Showing 27 results for Learning
M. Hariri, S. B. Shokouhi, N. Mozayani,
Volume 4, Issue 3 (10-2008)
Abstract
Dealing with uncertainty is one of the most critical problems in complicated
pattern recognition subjects. In this paper, we modify the structure of a useful Unsupervised
Fuzzy Neural Network (UFNN) of Kwan and Cai, and compose a new FNN with 6 types of
fuzzy neurons and its associated self organizing supervised learning algorithm. This
improved five-layer feed forward Supervised Fuzzy Neural Network (SFNN) is used for
classification and identification of shifted and distorted training patterns. It is generally
useful for those flexible patterns which are not certainly identifiable upon their features. To
show the identification capability of our proposed network, we used fingerprint, as the most
flexible and varied pattern. After feature extraction of different shapes of fingerprints, the
pattern of these features, “feature-map”, is applied to the network. The network first
fuzzifies the pattern and then computes its similarities to all of the learned pattern classes.
The network eventually selects the learned pattern of highest similarity and returns its
specific class as a non fuzzy output. To test our FNN, we applied the standard (NIST
database) and our databases (with 176×224 dimensions). The feature-maps of these
fingerprints contain two types of minutiae and three types of singular points, each of them
is represented by 22×28 pixels, which is less than real size and suitable for real time
applications. The feature maps are applied to the FNN as training patterns. Upon its setting
parameters, the network discriminates 3 to 7 subclasses for each main classes assigned to
one of the subjects.
M. M Daevaeiha, M. R Homaeinezhad, M. Akraminia, A. Ghaffari, M. Atarod,
Volume 6, Issue 3 (9-2010)
Abstract
The aim of this study is to introduce a new methodology for isolation of ectopic
rhythms of ambulatory electrocardiogram (ECG) holter data via appropriate statistical
analyses imposing reasonable computational burden. First, the events of the ECG signal are
detected and delineated using a robust wavelet-based algorithm. Then, using Binary
Neyman-Pearson Radius test, an appropriate classifier is designed to categorize ventricular
complexes into "Normal + Premature Atrial Contraction (PAC)" and "Premature
Ventricular Contraction (PVC)" beats. Afterwards, an innovative measure is defined based
on wavelet transform of the delineated P-wave namely as P-Wave Strength Factor (PSF)
used for the evaluation of the P-wave power. Finally, ventricular contractions pursuing
weak P-waves are categorized as PAC complexes however, those ensuing strong P-waves
are specified as normal complexes. The discriminant quality of the PSF-based feature space
was evaluated by a modified learning vector quantization (MLVQ) classifier trained with
the original QRS complexes and corresponding Discrete Wavelet Transform (DWT) dyadic
scale. Also, performance of the proposed Neyman-Pearson Classifier (NPC) is compared
with the MLVQ and Support Vector Machine (SVM) classifiers using a common feature
space. The processing speed of the proposed algorithm is more than 176,000 samples/sec
showing desirable heart arrhythmia classification performance. The performance of the
proposed two-lead NPC algorithm is compared with MLVQ and SVM classifiers and the
obtained results indicate the validity of the proposed method. To justify the newly defined
feature space (σi1, σi2, PSFi), a NPC with the proposed feature space and a MLVQ
classification algorithm trained with the original complex and its corresponding DWT as
well as RR interval are considered and their performances were compared with each other.
An accuracy difference about 0.15% indicates acceptable discriminant quality of the
properly selected feature elements. The proposed algorithm was applied to holter data of
the DAY general hospital (more than 1,500,000 beats) and the average values of Se =
99.73% and P+ = 99.58% were achieved for sensitivity and positive predictivity,
respectively.
