Implementation Of Learning Algorithms Research Articles

Targeted nano-energetic material exploration through active learning algorithm implementation

This paper presents an active learning-based method designed to guide experiments towards user-defined specific regions, termed ”regions of interest,” within vast and multi-dimensional thermite design spaces. Thermites composed of metallic reactant coupled to an inorganic oxidizer are non-explosive energetic materials which stay inert and stable until subjected to a sufficiently strong thermal stimulus, after which they undergo fast burning with release of high amount of chemical energy (up to 16 kJ.cm−3). They represent an interesting class of nano-engineered energetic material because of their high adiabatic flame temperature (> 2600 °C) and customizable combustion properties. We introduced a new acquisition function combining linearly two factors with the usual standard deviation of a Gaussian Process Regression algorithm. A first factor guides the sampling towards the defined zone of interest, and a second, is an incentive function that encourages the exploration of under-sampled regions of the design space. We found that our algorithm effectively provides up to several tens of nanothermites that achieve specific desired properties well-distributed within the design space, after only 200 samplings, whereas, Latin Hypercube Sampling procedure samples less than 10 points of interest. Our framework is tailored for typical discrete search spaces involving multiple measured physical properties and when only a small dataset is available, as it is the case in thermite materials. But more generally, this research represents a significant advancement for large-scale problems in energetic materials science and engineering, where predicting the effect of feature modification is desired but limited simulations or experiments can be afforded due to their high cost.

Improving Early Fault Detection in Machine Learning Systems Using Data Diversity-Driven Metamorphic Relation Prioritization

Metamorphic testing is a valuable approach to verifying machine learning programs where traditional oracles are unavailable or difficult to apply. This paper proposes a technique to prioritize metamorphic relations (MRs) in metamorphic testing for machine learning and deep learning systems, aiming to enhance early fault detection. We introduce five metrics based on diversity in source and follow-up test cases to prioritize MRs. The effectiveness of our proposed prioritization methods is evaluated on three machine learning and one deep learning algorithm implementation. We compare our approach against random-based, fault-based, and neuron activation coverage-based MR ordering. The results show that our data diversity-based prioritization performs comparably to fault-based prioritization, reducing fault detection time by up to 62% compared to random MR execution. Our proposed metrics outperformed neuron activation coverage-based prioritization, providing 5–550% higher fault detection effectiveness. Overall, our approach to prioritizing metamorphic relations leads to increased fault detection effectiveness and reduced average fault detection time. This improvement in efficiency can result in significant time and cost savings when applying metamorphic testing to machine learning and deep learning systems.

Heart patient health monitoring system using invasive and non-invasive measurement

The abnormal heart conduction, known as arrhythmia, can contribute to cardiac diseases that carry the risk of fatal consequences. Healthcare professionals typically use electrocardiogram (ECG) signals and certain preliminary tests to identify abnormal patterns in a patient’s cardiac activity. To assess the overall cardiac health condition, cardiac specialists monitor these activities separately. This procedure may be arduous and time-intensive, potentially impacting the patient’s well-being. This study automates and introduces a novel solution for predicting the cardiac health conditions, specifically identifying cardiac morbidity and arrhythmia in patients by using invasive and non-invasive measurements. The experimental analyses conducted in medical studies entail extremely sensitive data and any partial or biased diagnoses in this field are deemed unacceptable. Therefore, this research aims to introduce a new concept of determining the uncertainty level of machine learning algorithms using information entropy. To assess the effectiveness of machine learning algorithms information entropy can be considered as a unique performance evaluator of the machine learning algorithm which is not selected previously any studies within the realm of bio-computational research. This experiment was conducted on arrhythmia and heart disease datasets collected from Massachusetts Institute of Technology-Berth Israel Hospital-arrhythmia (DB-1) and Cleveland Heart Disease (DB-2), respectively. Our framework consists of four significant steps: 1) Data acquisition, 2) Feature preprocessing approach, 3) Implementation of learning algorithms, and 4) Information Entropy. The results demonstrate the average performance in terms of accuracy achieved by the classification algorithms: Neural Network (NN) achieved 99.74%, K-Nearest Neighbor (KNN) 98.98%, Support Vector Machine (SVM) 99.37%, Random Forest (RF) 99.76 % and Naïve Bayes (NB) 98.66% respectively. We believe that this study paves the way for further research, offering a framework for identifying cardiac health conditions through machine learning techniques.

