This document compares self-organizing feature maps (SOFM) with k-means clustering and artificial neural networks for pattern recognition and feature map creation. SOFM uses competitive learning to organize input vectors into clusters without supervision, mapping similar inputs close together on the map. K-means aims to partition inputs into a predefined number of clusters by minimizing within-cluster variation. Artificial neural networks implement classification in three phases - self-organizing feature map learning, followed by supervised learning phases. The document discusses algorithms, architectures, and training approaches for each method.
Neural networks Self Organizing Map by Engr. Edgar Carrillo IIEdgar Carrillo
This presentation talks about neural networks and self organizing maps. In this presentation,Engr. Edgar Caburatan Carrillo II also discusses its applications.
DESIGN AND IMPLEMENTATION OF BINARY NEURAL NETWORK LEARNING WITH FUZZY CLUSTE...cscpconf
In this paper, Design and Implementation of Binary Neural Network Learning with Fuzzy
Clustering (DIBNNFC), is proposed to classify semisupervised data, it is based on the
concept of binary neural network and geometrical expansion. Parameters are updated
according to the geometrical location of the training samples in the input space, and each
sample in the training set is learned only once. It’s a semisupervised based approach, the
training samples are semi-labelled i.e. for some samples, labels are known and for some
samples data labels are not known. The method starts with classification, which is done by
using the concept of ETL algorithm. In classification process various classes are formed.
These classes classify samples in to two classes after that considers each class as a region and calculates the average of the entire region separately. This average is centres of the region which is used for the purpose of clustering by using FCM algorithm. Once clustering process over labelling of semi supervised data is done, then whole samples would be classify by (DIBNNFC). The method proposes here is exhaustively tested with different benchmark datasets and it is found that, on increasing value of training parameters number of hidden neurons and training time both are getting decrease. The result reported, using real character recognition data set and result will compare with existing semi-supervised classifier, the proposed approach learned with semi-supervised leads to higher classification accuracy.
Self Organizing Maps: Fundamentals.
Introduction to Neural Networks.
1. What is a Self Organizing Map?
2. Topographic Maps
3. Setting up a Self Organizing Map
4. Kohonen Networks
5. Components of Self Organization
6. Overview of the SOM Algorithm
Neural networks Self Organizing Map by Engr. Edgar Carrillo IIEdgar Carrillo
This presentation talks about neural networks and self organizing maps. In this presentation,Engr. Edgar Caburatan Carrillo II also discusses its applications.
DESIGN AND IMPLEMENTATION OF BINARY NEURAL NETWORK LEARNING WITH FUZZY CLUSTE...cscpconf
In this paper, Design and Implementation of Binary Neural Network Learning with Fuzzy
Clustering (DIBNNFC), is proposed to classify semisupervised data, it is based on the
concept of binary neural network and geometrical expansion. Parameters are updated
according to the geometrical location of the training samples in the input space, and each
sample in the training set is learned only once. It’s a semisupervised based approach, the
training samples are semi-labelled i.e. for some samples, labels are known and for some
samples data labels are not known. The method starts with classification, which is done by
using the concept of ETL algorithm. In classification process various classes are formed.
These classes classify samples in to two classes after that considers each class as a region and calculates the average of the entire region separately. This average is centres of the region which is used for the purpose of clustering by using FCM algorithm. Once clustering process over labelling of semi supervised data is done, then whole samples would be classify by (DIBNNFC). The method proposes here is exhaustively tested with different benchmark datasets and it is found that, on increasing value of training parameters number of hidden neurons and training time both are getting decrease. The result reported, using real character recognition data set and result will compare with existing semi-supervised classifier, the proposed approach learned with semi-supervised leads to higher classification accuracy.
Self Organizing Maps: Fundamentals.
Introduction to Neural Networks.
1. What is a Self Organizing Map?
2. Topographic Maps
3. Setting up a Self Organizing Map
4. Kohonen Networks
5. Components of Self Organization
6. Overview of the SOM Algorithm
Here is my class on the multilayer perceptron where I look at the following:
1.- The entire backproagation algorithm based in the gradient descent
However, I am planning the tanning based in Kalman filters.
2.- The use of matrix computations to simplify the implementations.
I hope you enjoy it.