M. R. Homaeinezhad, E. Tavakkoli, A. Afshar, A. Atyabi, A. Ghaffari,
Volume 7, Issue 2 (6-2011)
Abstract
The paper addresses a new QRS complex geometrical feature extraction technique as well as its application for electrocardiogram (ECG) supervised hybrid (fusion) beat-type classification. To this end, after detection and delineation of the major events of ECG signal via a robust algorithm, each QRS region and also its corresponding discrete wavelet transform (DWT) are supposed as virtual images and each of them is divided into eight polar sectors. Then, the curve length of each excerpted segment is calculated and is used as the element of the feature space. To increase the robustness of the proposed classification algorithm versus noise, artifacts and arrhythmic outliers, a fusion structure consisting of three Multi Layer Perceptron-Back Propagation (MLP-BP) neural networks with different topologies and one Adaptive Network Fuzzy Inference System (ANFIS) were designed and implemented. To show the merit of the new proposed algorithm, it was applied to all MIT-BIH Arrhythmia Database records and the discrimination power of the classifier in isolation of different beat types of each record was assessed and as the result, the average accuracy value Acc=98.27% was obtained. Also, the proposed method was applied to 8 number of arrhythmias (Normal, LBBB, RBBB, PVC, APB, VE, PB, VF) belonging to 19 number of the aforementioned database and the average value of Acc=98.08% was achieved. To evaluate performance quality of the new proposed hybrid learning machine, the obtained results were compared with similar peer-reviewed studies in this area.
Sh. Yousefi, M. Parsa Moghaddam, V. Johari Majd,
Volume 7, Issue 3 (9-2011)
Abstract
In this paper, an agent-based structure of the electricity retail market is presented based on which day-ahead (DA) energy procurement for customers is modeled. Here, we focus on operation of only one Retail Energy Provider (REP) agent who purchases energy from DA pool-based wholesale market and offers DA real time tariffs to a group of its customers. As a model of customer response to the offered real time prices, an hourly acceptance function is proposed in order to represent the hourly changes in the customer’s effective demand according to the prices. Here, Q-learning (QL) approach is applied in day-ahead real time pricing for the customers enabling the REP agent to discover which price yields the most benefit through a trial-and-error search. Numerical studies are presented based on New England day-ahead market data which include comparing the results of RTP based on QL approach with that of genetic-based pricing.
A. Khoshsaadat , M. R. Mosavi, J. S. Moghani,
Volume 10, Issue 3 (9-2014)
Abstract
Static Synchronous Series Compensator (SSSC) is a series compensating Flexible AC Transmission System (FACTS) controller for maintaining to the power flow control on a transmission line by injecting a voltage in quadrature with the line current and in series mode with the line. In this work, an Adaptive Network-based Fuzzy Inference System controller (ANFISC) has been proposed for controlling of the SSSC-based damping system and applied to a Single Machine Infinite Bus (SMIB) power system. For implementation of the learning process in this controller, we use of the one approach of the learning ability that named as Forward Signal and Backward Error Back-Propagation (FSBEBP) method for improving of the system efficiency. This artificial intelligence-based control model leads to a controller with adaptive structure, improved correctness, high damping ability and dynamic performance. System implementation is easy and it requires 49 fuzzy rules for inference engine of the system. As compared with the other complex neuro-fuzzy systems, this controller has medium number of the fuzzy rules and low number of layers, but it has high accuracy. In order to demonstrate of the proposed controller ability, it is simulated and its output compared with that of classic Lead-Lag-based Controller (LLC) and PI controller.
H. Bakhshandeh, A. Akbari Foroud,
Volume 12, Issue 1 (3-2016)
Abstract
This paper addresses the possibility of capacity withholding by energy producers, who seek to increase the market price and their own profits. The energy market is simulated as an iterative game, where each state game corresponds to an hourly energy auction with uniform pricing mechanism. The producers are modeled as agents that interact with their environment through reinforcement learning (RL) algorithm. Each producer submits step-wise offer curves, which include the quantity-price pairs, to independent system operator (ISO) under incomplete information. An experimental change is employed in the producer's profit maximization model that causes the iterative algorithm converge to a withholding bidding value. The producer can withhold the energy of his own generating unit in a continuous range of its available capacity. The RL relation is developed to prevent from becoming invalid in certain situations. The results on a small test system demonstrate the emergence of the capacity withholding by the producers and its effect on the market price.