SALA-IoT: Self-Reduced Internet of Things With Learning Automaton Sleep Scheduling Algorithm

The extensive development of wireless communications has led to the popularity of the Self-Powered Internet of Things (SPIoT) networks. Even though significant advances made in keeping energy consumption at a low level, making such networks live longer are still one of the biggest challenges. In particular, a significant question is how to cover the maximum range of the environment by the sensor nodes with the lowest amount of energy. In this research, we have proposed a sleep Scheduling Algorithm based on Learning Automaton for IoT (SALA-IoT), utilizing machine learning approaches to find the optimal set of sensor nodes that can cover a wide range of the environment. This algorithm consists of learning and covering phases, in which the essential sensor nodes are activated, and the rest are turned off. Although the implementation of learning algorithms requires a high energy level, after the learning phase’s performance, the network gathers complete information about the status of the nodes. It can meet the needs of the network with a limited amount of energy. We have evaluated the proposed algorithm with several simulation scenarios and considered the coverage area and the number of active sensor nodes as the main evaluation metrics. The simulation results show that the sensors sensing and covering ranges are directly related to their transmitted power, and the transmitted power of the sensors can be adjusted to match the expected network requirements in terms of primary coverage factors; the proposed method improved by 6.743 for 50 nodes and 5.204 for 100 nodes, respectively, for distance and intervals in a different number of nodes.

Digital Hardware Implementation of ReSuMe Learning Algorithm for Spiking Neural Networks.

Within this paper, we demonstrate the feasibility of the FPGA implementation as well as the 180nm CMOS circuit design of a particular biologically plausible supervised learning algorithm (ReSuMe). Based on the Spike-Timing-Dependent Plasticity (STDP) learning phenomenon, this design proposes a fully configurable implementation of STDP learning window function to adjust the learning process for different applications, optimizing results for each use case. The CMOS implementation in 180nm technology node supplied with 1.8V shows a core area of 0.78mm2 and verifies the suitability of an on-chip ReSuMe learning algorithm implementation and its capability of integration with a multitude of external and already designed structures of Spiking Neural Networks (SNNs).

Typological gaps in iambic nonfinality correlate with learning difficulty

This paper discusses gaps in stress typology that are unexpected from the perspective of a foot-based theory and shows that the patterns pose difficulties for a computationally implemented learning algorithm. The unattested patterns result from combining theoretical elements whose effects are generally well-attested, including iambic footing, nonfinality, word edge alignment and a foot binarity requirement. The patterns can be found amongst the 124 target stress systems constructed by Tesar and Smolensky (2000) as a test of their approach to hidden structure learning. A learner with a Maximum Entropy grammar that uses a form of Expectation Maximization to deal with hidden structure was found to often fail on these unattested languages.

IoT Multi-Vector Cyberattack Detection Based on Machine Learning Algorithms: Traffic Features Analysis, Experiments, and Efficiency

Cybersecurity is a common Internet of Things security challenge. The lack of security in IoT devices has led to a great number of devices being compromised, with threats from both inside and outside the IoT infrastructure. Attacks on the IoT infrastructure result in device hacking, data theft, financial loss, instability, or even physical damage to devices. This requires the development of new approaches to ensure high-security levels in IoT infrastructure. To solve this problem, we propose a new approach for IoT cyberattack detection based on machine learning algorithms. The core of the method involves network traffic analyses that IoT devices generate during communication. The proposed approach deals with the set of network traffic features that may indicate the presence of cyberattacks in the IoT infrastructure and compromised IoT devices. Based on the obtained features for each IoT device, the feature vectors are formed. To conclude the possible attack presence, machine learning algorithms were employed. We assessed the complexity and time of machine learning algorithm implementation considering multi-vector cyberattacks on IoT infrastructure. Experiments were conducted to approve the method’s efficiency. The results demonstrated that the network traffic feature-based approach allows the detection of multi-vector cyberattacks with high efficiency.