Fixed-Point Code Synthesis for Neural Networksgerogepatton
Over the last few years, neural networks have started penetrating safety critical systems to take decisions in robots, rockets, autonomous driving car, etc. A problem is that these critical systems often have limited computing resources. Often, they use the fixed-point arithmetic for its many advantages (rapidity, compatibility with small memory devices.) In this article, a new technique is introduced to tune the formats (precision) of already trained neural networks using fixed-point arithmetic, which can be implemented using integer operations only. The new optimized neural network computes the output with fixed-point numbers without modifying the accuracy up to a threshold fixed by the user. A fixed-point code is synthesized for the new optimized neural network ensuring the respect of the threshold for any input vector belonging the range [xmin, xmax] determined during the analysis. From a technical point of view, we do a preliminary analysis of our floating neural network to determine the worst cases, then we generate a system of linear constraints among integer variables that we can solve by linear programming. The solution of this system is the new fixed-point format of each neuron. The experimental results obtained show the efficiency of our method which can ensure that the new fixed-point neural network has the same behavior as the initial floating-point neural network.
Radial basis function network ppt bySheetal,Samreen and Dhanashrisheetal katkar
Radial Basis Functions are nonlinear activation functions used by artificial neural networks.Explained commonly used RBFs ,cover's theorem,interpolation problem and learning strategies.
Classification by back propagation, multi layered feed forward neural network...bihira aggrey
Classification by Back Propagation, Multi-layered feed forward Neural Networks - Provides a basic introduction of classification in data mining with neural networks
Here is my class on the multilayer perceptron where I look at the following:
1.- The entire backproagation algorithm based in the gradient descent
However, I am planning the tanning based in Kalman filters.
2.- The use of matrix computations to simplify the implementations.
I hope you enjoy it.
Fixed-Point Code Synthesis for Neural Networksgerogepatton
Over the last few years, neural networks have started penetrating safety critical systems to take decisions in robots, rockets, autonomous driving car, etc. A problem is that these critical systems often have limited computing resources. Often, they use the fixed-point arithmetic for its many advantages (rapidity, compatibility with small memory devices.) In this article, a new technique is introduced to tune the formats (precision) of already trained neural networks using fixed-point arithmetic, which can be implemented using integer operations only. The new optimized neural network computes the output with fixed-point numbers without modifying the accuracy up to a threshold fixed by the user. A fixed-point code is synthesized for the new optimized neural network ensuring the respect of the threshold for any input vector belonging the range [xmin, xmax] determined during the analysis. From a technical point of view, we do a preliminary analysis of our floating neural network to determine the worst cases, then we generate a system of linear constraints among integer variables that we can solve by linear programming. The solution of this system is the new fixed-point format of each neuron. The experimental results obtained show the efficiency of our method which can ensure that the new fixed-point neural network has the same behavior as the initial floating-point neural network.
Radial basis function network ppt bySheetal,Samreen and Dhanashrisheetal katkar
Radial Basis Functions are nonlinear activation functions used by artificial neural networks.Explained commonly used RBFs ,cover's theorem,interpolation problem and learning strategies.
Classification by back propagation, multi layered feed forward neural network...bihira aggrey
Classification by Back Propagation, Multi-layered feed forward Neural Networks - Provides a basic introduction of classification in data mining with neural networks
Paths of Wellbeing on Self-Organizing Maps + excerpts from other presentationsTimo Honkela
In this presentation in WSOM 2012 conference, we introduce the concept of pathways of wellbeing
and examine how such paths can be discovered from large data
sets using the self-organizing map. Data sets used in the illustrative experiments
include measurements of physical fitness and subjective assessments
related to diagnosing work stress. In addition, we show results from related projects.
Intrusion Detection System using Self Organizing Map: A SurveyIJERA Editor
Due to usage of computer every field, Network Security is the major concerned in today’s scenario. Every year the number of users and speed of network is increasing, along with it online fraud or security threats are also increasing. Every day a new attack is generated to harm the system or network. It is necessary to protect the system or networks from various threats by using Intrusion Detection System which can detect “known” as well as “unknown” attack and generate alerts if any unusual behavior in the traffic. There are various approaches for IDS, but in this paper, survey is focused on IDS using Self Organizing Map. SOM is unsupervised, fast conversion and automatic clustering algorithm which is able to handle novelty detection. The main objective of the survey is to find and address the current challenges of SOM. Our survey shows that the existing IDS based on SOM have poor detection rate for U2R and R2L attacks. To improve it, proper normalization technique should be used. During the survey we also found that HSOM and GHSOM are advance model of SOM which have their own unique feature for better performance of IDS. GHSOM is efficient due to its low computation time. This survey is beneficial to design and develop efficient SOM based IDS having less computation time and better detection rate.