H. Shayeghi, A. Younesi,
Volume 13, Issue 4 (12-2017)
Abstract
This paper presents an online two-stage Q-learning based multi-agent (MA) controller for load frequency control (LFC) in an interconnected multi-area multi-source power system integrated with distributed energy resources (DERs). The proposed control strategy consists of two stages. The first stage is employed a PID controller which its parameters are designed using sine cosine optimization (SCO) algorithm and are fixed. The second one is a reinforcement learning (RL) based supplementary controller that has a flexible structure and improves the output of the first stage adaptively based on the system dynamical behavior. Due to the use of RL paradigm integrated with PID controller in this strategy, it is called RL-PID controller. The primary motivation for the integration of RL technique with PID controller is to make the existing local controllers in the industry compatible to reduce the control efforts and system costs. This novel control strategy combines the advantages of the PID controller with adaptive behavior of MA to achieve the desired level of robust performance under different kind of uncertainties caused by stochastically power generation of DERs, plant operational condition changes, and physical nonlinearities of the system. The suggested decentralized controller is composed of the autonomous intelligent agents, who learn the optimal control policy from interaction with the system. These agents update their knowledge about the system dynamics continuously to achieve a good frequency oscillation damping under various severe disturbances without any knowledge of them. It leads to an adaptive control structure to solve LFC problem in the multi-source power system with stochastic DERs. The results of RL-PID controller in comparison to the traditional PID and fuzzy-PID controllers is verified in a multi-area power system integrated with DERs through some performance indices.
A. Younesi, H. Shayeghi,
Volume 15, Issue 1 (3-2019)
Abstract
The purpose of this paper is to design a supplementary controller for traditional PID controller in order to damp the frequency oscillations in a micro-grid. Q-learning, which is used for supervise a classical PID controller in this paper, is a model free and a simple solution method of reinforcement learning (RL). RL is one of the branches of the machine learning, which is the main solution method of Markov decision process (MDPs). The proposed control mechanism is consisting of two main parts. The first part is a classical PID controller which is fixed tuned using Salp swarm algorithm. The second part is a Q‑learning based control strategy which is consistent and updates its characteristics according to the changes in the system continuously. Eventually, a hybrid micro-grid is considered to evaluate the performance of the suggested control method compared to classical PID and fractional order fuzzy PID (FOFPID) controllers. The considered hybrid system is consisting of renewable energy resources such as solar-thermal power station (STPS) and wind turbine generation (WTG), along with several energy storage devices such as batteries, flywheel and ultra-capacitor with physical constraints and time delays. Simulations are carried out in various realistic scenarios considering system parameter variations along with changing in operating conditions. Results indicate that the proposed control strategy has an excellent dynamic response compared to the traditional PID and FOFPID controllers for damping the frequency oscillations in different operating conditions.
S. Mavaddati,
Volume 15, Issue 2 (6-2019)
Abstract
A new single channel singing voice separation algorithm is presented in this paper. This field of signal processing provides important capability in various areas dealing with singer identification, voice recognition, data retrieval. This separation procedure is done using a decomposition model based on the spectrogram of singing voice signals. The novelty of the proposed separation algorithm is related to different issues listed in the following: 1) The decomposition scheme employs the vocal and music models learned using sparse non-negative matrix factorization algorithm. The vocal signal and music accompaniment can be considered as sparse and low-rank components of a singing voice segment, respectively. 2) An alternating factorization algorithm is used to decompose input data based on the modeled structures of the vocal and musical components. 3) A voice activity detection algorithm is introduced based on the energy of coding coefficients matrix in the training step to learn the basis vectors that are related to instrumental parts. 4) In the separation phase, these non-vocal atoms are updated to the new test conditions using the domain transfer approach to result in a proper separation procedure with low reconstruction error. The performance evaluation of the proposed algorithm is done using different measures and leads to significantly better results in comparison with the earlier methods in this context and the traditional procedures. The average improvement values of the proposed separation algorithm for PESQ, fwSegSNR, SDI, and GNSDR measures in comparison with previous separation methods in two defined test scenario and three mentioned SMR levels are 0.53, 0.84, 0.39, and 2.19, respectively.