Tomato Leaf Diseases Classification Based on Leaf Images: A Comparison between Classical Machine Learning and Deep Learning Methods

Tomato production can be greatly reduced due to various diseases, such as bacterial spot, early blight, and leaf mold. Rapid recognition and timely treatment of diseases can minimize tomato production loss. Nowadays, a large number of researchers (including different institutes, laboratories, and universities) have developed and examined various traditional machine learning (ML) and deep learning (DL) algorithms for plant disease classification. However, through pass survey analysis, we found that there are no studies comparing the classification performance of ML and DL for the tomato disease classification problem. The performance and outcomes of different traditional ML and DL (a subset of ML) methods may vary depending on the datasets used and the tasks to be solved. This study generally aimed to identify the most suitable ML/DL models for the PlantVillage tomato dataset and the tomato disease classification problem. For machine learning algorithm implementation, we used different methods to extract disease features manually. In our study, we extracted a total of 52 texture features using local binary pattern (LBP) and gray level co-occurrence matrix (GLCM) methods and 105 color features using color moment and color histogram methods. Among all the feature extraction methods, the COLOR+GLCM method obtained the best result. By comparing the different methods, we found that the metrics (accuracy, precision, recall, F1 score) of the tested deep learning networks (AlexNet, VGG16, ResNet34, EfficientNet-b0, and MobileNetV2) were all better than those of the measured machine learning algorithms (support vector machine (SVM), k-nearest neighbor (kNN), and random forest (RF)). Furthermore, we found that, for our dataset and classification task, among the tested ML/DL algorithms, the ResNet34 network obtained the best results, with accuracy of 99.7%, precision of 99.6%, recall of 99.7%, and F1 score of 99.7%.

The Efficiency of Learning Methodology for Privacy Protection in Context-aware Environment during the COVID-19 Pandemic.

When the COVID-19 coronavirus hit, the context-aware application users were willing to relax their context privacy preferences during the lockdown to cope their lives while staying home. Such disturbance in the privacy behavior affected the performance of Machine Learning (ML) algorithm that is trained on normal behavior. In this paper, we present the impact of the pandemic on the efficiency of the learning algorithm implementation of a privacy protection system. The system is composed of three modules, in this work we focus on Privacy Preferences Manager (PPM) module which is implemented using hybrid methodology based on a Statistical Model (SM) and Logistic Regression (LR) learning algorithm. The efficiency of the hybrid methodology is assessed using two real-world datasets collected prior and during the COVID-19 pandemic. The results show that the pandemic significantly impacted the efficiency of the hybrid methodology by 13.05% and 15.22% for the accuracy and F1 score respectively.

Data driven estimation of electric vehicle battery state-of-charge informed by automotive simulations and multi-physics modeling

State-of-charge (SOC) estimation in a lithium-ion battery (LIB) is a crucial task of the battery management system (BMS) in battery electric vehicle (BEV) applications. In this work, we propose a modeling framework for SOC estimation using different machine learning (ML) methods, i.e. support vector regressor (SVR), artificial neural network (ANN), and long-short term memory (LSTM) network. The necessary training data have been developed using Matlab/Simulink automotive simulations of BEV, integrated with an electrochemical Comsol Multiphysics model of LIBs. The developed multi-physics model of BEV and LIBs operation allows to investigate the effect of driving conditions on the electrochemical and degradation (i.e., the solid electrolyte interphase – SEI – formation and decomposition) processes occurring inside batteries of different chemistries adopted in the Tesla S and Nissan Leaf BEVs. Our study remarks also the importance of taking into account the different components of BEV in the development of informative datasets, which are required for the implementation of learning algorithms for SOC evaluation. Thus, the proposed work establishes a basis for the generation of realistic training data based on simulations of BEV and LIBs dynamic response, which allows a more precise SOC estimation based on data-driven approaches.

Data Science Methods and Machine Learning Algorithm Implementations for Customized Pratical Usage

Due to the unprecedented rise of data content over the last decade, an opportunity for databased personalization and analysis has become a norm in the modern world. By implementing Machine Learning algorithms and Data Science methods no industry remained unchanged. This paper applied these methods and algorithms in personal, practical examples in order to see their benefits in our day-today lives. For the purpose of this case study, we analyzed three cases: a personal movie recommender, messages analysis and real estate trends and predictions on the local market. For this research we used global and personal data, and applied a suitable machine learning model. The purpose of this paper is to establish how one individual, and in what measure, with the use of these new technological tools, can ease his decision making process and manage a more tailored lifestyle.