SOM (Self-Organizing Map) is one of the most popular artificial neural network algorithms in the unsupervised learning category. For efficient construction of large maps searching the best-matching unit is usually the computationally heaviest operation in the SOM. The parallel nature of the algorithm and the huge computations involved makes it a good target for GPU based parallel implementation. This paper presents an overall idea of the optimization strategies used for the parallel implementation of Basic-SOM on GPU using CUDA programming paradigm.
Food Recommendation System Using Clustering Analysis for Diabetic patientsMaiyaporn Phanich
Food and nutrition are a key to have good health. They are important for everyone to maintain a healthy diet, especially for diabetic patients who have several limitations. Nutrition therapy is a major solution to prevent, manage and control diabetes by managing the nutrition based on the belief that food provides vital medicine and maintains a good health. Typically, diabetic patients need to avoid additional sugar and fat, so the food pyramid is recommended to the patients for finding the substitution from the same food group. However, there is still a dietary diversity within food groups that can affect the diabetic patients. In this study, we proposed Food Recommendation System (FRS) by using food clustering analysis for diabetic patients. Our system will recommend the proper substituted foods in the context of nutrition and food characteristic. We used Self-Organizing Map (SOM) and K-mean clustering for food clustering analysis, which is based on the similarity of eight significant nutrients for diabetic patient. At the end, the FRS was evaluated by nutritionists and it has performed very well and useful for nutrition area.
KohonAnts: A Self-Organizing Ant Algorithm for Clustering and Pattern Classif...Antonio Mora
You can try it at:
http://geneura.ugr.es/~amorag/kohonants/KAnts.zip
This is the presentation of KohonAnts (KANTS), an hybrid Ant Colony and Self-organizing Map algorithm for clustering and pattern classification.
Esta es la presentación de KohonAnts (KANTS), un algoritmo que combina conceptos de los algortimos de hormigas y los mapas autoorganizativos y que se puede utilizar para agrupamiento (clustering) o clasificación de patrones.
Mobile Network Coverage Determination at 900MHz for Abuja Rural Areas using A...ijtsrd
This study proposes Artificial Neural Network ANN based field strength prediction models for the rural areas of Abuja, the federal capital territory of Nigeria. The ANN based models were created on bases of the Generalized Regression Neural network GRNN and the Multi Layer Perceptron Neural Network MLP NN . These networks were created, trained and tested for field strength prediction using received power data recorded at 900MHz from multiple Base Transceiver Stations BTSs distributed across the rural areas. Results indicate that the GRNN and MLP NN based models with Root Mean Squared Error RMSE values of 4.78dBm and 5.56dBm respectively, offer significant improvement over the empirical Hata Okumura counterpart, which overestimates the signal strength by an RMSE value of 20.17dBm. Deme C. Abraham ""Mobile Network Coverage Determination at 900MHz for Abuja Rural Areas using Artificial Neural Networks"" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-4 | Issue-2 , February 2020,
URL: https://www.ijtsrd.com/papers/ijtsrd30228.pdf
Paper Url : https://www.ijtsrd.com/computer-science/artificial-intelligence/30228/mobile-network-coverage-determination-at-900mhz-for-abuja-rural-areas-using-artificial-neural-networks/deme-c-abraham
A New Classifier Based onRecurrent Neural Network Using Multiple Binary-Outpu...iosrjce
IOSR Journal of Computer Engineering (IOSR-JCE) is a double blind peer reviewed International Journal that provides rapid publication (within a month) of articles in all areas of computer engineering and its applications. The journal welcomes publications of high quality papers on theoretical developments and practical applications in computer technology. Original research papers, state-of-the-art reviews, and high quality technical notes are invited for publications.