A. Afrush, M. Shahriari-Kahkeshi,
Volume 15, Issue 2 (6-2019)
Abstract
This paper proposes an adaptive approximation-based controller for uncertain strict-feedback nonlinear systems with unknown dead-zone nonlinearity. Dead-zone constraint is represented as a combination of a linear system with a disturbance-like term. This work invokes neural networks (NNs) as a linear-in-parameter approximator to model uncertain nonlinear functions that appear in virtual and actual control laws. Minimal learning parameter (MLP) algorithm is proposed to decrease the computational load, the number of adjustable parameters, and to avoid the “explosion of learning parameters” problem. An adaptive TSK-type fuzzy system is proposed to estimate the disturbance-like term in the dead-zone description which further will be used to compensate the effect of the dead-zone, and it does not require the availability of the dead-zone input. Then, the proposed method based on the dynamic surface control (DSC) method is designed which avoids the “explosion of complexity” problem. Proposed scheme deals with dead-zone nonlinearity and uncertain dynamics without requiring the availability of any knowledge about them, and it develops a control input without singularity concern. Stability analysis shows that all the signals of the closed-loop system are semi-globally uniformly ultimately bounded and the tracking error converges to the vicinity of the origin. Simulation and comparison results verify the acceptable performance of the presented controller.
N. Sayyadi Shahraki, S. H. Zahiri,
Volume 16, Issue 2 (6-2020)
Abstract
In this paper, we propose an efficient approach to design optimization of analog circuits that is based on the reinforcement learning method. In this work, Multi-Objective Learning Automata (MOLA) is used to design a two-stage CMOS operational amplifier (op-amp) in 0.25μm technology. The aim is optimizing power consumption and area so as to achieve minimum Total Optimality Index (TOI), as a new and comprehensive proposed criterion, and also meet different design specifications such as DC gain, Gain-Band Width product (GBW), Phase Margin (PM), Slew Rate (SR), Common Mode Rejection Ratio (CMRR), Power Supply Rejection Ratio (PSRR), etc. The proposed MOLA contains several automata and each automaton is responsible for searching one dimension. The workability of the proposed approach is evaluated in comparison with the most well-known category of intelligent meta-heuristic Multi-Objective Optimization (MOO) methods such as Particle Swarm Optimization (PSO), Inclined Planes system Optimization (IPO), Gray Wolf Optimization (GWO) and Non-dominated Sorting Genetic Algorithm II (NSGA-II). The performance of the proposed MOLA is demonstrated in finding optimal Pareto fronts with two criteria Overall Non-dominated Vector Generation (ONVG) and Spacing (SP). In simulations, for the desired application, it has been shown through Computer-Aided Design (CAD) tool that MOLA-based solutions produce better results.
M. Dodangeh, N. Ghaffarzadeh,
Volume 18, Issue 4 (12-2022)
Abstract
An intelligent strategy for the protection of AC microgrids is presented in this paper. This method was halving to an initial signal processing step and a machine learning-based forecasting step. The initial stage investigates currents and voltages with a window-based approach based on the dynamic decomposition method (DDM) and then involves the norms of the signals to the resultant DDM data. The results of the currents and voltages norms are applied as features for a topology data analysis algorithm for fault type classifying in the AC microgrid for fault location purposes. The Algorithm was tested on a microgrid that operates with precision equal to 100% in fault classification and a mean error lower than 20 m when forecasting the fault location. The proposed method robustly operates in sampling frequency, fault resistance variation, and noisy and high impedance fault conditions.
Mohammad Hasheminejad,
Volume 19, Issue 4 (12-2023)
Abstract
The Nonparametric Speech Kernel (NSK), a nonparametric kernel technique, is presented in this study as a novel way to improve Speech Emotion Recognition (SER). The method aims to effectively reduce the size of speech features to improve recognition accuracy. The proposed approach addresses the need for efficient and compact low-dimensional features for speech emotion recognition. Having acknowledged the intrinsic distinctions between speech and picture data, we have refined the Kernel Nonparametric Weighted Feature Extraction (KNWFE) formulation to suggest NSK, which is especially intended for speech emotion identification. The output of NSK can be used as input features for deep learning models such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), or hybrid architectures. In deep learning, NSK can also be used as a kernel function for kernel-based methods such as kernelized support vector machines (SVM) or kernelized neural networks. Our tests demonstrate that NSK outperforms current techniques, outperforming the best-tested approach by 5.02% and 3.05%, respectively, with an average accuracy of 96.568% for the Persian speech emotion dataset and 82.56% for the Berlin speech emotion dataset.
