Most enterprise customers now choose to divide a large monolithic service into large numbers of loosely-coupled, specialized microservices, which can be developed and deployed separately. Docker, as ...a light-weight virtualization technology, has been widely adopted to support diverse microservices. At the moment, Kubernetes is a portable, extensible, and open-source orchestration platform for managing these containerized microservice applications. To adapt to frequently changing user requests, it offers an automated scaling method, Horizontal Pod Autoscaler (HPA), that can scale itself based on the system’s current workload. The native reactive auto-scaling method, however, is unable to foresee the system workload scenario in the future to complete proactive scaling, leading to QoS (quality of service) violations, long tail latency, and insufficient server resource usage. In this paper, we suggest a new proactive scaling scheme based on deep learning approaches to make up for HPA’s inadequacies as the default autoscaler in Kubernetes. After meticulous experimental evaluation and comparative analysis, we use the Gated Recurrent Unit (GRU) model with higher prediction accuracy and efficiency as the prediction model, supplemented by a stability window mechanism to improve the accuracy and stability of the prediction model. Finally, with the third-party custom autoscaling framework, Custom Pod Autoscaler (CPA), we packaged our custom autoscaling algorithm into a framework and deployed the framework into the real Kubernetes cluster. Comprehensive experiment results prove the feasibility of our autoscaling scheme, which significantly outperforms the existing Horizontal Pod Autoscaler (HPA) approach.
•Gated Recurrent Units with attention mechanism learn time-series features efficiently.•Multi-channel model exploits correlations among the various cryptocurrency prices.•Convolutional neural ...networks extract local features effectively.
After the invention of Bitcoin as well as other blockchain-based peer-to-peer payment systems, the cryptocurrency market has rapidly gained popularity. Consequently, the volatility of the various cryptocurrency prices attracts substantial attention from both investors and researchers. It is a challenging task to forecast the prices of cryptocurrencies due to the non-stationary prices and the stochastic effects in the market. Current cryptocurrency price forecasting models mainly focus on analyzing exogenous factors, such as macro-financial indicators, blockchain information, and social media data – with the aim of improving the prediction accuracy. However, the intrinsic systemic noise, caused by market and political conditions, is complex to interpret. Inspired by the strong correlations among cryptocurrencies and the powerful modelling capability displayed by deep learning techniques, we propose a Weighted & Attentive Memory Channels model to predict the daily close price and the fluctuation of cryptocurrencies. In particular, our proposed model consists of three modules: an Attentive Memory module combines a Gated Recurrent Unit with a self-attention component to establish attentive memory for each input sequence; a Channel-wise Weighting module receives the price of several heavyweight cryptocurrencies and learns their interdependencies by recalibrating the weights for each sequence; and a Convolution & Pooling module extracts local temporal features, thereby improving the generalization ability of the overall model. In order to validate the proposed model, we conduct a battery of experiments. The results show that our proposed scheme achieves state-of-the-art performance and outperforms the baseline models in prediction error, accuracy, and profitability.
Transfer learning with spinally shared layers Kabir, H.M. Dipu; Mondal, Subrota Kumar; Alam, Syed Bahauddin ...
Applied soft computing,
September 2024, 2024-09-00, Volume:
163
Journal Article
Peer reviewed
Open access
Transfer-learned models have achieved promising performance in numerous fields. However, high-performing transfer-learned models contain a large number of parameters. In this paper, we propose a ...transfer learning approach with parameter reduction and potential high performance. Although the high performance depends on the nature of the dataset, we ensure the parameter reduction. In the proposed SpinalNet shared parameters, all intermediate-split-incoming parameters except the first-intermediate-split contain a shared value. Therefore, the SpinalNet shared parameters network contains three parameter groups: (1) first input-split to intermediate-split parameters, (2) shared intermediate-split-incoming parameters, and (3) intermediate-split-to-output-split parameters. The total number of parameters becomes lower than the SpinalNet and traditional fully connected layers due to parameter sharing. Besides the overall accuracy, this paper compares the precision, recall, and F1-score of each class as performance criteria. As a result, both parameter reduction and potential performance improvement become possible for the ResNet-type models, VGG-type traditional models, and Vision Transformers. We applied the proposed model to MNIST, STL-10, and COVID-19 datasets to validate our claims. We also provided a posterior plot of the sample from different models for medical practitioners to understand the uncertainty. Example model training scripts of the proposed model are also shared to GitHub.