ARTIFICIAL NEURAL NETWORK APPROACH TO MODELING OF POLYPROPYLENE REACTORijac123
This paper shows modeling of highly nonlinear polymerization process using the artificial neural network approach for the model predictive purposes. Polymerization occurs in a fluidized bed polypropylene reactor using Ziegler - Natta catalyst and the main objective was modeling of the reactor production rate.
The data set used for an identification of the model is a real process data received from an existing polypropylene plant and the identified model is a nonlinear autoregressive neural network with the exogenous input. Performance of a trained network has been verified using the real process data and the
ability of the production rate prediction is shown in the conclusion.
Web spam classification using supervised artificial neural network algorithmsaciijournal
Due to the rapid growth in technology employed by the spammers, there is a need of classifiers that are more efficient, generic and highly adaptive. Neural Network based technologies have high ability of adaption as well as generalization. As per our knowledge, very little work has been done in this field using neural network. We present this paper to fill this gap. This paper evaluates performance of three supervised learning algorithms of artificial neural network by creating classifiers for the complex problem of latest web spam pattern classification. These algorithms are Conjugate Gradient algorithm, Resilient Backpropagation learning, and Levenberg-Marquardt algorithm.
A STUDY OF METHODS FOR TRAINING WITH DIFFERENT DATASETS IN IMAGE CLASSIFICATIONADEIJ Journal
This research developed a training method of Convolutional Neural Network model with multiple datasets to achieve good performance on both datasets. Two different methods of training with two characteristically different datasets with identical categories, one with very clean images and one with real-world data, were proposed and studied. The model used for the study was a neural network derived from ResNet. Mixed training was shown to produce the best accuracies for each dataset when the dataset is mixed into the training set at the highest proportion, and the best combined performance when the realworld dataset was mixed in at a ratio of around 70%. This ratio produced a top-1 combined performance of 63.8% (no mixing produced 30.8%) and a top-3 combined performance of 83.0% (no mixing produced 55.3%). This research also showed that iterative training has a worse combined performance than mixed training due to the issue of fast forgetting.
Web Spam Classification Using Supervised Artificial Neural Network Algorithmsaciijournal
Due to the rapid growth in technology employed by the spammers, there is a need of classifiers that are
more efficient, generic and highly adaptive. Neural Network based technologies have high ability of
adaption as well as generalization. As per our knowledge, very little work has been done in this field using
neural network. We present this paper to fill this gap. This paper evaluates performance of three supervised
learning algorithms of artificial neural network by creating classifiers for the complex problem of latest
web spam pattern classification. These algorithms are Conjugate Gradient algorithm, Resilient Backpropagation learning, and Levenberg-Marquardt algorithm.
Optimization of Number of Neurons in the Hidden Layer in Feed Forward Neural ...IJERA Editor
The architectures of Artificial Neural Networks (ANN) are based on the problem domain and it is applied during
the „training phase‟ of sample data and used to infer results for the remaining data in the testing phase.
Normally, the architecture consist of three layers as input, hidden, output layers with the number of nodes in the
input layer as number of known values on hand and the number of nodes as result to be computed out of the
values of input nodes and hidden nodes as the output layer. The number of nodes in the hidden layer is
heuristically decided so that the optimum value is obtained with reasonable number of iterations with other
parameters with its default values. This study mainly focuses on Cascade-Correlation Neural Networks (CCNN)
using Back-Propagation (BP) algorithm which finds the number of neurons during the training phase itself by
appending one from the previous iteration satisfying the error condition gives a promising result on the optimum
number of neurons in the hidden layer
Soft computing based cryptographic technique using kohonen's selforganizing m...ijfcstjournal
In this paper a novel soft computing based cryptographic technique based on synchronization of two
Kohonen's Self-Organizing Feature Map (KSOFM) between sender and receiver has been proposed. In this
proposed technique KSOFM based synchronization is performed for tuning both sender and receiver. After
the completion of the tuning identical session key get generates at the both sender and receiver end with the
help of synchronized KSOFM. This synchronized network can be used for transmitting message using any
light weight encryption/decryption algorithm with the help of identical session key of the synchronized
network. Exhaustive parametric tests are done and results are compared with some existing classical
techniques, which show comparable results for the proposed system.