Machine Learning Framework To Analyze Against Spear Phishing

The objective of this paper is to design and implement machine learning based ensemble algorithm on dataset to fit into the models that can be understood and executed by machines. In this paper we discussed different algorithms and machine learning concepts that can be implemented on the datasets, we taken email spam filter dataset for experiment and analysis, as the Advanced persistent threat the latest threat is intruded using the emails and major intrusion is done through spam emails. Machine learning uses different datamining techniques and mechanisms and accepts the input-data and gives the output as the statistical analysis. We implemented different email classification algorithms on the datasets bsed on spam and ham emails where spear phishing methods are identified and implemented different classification and regression methods to get the accurate results. In this paper for the better results in spite of existing algorithms we introduced the ensemble methods such as boosting, bagging, stacking and voting for much accuracy and higher level of classification and combining different algorithm. This paper will measure different machine learning algorithms performance on spam email filtering on the huge datasets. The framework provides implementation of learning algorithms that you can apply to larger datasets. An obvious approach to making decisions more reliable is to combine the output of different models. We even compared the existing algorithms and proposed algorithm; comparison tables are drawn along with the statistical analysis, data and graphical analysis is given.

Learning in Memristive Neural Network Architectures Using Analog Backpropagation Circuits

The on-chip implementation of learning algorithms would speed-up the training of neural networks in crossbar arrays. The circuit level design and implementation of backpropagation algorithm using gradient descent operation for neural network architectures is an open problem. In this paper, we proposed the analog backpropagation learning circuits for various memristive learning architectures, such as Deep Neural Network (DNN), Binary Neural Network (BNN), Multiple Neural Network (MNN), Hierarchical Temporal Memory (HTM) and Long-Short Term Memory (LSTM). The circuit design and verification is done using TSMC 180nm CMOS process models, and TiO2 based memristor models. The application level validations of the system are done using XOR problem, MNIST character and Yale face image databases

An analysis of hierarchical text classification using word embeddings

Efficient distributed numerical word representation models (word embeddings) combined with modern machine learning algorithms have recently yielded considerable improvement on automatic document classification tasks. However, the effectiveness of such techniques has not been assessed for the hierarchical text classification (HTC) yet. This study investigates the application of those models and algorithms on this specific problem by means of experimentation and analysis. We trained classification models with prominent machine learning algorithm implementations—fastText, XGBoost, SVM, and Keras’ CNN—and noticeable word embeddings generation methods—GloVe, word2vec, and fastText—with publicly available data and evaluated them with measures specifically appropriate for the hierarchical context. FastText achieved an lcaF1 of 0.893 on a single-labeled version of the RCV1 dataset. An analysis indicates that using word embeddings and its flavors is a very promising approach for HTC.

What can Android mobile app developers do about the energy consumption of machine learning?

Machine learning is a popular method of learning functions from data to represent and to classify sensor inputs, multimedia, emails, and calendar events. Smartphone applications have been integrating more and more intelligence in the form of machine learning. Machine learning functionality now appears on most smartphones as voice recognition, spell checking, word disambiguation, face recognition, translation, spatial reasoning, and even natural language summarization. Excited app developers who want to use machine learning on mobile devices face one serious constraint that they did not face on desktop computers or cloud virtual machines: the end-user’s mobile device has limited battery life, thus computationally intensive tasks can harm end users’ phone availability by draining batteries of their stored energy. Currently, there are few guidelines for developers who want to employ machine learning on mobile devices yet are concerned about software energy consumption of their applications. In this paper, we combine empirical measurements of different machine learning algorithm implementations with complexity theory to provide concrete and theoretically grounded recommendations to developers who want to employ machine learning on smartphones. We conclude that some implementations of algorithms, such as J48, MLP, and SMO, do generally perform better than others in terms of energy consumption and accuracy, and that energy consumption is well-correlated to algorithmic complexity. However, to achieve optimal results a developer must consider their specific application as many factors — dataset size, number of data attributes, whether the model will require updating, etc. — affect which machine learning algorithm and implementation will provide the best results.