Ramin Safaeian, Mahmoud Tabandeh ,
Volume 20, Issue 2 (6-2024)
Abstract
Directed Acyclic Graphs stand as one of the prevailing approaches for representing causal relationships within a set of variables. With observational or interventional data, certain undirected edges within a causal DAG can be oriented. Performing intervention can be done in two different settings, passive and active. Here, we prove that an optimal intervention set can be obtained based on the minimum vertex cover of a graph. We propose an algorithm that efficiently identifies such an optimal intervention set for chordal graphs within polynomial time. Performing intervention on this optimal set recovers all the undirected edges in graph G, regardless of the underlying ground truth DAG. Furthermore, we present an algorithm for evaluating the performance of passive algorithms. This evaluation provides insights into how many intervention steps of a specific algorithm are required to recover all edges in the causal graph for any possible underlying ground truth in the equivalence class. Experimental findings underscore that the number of nodes in the optimal intervention set increases with growing the number of nodes in a graph, where the edge density is fixed, and also increases with the rising edge density in a graph with a fixed number of nodes.
Biswapriyo Sen, Maharishi Kashyap, Jitendra Singh Tamang, Sital Sharma, Rijhi Dey,
Volume 20, Issue 2 (6-2024)
Abstract
Cardiovascular arrhythmia is indeed one of the most prevalent cardiac issues globally. In this paper, the primary objective was to develop and evaluate an automated classification system. This system utilizes a comprehensive database of electro- cardiogram (ECG) data, with a particular focus on improving the detection of minority arrhythmia classes.
In this study, the focus was on investigating the performance of three different supervised machine learning models in the context of arrhythmia detection. These models included Support Vector Machine (SVM), Logistic Regression (LR) and Random Forest (RF). An analysis was conducted using real inter-patient electrocardiogram (ECG) records, which is a more realistic scenario in a clinical environment where ECG data comes from various patients.
The study evaluated the models’ performances based on four important metrics: accuracy, precision, recall, and f1-score. After thorough experimentation, the results highlighted that the Random Forest (RF) classifier outperformed the other methods in all of the metrics used in the experiments. This classifier achieved an impressive accuracy of 0.94, indicating its effectiveness in accurately detecting arrhythmia in diverse ECG signals collected from different patients.
Eisa Zarepour, Mohammad Reza Mohammadi, Morteza Zakeri-Nasrabadi, Sara Aein, Razieh Sangsari, Leila Taheri, Mojtaba Akbari, Ali Zabihallahpour,
Volume 20, Issue 3 (9-2024)
Abstract
Using mobile phones for medical applications are proliferating due to high-quality embedded sensors. Jaundice, a yellow discoloration of the skin caused by excess bilirubin, is a prevalent physiological problem in newborns. While moderate amounts of bilirubin are safe in healthy newborns, extreme levels are fatal and cause devastating and irreversible brain damage. Accurate tests to measure jaundice require a blood draw or dedicated clinical devices facing difficulty where clinical technology is unavailable. This paper presents a smartphone-based screening tool to detect neonatal hyperbilirubinemia caused by the high bilirubin production rate. A machine learning regression model is trained on a pretty large dataset of images, including 446 samples, taken from newborns' sternum skin in four medical centers in Iran. The learned model is then used to estimate the level of bilirubin. Experimental results show a mean absolute error of 1.807 mg/dl and a correlation of 0.701 between predicted bilirubin by the proposed method and the TSB values as ground truth.