ExcelR is a proud partner of Universiti Malaysia Saravak (UNIMAS), Malaysia’s 1st public University and ranked 8th top university in Malaysia and ranked among top 200th in Asian University Rankings 2017 by QS World University Rankings. Participants will be awarded Data Science international certification from UNIMAS, after succesfully clearning the online examination
ExcelR provides the best Data Science Training in Kolkata covering its entire scope and much more.You will be given access to all the training material for lifetime with LMS access which you can access anytime and anywhere.Post training assistance will also be provided.Enroll Now!
ExcelR is the fastest growing company providing Python Data science course. We provide you the best faculty for training and after completing of training ExcelR will provide you certificate from Malaysian University. If you are searching for Python Data science course your searching ends with ExcelR.
Looking for a great career in Data Science Training? Here is the answer to your question related to Data Science.
Excelr is providing core Data Science Training in Pune. Learn with experts with extensive corporate training experience. Faculty of best in the business will help you make market leaders in data science. Real time work and task will be provided for better results.
Please do go through the link below for all your queries regarding the course/ training.
https://www.excelr.com/business-analytics-training-in-pune/
Similar to International Journal of Engineering Research and Development (IJERD) (20)
A Novel Method for Prevention of Bandwidth Distributed Denial of Service AttacksIJERD Editor
Distributed Denial of Service (DDoS) Attacks became a massive threat to the Internet. Traditional
Architecture of internet is vulnerable to the attacks like DDoS. Attacker primarily acquire his army of Zombies,
then that army will be instructed by the Attacker that when to start an attack and on whom the attack should be
done. In this paper, different techniques which are used to perform DDoS Attacks, Tools that were used to
perform Attacks and Countermeasures in order to detect the attackers and eliminate the Bandwidth Distributed
Denial of Service attacks (B-DDoS) are reviewed. DDoS Attacks were done by using various Flooding
techniques which are used in DDoS attack.
The main purpose of this paper is to design an architecture which can reduce the Bandwidth
Distributed Denial of service Attack and make the victim site or server available for the normal users by
eliminating the zombie machines. Our Primary focus of this paper is to dispute how normal machines are
turning into zombies (Bots), how attack is been initiated, DDoS attack procedure and how an organization can
save their server from being a DDoS victim. In order to present this we implemented a simulated environment
with Cisco switches, Routers, Firewall, some virtual machines and some Attack tools to display a real DDoS
attack. By using Time scheduling, Resource Limiting, System log, Access Control List and some Modular
policy Framework we stopped the attack and identified the Attacker (Bot) machines
Hearing loss is one of the most common human impairments. It is estimated that by year 2015 more
than 700 million people will suffer mild deafness. Most can be helped by hearing aid devices depending on the
severity of their hearing loss. This paper describes the implementation and characterization details of a dual
channel transmitter front end (TFE) for digital hearing aid (DHA) applications that use novel micro
electromechanical- systems (MEMS) audio transducers and ultra-low power-scalable analog-to-digital
converters (ADCs), which enable a very-low form factor, energy-efficient implementation for next-generation
DHA. The contribution of the design is the implementation of the dual channel MEMS microphones and powerscalable
ADC system.
Influence of tensile behaviour of slab on the structural Behaviour of shear c...IJERD Editor
-A composite beam is composed of a steel beam and a slab connected by means of shear connectors
like studs installed on the top flange of the steel beam to form a structure behaving monolithically. This study
analyzes the effects of the tensile behavior of the slab on the structural behavior of the shear connection like slip
stiffness and maximum shear force in composite beams subjected to hogging moment. The results show that the
shear studs located in the crack-concentration zones due to large hogging moments sustain significantly smaller
shear force and slip stiffness than the other zones. Moreover, the reduction of the slip stiffness in the shear
connection appears also to be closely related to the change in the tensile strain of rebar according to the increase
of the load. Further experimental and analytical studies shall be conducted considering variables such as the
reinforcement ratio and the arrangement of shear connectors to achieve efficient design of the shear connection
in composite beams subjected to hogging moment.
Gold prospecting using Remote Sensing ‘A case study of Sudan’IJERD Editor
Gold has been extracted from northeast Africa for more than 5000 years, and this may be the first
place where the metal was extracted. The Arabian-Nubian Shield (ANS) is an exposure of Precambrian
crystalline rocks on the flanks of the Red Sea. The crystalline rocks are mostly Neoproterozoic in age. ANS
includes the nations of Israel, Jordan. Egypt, Saudi Arabia, Sudan, Eritrea, Ethiopia, Yemen, and Somalia.