•Proposed a transfer learning model with an optimal size-performance trade-off.•Reduced the number of parameters in the fully connected or the head layer.•Achieved both parameter reduction and improved performance for several combinations.•Provided an example NN training script at: www.github.com/dipuk0506/SpinalNet/.•Proposed a posterior plot for a closer understanding of the level of uncertainty.
Deep question generation (DQG) refers to generating a complex question from different sentences in context. Existing methods mainly focus on enhancing information extraction based on the ...encoder–decoder neural networks though they cannot perform well in DQG tasks. To address this issue, we consider combining reinforcement learning with semantic-rich information to generate deep questions in this paper. In particular, we propose a Semantic-Rich Reinforcement Learning Deep Question Generation (SRL-DQG) model, which better utilizes the semantic graphs of document representations based on the Gated Graph Neural Network (GGNN). In order to generate high-quality questions, we also optimize specific objectives via reinforcement learning with consideration of four evaluation factors including naturality, relevance, answerability, and difficulty. Empirical evaluations demonstrate that our SRL-DQG effectively improves the quality of generated questions and achieves superior performance than existing methods in terms of multiple performance metrics. Specifically, we show that several BLEU-n scores were improved by 3.5% to 10% after running SRL-DQG on 6072 samples of HotPotQA.
•Integrating reinforcement learning and semantic information methods for deep question generation.•Using multiple evaluation metrics: naturality, relevance, answerability, and difficulty in reward system.•Extensive experiments demonstrate the proposed method achieving state-of-the-art performance.
Machine translation and its evaluation: a study Mondal, Subrota Kumar; Zhang, Haoxi; Kabir, H. M. Dipu ...
The Artificial intelligence review,
09/2023, Volume:
56, Issue:
9
Journal Article
Peer reviewed
Machine translation (namely MT) has been one of the most popular fields in computational linguistics and Artificial Intelligence (AI). As one of the most promising approaches, MT can potentially ...break the language barrier of people from all over the world. Despite a number of studies in MT, there are few studies in summarizing and comparing MT methods. To this end, in this paper, we principally focus on presenting the two mainstream MT schemes: statistical machine translation (SMT) and neural machine translation (NMT), including their basic rationales and developments. Meanwhile, the detailed translation models are also presented, such as the word-based model, syntax-based model, and phrase-based model in statistical machine translation. Similarly, approaches in NMT, such as the recurrent neural network-based, attention mechanism-based, and transformer-based models are presented. Last but not least, the evaluation approaches also play an important role in helping developers to improve their methods better in MT. The prevailing machine translation evaluation methodologies are also presented in this article.
As an emulous alternative to deep neural networks, Deep Forest emerges with features like low complexity, fewer hyper-parameters, and good robustness, which are predominantly desired in distributed ...computing applications and ecosystems. Recently, an efficient distributed Deep Forest system, named ForestLayer, was proposed, designing a fine-grained sub-Forest-based task-parallel algorithm to improve the parallel computing efficiency of Deep Forest. However, the sub-Forest splitting of ForestLayer is static and one-off without adaptability to the computing environment, nevertheless, the size of splitting granularity has a significant impact on the system performance. To further improve the computing efficiency and scalability of the distributed Deep Forest, in this paper, we propose a novel distributed Deep Forest algorithm, named BLB-gcForest (Bag of Little Bootstraps-gcForest), which augments the gcForest (multi-Grained Cascade Forest) approach for constructing Deep Forest. BLB-gcForest carries out parallel computation for each tree in sub-Forests at a finer parallel granularity and integrates with the Bag of Little Bootstraps (BLB) mechanism to reduce massive transmitted feature instances for Cascade Forest Layers, utterly improving both computation efficiency and communication efficiency. Moreover, to solve the problem of the forest splitting granularity, we further design an adaptive sub-Forest splitting algorithm to ensure the maximum resource utilization for parallel computation of each sub-Forest. Experimental results on four well-known large-scale datasets, namely YEAST, LETTER, MNIST, CIFAR10, show that the training efficiency of BLB-gcForest achieves up to 20.3x and 1.64x speedups compared with the state-of-the-art gcForest and ForestLayer, respectively while guaranteeing higher accuracy and better robustness