Sandra D’souza, Niranjan Reddy S, Saikonda Krishna Tarun, Sohan P, Aneesha Acharya K,
Volume 20, Issue 4 (11-2024)
Abstract
The incidence of heart-related illnesses is on the rise worldwide. Heart diseases are primarily caused by a multitude of parameters, including high blood pressure, diabetes, and excessive cholesterol, which are controlled by poor dietary and lifestyle choices. The growth in cardiovascular diseases (CVD) is mostly due to several other behaviors, such as smoking, drinking, and sleeplessness. In the research, machine learning-based prediction methods work on the audio recordings of heartbeats known as phonocardiograms (PCG) to develop an algorithm that differentiates a normal healthy heart from an abnormal heart based on the heart sounds. The data set consists of 831 normal and 260 abnormal data, and the duration of each sample is 5 seconds. Features extracted from the data are up-sampled and applied to the logistic regression and random forest classification models. The developed models record a classification accuracy of 71% for logistic regression and 94% for the random forest model. Further, artificial neural networks (ANN) and Deep learning networks have been trained to improve performance and demonstrated an accuracy of 94.5%.
Elahe Moradi,
Volume 20, Issue 4 (11-2024)
Abstract
With the intricate interplay between clinical and pathological data in coronary heart disease (CHD) diagnosis, there is a growing interest among researchers and healthcare providers in developing more accurate and reliable predictive methods. In this paper, we propose a new method entitled the robust artificial neural network classifier (RANNC) technique for the prediction of CHD. The dataset CHD in this paper has imbalanced data, and in addition, it has some outlier values. The dataset consists of information related to 4240 samples with 16 attributes. Due to the presence of outliers, a robust method has been used to scale the dataset. On the other hand, due to the imbalance of CHD data, three data balancing methods, including Random Over Sampling (ROS), Synthetic Minority Over Sampling Technique (SMOTE), and Adaptive Synthetic Sampling (ADASYN) approaches, have been applied to the CHD data set. Also, six artificial intelligence algorithms, including LRC, DTC, RFC, KNNC, SVC, and ANN, have been evaluated on the considered dataset with criteria such as precision, accuracy, recall, F1-score, and MCC. The RANNC, leveraging ADASYN to address data imbalance and outliers, significantly improved CHD diagnostic accuracy and the reliability of healthcare predictive models. It outperformed other artificial intelligence methods, achieving precision, accuracy, recall, F1-score, and MCC scores of 95.57%, 96.90%, 99.70%, 97.59%, and 93.42%, respectively.
Reza Bayat Rizi, Amir R. Forouzan, Farshad Miramirkhani, Mohamad F. Sabahi,
Volume 20, Issue 4 (11-2024)
Abstract
Visible Light Communication, a key optical wireless technology, offers reliable, high-bandwidth, and secure communication, making it a promising soloution for a variety of applications. Despite its many advantages, optical wireless communication faces challenges in medical environments due to fluctuating signal strength caused by patient movement. Smart transmitter structures can improve system performance by adjusting system parameters to the fluctuating channel conditions. The purpose of this research is to examine how adaptive modulation performs in a medical body sensor network system that uses visible light communication. The analysis focuses on various medical situations and investigates machine learning algorithms. The study compares adaptive modulation based on supervised learning with that based on reinforcement learning. The findings indicate that both approaches greatly improve spectral efficiency, emphasizing the significance of implementing link adaptation in visible light communication-based medical body sensor networks. The use of the Q-learning algorithm in adaptive modulation enables real-time training and enables the system to adjust to the changing environment without any prior knowledge about the environment. A remarkable improvement is observed for photodetectors on the shoulder and wrist since they experience more DC gain.
Robinson Jimenez-Moreno, Anny Astrid Espitia Cubillos, Esperanza Rodríguez Carmona,
Volume 20, Issue 4 (11-2024)
Abstract
This document presents the design of a virtual robotic system for the supervision of physical training exercises, to be carried out in a closed environment, which only requires a computer equipment with a web camera. To do this, deep learning algorithms such as convolutional networks and short- and long-term memory networks are used to recognize voice commands and the user's video actions. A predefined dialogue template is used to guide a user's training cycle based on the execution of the exercises: push-ups, abdominal, jump or squat. The contribution of the work focuses on the integration of deep learning techniques to design and personalize virtual robotic assistants for everyday task. The results show a high level of accuracy by the virtual robot both in understanding the audio and in predicting the exercise to be performed, with a final accuracy value of 97.75% and 100%, respectively.