Arabian Nubian Shield Consists of juvenile continental crest that formed between 900 550 Ma, when intra
oceanic arc welded together along ophiolite decorated arc. Primary Au mineralization probably developed in
association with the growth of intra oceanic arc and evolution of back arc. Multiple episodes of deformation
have obscured the primary metallogenic setting, but at least some of the deposits preserve evidence that they
originate as sea floor massive sulphide deposits.
The Red Sea Hills Region is a vast span of rugged, harsh and inhospitable sector of the Earth with
inimical moon-like terrain, nevertheless since ancient times it is famed to be an abode of gold and was a major
source of wealth for the Pharaohs of ancient Egypt. The Pharaohs old workings have been periodically
rediscovered through time. Recent endeavours by the Geological Research Authority of Sudan led to the
discovery of a score of occurrences with gold and massive sulphide mineralizations. In the nineties of the
previous century the Geological Research Authority of Sudan (GRAS) in cooperation with BRGM utilized
satellite data of Landsat TM using spectral ratio technique to map possible mineralized zones in the Red Sea
Hills of Sudan. The outcome of the study mapped a gossan type gold mineralization. Band ratio technique was
applied to Arbaat area and a signature of alteration zone was detected. The alteration zones are commonly
associated with mineralization. The alteration zones are commonly associated with mineralization. A filed check
confirmed the existence of stock work of gold bearing quartz in the alteration zone. Another type of gold
mineralization that was discovered using remote sensing is the gold associated with metachert in the Atmur
Desert.
Reducing Corrosion Rate by Welding DesignIJERD Editor
The paper addresses the importance of welding design to prevent corrosion at steel. Welding is
used to join pipe, profiles at bridges, spindle, and a lot more part of engineering construction. The
problems happened associated with welding are common issues in these fields, especially corrosion.
Corrosion can be reduced with many methods, they are painting, controlling humidity, and also good
welding design. In the research, it can be found that reducing residual stress on the welding can be
solved in corrosion rate reduction problem.
Preheating on 500oC and 600oC give better condition to reduce corosion rate than condition after
preheating 400oC. For all welding groove type, material with 500oC and 600oC preheating after 14 days
corrosion test is 0,5%-0,69% lost. Material with 400oC preheating after 14 days corrosion test is 0,57%-0,76%
lost.
Welding groove also influence corrosion rate. X and V type welding groove give better condition to reduce
corrosion rate than use 1/2V and 1/2 X welding groove. After 14 days corrosion test, the samples with
X welding groove type is 0,5%-0,57% lost. The samples with V welding groove after 14 days corrosion test is
0,51%-0,59% lost. The samples with 1/2V and 1/2X welding groove after 14 days corrosion test is 0,58%-
0,71% lost.
Router 1X3 – RTL Design and VerificationIJERD Editor
Routing is the process of moving a packet of data from source to destination and enables messages
to pass from one computer to another and eventually reach the target machine. A router is a networking device
that forwards data packets between computer networks. It is connected to two or more data lines from different
networks (as opposed to a network switch, which connects data lines from one single network). This paper,
mainly emphasizes upon the study of router device, it‟s top level architecture, and how various sub-modules of
router i.e. Register, FIFO, FSM and Synchronizer are synthesized, and simulated and finally connected to its top
module.
Active Power Exchange in Distributed Power-Flow Controller (DPFC) At Third Ha...IJERD Editor
This paper presents a component within the flexible ac-transmission system (FACTS) family, called
distributed power-flow controller (DPFC). The DPFC is derived from the unified power-flow controller (UPFC)
with an eliminated common dc link. The DPFC has the same control capabilities as the UPFC, which comprise
the adjustment of the line impedance, the transmission angle, and the bus voltage. The active power exchange
between the shunt and series converters, which is through the common dc link in the UPFC, is now through the
transmission lines at the third-harmonic frequency. DPFC multiple small-size single-phase converters which
reduces the cost of equipment, no voltage isolation between phases, increases redundancy and there by
reliability increases. The principle and analysis of the DPFC are presented in this paper and the corresponding
simulation results that are carried out on a scaled prototype are also shown.