Today’s industry has gradually realized the importance of lifting efficiency and saving costs during the life-cycle of an application. In particular, we see that most of the cloud-based applications ...and services often consist of hundreds of micro-services; however, the traditional monolithic pattern is no longer suitable for today’s development life-cycle. This is due to the difficulties of maintenance, scale, load balance, and many other factors associated with it. Consequently, people switch their focus on containerization—a lightweight virtualization technology. The saving grace is that it can use machine resources more efficiently than the virtual machine (VM). In VM, a guest OS is required to simulate on the host machine, whereas containerization enables applications to share a common OS. Furthermore, containerization facilitates users to create, delete, or deploy containers effortlessly. In order to manipulate and manage the multiple containers, the leading Cloud providers introduced the container orchestration platforms, such as Kubernetes, Docker Swarm, Nomad, and many others. In this paper, a rigorous study on Kubernetes from an administrator’s perspective is conducted. In a later stage, serverless computing paradigm was redefined and integrated with Kubernetes to accelerate the development of software applications. Theoretical knowledge and experimental evaluation show that this novel approach can be accommodated by the developers to design software architecture and development more efficiently and effectively by minimizing the cost charged by public cloud providers (such as AWS, GCP, Azure). However, serverless functions are attached with several issues, such as security threats, cold start problem, inadequacy of function debugging, and many other. Consequently, the challenge is to find ways to address these issues. However, there are difficulties and hardships in addressing all the issues altogether. Respectively, in this paper, we simply narrow down our analysis toward the security aspects of serverless. In particular, we quantitatively measure the success probability of attack in serverless (using Attack Tree and Attack–Defense Tree) with the possible attack scenarios and the related countermeasures. Thereafter, we show how the quantification can reflect toward the end-to-end security enhancement. In fine, this study concludes with research challenges such as the burdensome and error-prone steps of setting the platform, and investigating the existing security vulnerabilities of serverless computing, and possible future directions.
With the rapidly increasing demands of e-government systems in smart cities, a myriad of challenges and issues are required to be addressed. Among them, security is one of the prime concerns. To this ...end, we analyze different e-government systems and find that an e-government system built with container-based technology is endowed with many features. In addition, overhauling the architecture of container-technology-driven e-government systems, we observe that securing an e-government system demands quantifying security issues (vulnerabilities, threats, attacks, and risks) and the related countermeasures. Notably, we find that the Attack Tree and Attack–Defense Tree methods are state-of-the-art approaches in these aspects. Consequently, in this paper, we work on quantifying the security attributes, measures, and metrics of an e-government system using Attack Trees and Attack–Defense Trees—in this context, we build a working prototype of an e-government system aligned with the United Kingdom (UK) government portal, which is in line with our research scope. In particular, we propose a novel measure to quantify the probability of attack success using a risk matrix and normal distribution. The probabilistic analysis distinguishes the attack and defense levels more intuitively in e-government systems. Moreover, it infers the importance of enhancing security in e-government systems. In particular, the analysis shows that an e-government system is fairly unsafe with a 99% probability of being subject to attacks, and even with a defense mechanism, the probability of attack lies around 97%, which directs us to pay close attention to e-government security. In sum, our implications can serve as a benchmark for evaluation for governments to determine the next steps in consolidating e-government system security.
The user does not have any idea about the credibility of outcomes from deep neural networks (DNN) when uncertainty quantification (UQ) is not employed. However, current Deep UQ classification models ...capture mostly epistemic uncertainty. Therefore, this paper aims to propose an aleatory-aware Deep UQ method for classification problems. First, we train DNNs through transfer learning and collect numeric output posteriors for all training samples instead of logical outputs. Then we determine the probability of happening a certain class from K-nearest output posteriors of the same DNN in training samples. We name this probability as opacity score, as the paper focuses on the detection of opacity on X-ray images. This score reflects the level of aleatory on the sample. When the NN is certain on the classification of the sample, the probability of happening a class becomes much higher than the probabilities of others. Probabilities for different classes become close to each other for a highly uncertain classification outcome. To capture the epistemic uncertainty, we train multiple DNNs with different random initializations, model selection, and augmentations to observe the effect of these training parameters on prediction and uncertainty. To reduce execution time, we first obtain features from the pre-trained NN. Then we apply features to the ensemble of fully connected layers to get the distribution of opacity score during the test. We also train several ResNet and DenseNet DNNs to observe the effect of model selection on prediction and uncertainty. The paper also demonstrates a patient referral framework based on the proposed uncertainty quantification. The scripts of the proposed method are available at the following link: https://github.com/dipuk0506/Aleatory-aware-UQ.
•Considering asymmetric and heteroscedastic aleatoric uncertainty in image classification.•Consideration of the model-selection uncertainty.•Formulation of relative probability from K-nearest neighbors and relative to realistic probability conversion.•Study on the effect of augmentations on overall uncertainty.•An uncertainty-aware patient referral flow chart.
