procedure is more efficient than both the two-level and the multi-level procedures at an infection rate of 0.01, which is the anticipated rate at the University of Florida during the Spring 2021 semester. Additionally, at infection rates between 0.05 and 0.30, the multi two-level batch testing method slightly outperforms multi-level. When the infection rate exceeds 0.30, all methods are unviable and begin to require more tests than necessary to test each person in the population individually. If laboratories implement multi two-level batch testing, they may reduce costs and labor. Additionally, the novel batch testing procedure can be applied to other diseases and future pandemics.

An Improved Tool for Detection of XSS Attacks by Combining CNN with LSTM

Cross-Site Scripting (XSS) is still a significant threat to web applications. By combining Convolutional Neural Networks (CNN) with Long ShortTerm Memory (LSTM) techniques, researchers have developed a deep learning system called 3C-LSTM that achieves upwards of 99.4% accuracy when predicting whether a new URL corresponds to a benign locator or an XSS attack. This paper improves on 3C-LSTM by proposing different network architectures and validation strategies and identifying the optimal structure for a more efficient, yet similarly accurate, version of 3C-LSTM. The authors identify larger batch sizes, smaller inputs, and cross-validation removal as modifications to achieve a speedup of around 3.9 times in the training step.

An Efficient CNN-Based Deep Learning Model to Detect Malware Attacks (CNN-DMA) in 5G-IoT Healthcare Applications

The role of 5G-IoT has become indispensable in smart applications and it plays a crucial part in e-health applications. E-health applications require intelligent schemes and architectures to overcome the security threats against the sensitive data of patients. The information in e-healthcare applications is stored in the cloud which is vulnerable to security attacks. However, with deep learning techniques, these attacks can be detected, which needs hybrid models. In this article, a new deep learning model (CNN-DMA) is proposed to detect malware attacks based on a classifier—Convolution Neural Network (CNN). The model uses three layers, i.e., Dense, Dropout, and Flatten. Batch sizes of 64, 20 epoch, and 25 classes are used to train the network. An input image of 32 × 32 × 1 is used for the initial convolutional layer. Results are retrieved on the Malimg dataset where 25 families of malware are fed as input and our model has detected is Alueron.gen!J malware. The proposed model CNN-DMA is 99% accurate and it is validated with state-of-the-art techniques.

Using machine learning to model the training scalability of convolutional neural networks on clusters of GPUs

In this work, we build a general piece-wise model to analyze data-parallel (DP) training costs of convolutional neural networks (CNNs) on clusters of GPUs. This general model is based on i) multi-layer perceptrons (MLPs) in charge of modeling the NVIDIA cuDNN/cuBLAS library kernels involved in the training of some of the state-of-the-art CNNs; and ii) an analytical model in charge of modeling the NVIDIA NCCL Allreduce collective primitive using the Ring algorithm. The CNN training scalability study performed using this model in combination with the Roofline technique on varying batch sizes, node (floating-point) arithmetic performance, node memory bandwidth, network link bandwidth, and cluster dimension unveil some crucial bottlenecks at both GPU and cluster level. To provide evidence of this analysis, we validate the accuracy of the proposed model against a Python library for distributed deep learning training.

The origin of end flare in roll formed profiles

Roll forming is a continuous manufacturing process designed for large batch sizes. In order to economically produce roll formed parts with smaller batch sizes, the process setup times have to be reduced. During the setup, profile defects and especially the deformation caused by the release of the process-inherent residual stresses, also known as end flare, have to be counteracted. However, the knowledge regarding the creation of residual stresses is limited and the ability to reduce end flare usually depends on the experience of the process designer and the machine operator, which makes the setup time-consuming and cost-intensive. Therefore, in this paper the creation of end flare during the roll forming process is investigated in depth. As a result of this study explanation models for U-, C- and Hat-profiles, which link the creation of residual stresses to the local deformation during the forming process, are developed. Knowing how changes in the forming curve affect the creation of end flare allows to use a knowledge-based approach during the design and setup process, thereby reducing time and costs.