Mitigation of Voltage Sag/Swell with Fuzzy Control Reduced Rating DVRIJERD Editor
Power quality has been an issue that is becoming increasingly pivotal in industrial electricity
consumers point of view in recent times. Modern industries employ Sensitive power electronic equipments,
control devices and non-linear loads as part of automated processes to increase energy efficiency and
productivity. Voltage disturbances are the most common power quality problem due to this the use of a large
numbers of sophisticated and sensitive electronic equipment in industrial systems is increased. This paper
discusses the design and simulation of dynamic voltage restorer for improvement of power quality and
reduce the harmonics distortion of sensitive loads. Power quality problem is occurring at non-standard
voltage, current and frequency. Electronic devices are very sensitive loads. In power system voltage sag,
swell, flicker and harmonics are some of the problem to the sensitive load. The compensation capability
of a DVR depends primarily on the maximum voltage injection ability and the amount of stored
energy available within the restorer. This device is connected in series with the distribution feeder at
medium voltage. A fuzzy logic control is used to produce the gate pulses for control circuit of DVR and the
circuit is simulated by using MATLAB/SIMULINK software.
Study on the Fused Deposition Modelling In Additive ManufacturingIJERD Editor
Additive manufacturing process, also popularly known as 3-D printing, is a process where a product
is created in a succession of layers. It is based on a novel materials incremental manufacturing philosophy.
Unlike conventional manufacturing processes where material is removed from a given work price to derive the
final shape of a product, 3-D printing develops the product from scratch thus obviating the necessity to cut away
materials. This prevents wastage of raw materials. Commonly used raw materials for the process are ABS
plastic, PLA and nylon. Recently the use of gold, bronze and wood has also been implemented. The complexity
factor of this process is 0% as in any object of any shape and size can be manufactured.
Spyware triggering system by particular string valueIJERD Editor
This computer programme can be used for good and bad purpose in hacking or in any general
purpose. We can say it is next step for hacking techniques such as keylogger and spyware. Once in this system if
user or hacker store particular string as a input after that software continually compare typing activity of user
with that stored string and if it is match then launch spyware programme.
A Blind Steganalysis on JPEG Gray Level Image Based on Statistical Features a...IJERD Editor
This paper presents a blind steganalysis technique to effectively attack the JPEG steganographic
schemes i.e. Jsteg, F5, Outguess and DWT Based. The proposed method exploits the correlations between
block-DCTcoefficients from intra-block and inter-block relation and the statistical moments of characteristic
functions of the test image is selected as features. The features are extracted from the BDCT JPEG 2-array.
Support Vector Machine with cross-validation is implemented for the classification.The proposed scheme gives
improved outcome in attacking.
Secure Image Transmission for Cloud Storage System Using Hybrid SchemeIJERD Editor
- Data over the cloud is transferred or transmitted between servers and users. Privacy of that
data is very important as it belongs to personal information. If data get hacked by the hacker, can be
used to defame a person’s social data. Sometimes delay are held during data transmission. i.e. Mobile
communication, bandwidth is low. Hence compression algorithms are proposed for fast and efficient
transmission, encryption is used for security purposes and blurring is used by providing additional
layers of security. These algorithms are hybridized for having a robust and efficient security and
transmission over cloud storage system.
Application of Buckley-Leverett Equation in Modeling the Radius of Invasion i...IJERD Editor
A thorough review of existing literature indicates that the Buckley-Leverett equation only analyzes
waterflood practices directly without any adjustments on real reservoir scenarios. By doing so, quite a number
of errors are introduced into these analyses. Also, for most waterflood scenarios, a radial investigation is more
appropriate than a simplified linear system. This study investigates the adoption of the Buckley-Leverett
equation to estimate the radius invasion of the displacing fluid during waterflooding. The model is also adopted
for a Microbial flood and a comparative analysis is conducted for both waterflooding and microbial flooding.
Results shown from the analysis doesn’t only records a success in determining the radial distance of the leading
edge of water during the flooding process, but also gives a clearer understanding of the applicability of
microbes to enhance oil production through in-situ production of bio-products like bio surfactans, biogenic
gases, bio acids etc.