The More the Merrier? Effects of Humanlike Learning Abilities on Humans’ Perception and Evaluation of a Robot

In this paper, we present three experimental studies in which subjects trained a robot to do a card game via reinforcement learning. In the first two studies participants interacted with the robot either without any learning ability (control group) or with one out of three versions of a learning algorithm implementing gradually aspects of more humanlike learning abilities. Results show that the implementation of a learning algorithm had positive effects regarding the evaluation of the robot, its learning abilities and the interaction. We found that more humanlike learning abilities do not always lead to better performance and evaluation and that results were to some extend influenced by longer or shorter interaction times. In a third study, we additionally explored the influence of other behavioral variations such as low or high verbal skills and interaction modalities in perceived intelligence of the robot irrespective of the implemented learning algorithm, but did not find significant effects. Results are discussed with regard to the socialness of future interaction scenarios.

Low-VDD Operation of SRAM Synaptic Array for Implementing Ternary Neural Network

For Internet of Things (IoT) edge devices, it is very attractive to have the local sensemaking capability instead of sending all the data back to the cloud for information processing. For image pattern recognition, neuro-inspired machine learning algorithms have demonstrated enormous powerfulness. To effectively implement learning algorithms on-chip for IoT edge devices, on-chip synaptic memory architectures have been proposed to implement the key operations such as weighted-sum or matrix-vector multiplication. In this paper, we proposed a low-power design of static random access memory (SRAM) synaptic array for implementing a low-precision ternary neural network. We experimentally demonstrated that the supply voltage (VDD) of the SRAM array could be aggressively reduced to a level, where the SRAM cell is susceptible to bit failures. The testing results from 65-nm SRAM chips indicate that VDD could be reduced from the nominal 1–0.55 V (or 0.5 V) with a bit error rate ~0.23% (or ~1.56%), which only introduced ~0.08% (or ~1.68%) degradation in the classification accuracy. As a result, the power consumption could be reduced by more than $8\times $ (or $10\times $ ).

Anatomy of machine learning algorithm implementations in MPI, Spark, and Flink

With the ever-increasing need to analyze large amounts of data to get useful insights, it is essential to develop complex parallel machine learning algorithms that can scale with data and number of parallel processes. These algorithms need to run on large data sets as well as they need to be executed with minimal time in order to extract useful information in a time-constrained environment. Message passing interface (MPI) is a widely used model for developing such algorithms in high-performance computing paradigm, while Apache Spark and Apache Flink are emerging as big data platforms for large-scale parallel machine learning. Even though these big data frameworks are designed differently, they follow the data flow model for execution and user APIs. Data flow model offers fundamentally different capabilities than the MPI execution model, but the same type of parallelism can be used in applications developed in both models. This article presents three distinct machine learning algorithms implemented in MPI, Spark, and Flink and compares their performance and identifies strengths and weaknesses in each platform.

Filament formation in lithium niobate memristors supports neuromorphic programming capability

Memristor crossbars are capable of implementing learning algorithms in a much more energy and area efficient manner compared to traditional systems. However, the programmable nature of memristor crossbars must first be explored on a smaller scale to see which memristor device structures are most suitable for applications in reconfigurable computing. In this paper, we demonstrate the programmability of memristor devices with filamentary switching based on LiNbO3, a new resistive switching oxide. We show that a range of resistance values can be set within these memristor devices using a pulse train for programming. We also show that a neuromorphic crossbar containing eight memristors was capable of correctly implementing an OR function. This work demonstrates that lithium niobate memristors are strong candidates for use in neuromorphic computing.

Reconfigurable neuromorphic crossbars based on titanium oxide memristors

Memristor crossbars are capable of implementing learning algorithms in a much more energy- and area-efficient manner compared with traditional systems. However, the programmable nature of memristor crossbars must first be explored on a smaller scale to see if physical devices are suitable for applications of reconfigurable computing. The reconfigurability of these devices through small scale memristor crossbar implementations is demonstrated. It is shown that a crossbar containing eight memristors is capable of learning several different two-input Boolean logic functions. A strong foundation is provided to build on demonstrating that physical memristor crossbars can be programmed as linear classifiers.