Gesture Gaming on the World Wide Web Using an Ordinary Web CameraIJERD Editor
- Gesture gaming is a method by which users having a laptop/pc/x-box play games using natural or
bodily gestures. This paper presents a way of playing free flash games on the internet using an ordinary webcam
with the help of open source technologies. Emphasis in human activity recognition is given on the pose
estimation and the consistency in the pose of the player. These are estimated with the help of an ordinary web
camera having different resolutions from VGA to 20mps. Our work involved giving a 10 second documentary to
the user on how to play a particular game using gestures and what are the various kinds of gestures that can be
performed in front of the system. The initial inputs of the RGB values for the gesture component is obtained by
instructing the user to place his component in a red box in about 10 seconds after the short documentary before
the game is finished. Later the system opens the concerned game on the internet on popular flash game sites like
miniclip, games arcade, GameStop etc and loads the game clicking at various places and brings the state to a
place where the user is to perform only gestures to start playing the game. At any point of time the user can call
off the game by hitting the esc key and the program will release all of the controls and return to the desktop. It
was noted that the results obtained using an ordinary webcam matched that of the Kinect and the users could
relive the gaming experience of the free flash games on the net. Therefore effective in game advertising could
also be achieved thus resulting in a disruptive growth to the advertising firms.
Hardware Analysis of Resonant Frequency Converter Using Isolated Circuits And...IJERD Editor
-LLC resonant frequency converter is basically a combo of series as well as parallel resonant ckt. For
LCC resonant converter it is associated with a disadvantage that, though it has two resonant frequencies, the
lower resonant frequency is in ZCS region[5]. For this application, we are not able to design the converter
working at this resonant frequency. LLC resonant converter existed for a very long time but because of
unknown characteristic of this converter it was used as a series resonant converter with basically a passive
(resistive) load. . Here, it was designed to operate in switching frequency higher than resonant frequency of the
series resonant tank of Lr and Cr converter acts very similar to Series Resonant Converter. The benefit of LLC
resonant converter is narrow switching frequency range with light load[6] . Basically, the control ckt plays a
very imp. role and hence 555 Timer used here provides a perfect square wave as the control ckt provides no
slew rate which makes the square wave really strong and impenetrable. The dead band circuit provides the
exclusive dead band in micro seconds so as to avoid the simultaneous firing of two pairs of IGBT’s where one
pair switches off and the other on for a slightest period of time. Hence, the isolator ckt here is associated with
each and every ckt used because it acts as a driver and an isolation to each of the IGBT is provided with one
exclusive transformer supply[3]. The IGBT’s are fired using the appropriate signal using the previous boards
and hence at last a high frequency rectifier ckt with a filtering capacitor is used to get an exact dc
waveform .The basic goal of this particular analysis is to observe the wave forms and characteristics of
converters with differently positioned passive elements in the form of tank circuits.
Simulated Analysis of Resonant Frequency Converter Using Different Tank Circu...IJERD Editor
LLC resonant frequency converter is basically a combo of series as well as parallel resonant ckt. For
LCC resonant converter it is associated with a disadvantage that, though it has two resonant frequencies, the
lower resonant frequency is in ZCS region [5]. For this application, we are not able to design the converter
working at this resonant frequency. LLC resonant converter existed for a very long time but because of
unknown characteristic of this converter it was used as a series resonant converter with basically a passive
(resistive) load. . Here, it was designed to operate in switching frequency higher than resonant frequency of the
series resonant tank of Lr and Cr converter acts very similar to Series Resonant Converter. The benefit of LLC
resonant converter is narrow switching frequency range with light load[6] . Basically, the control ckt plays a
very imp. role and hence 555 Timer used here provides a perfect square wave as the control ckt provides no
slew rate which makes the square wave really strong and impenetrable. The dead band circuit provides the
exclusive dead band in micro seconds so as to avoid the simultaneous firing of two pairs of IGBT’s where one
pair switches off and the other on for a slightest period of time. Hence, the isolator ckt here is associated with
each and every ckt used because it acts as a driver and an isolation to each of the IGBT is provided with one
exclusive transformer supply[3]. The IGBT’s are fired using the appropriate signal using the previous boards
and hence at last a high frequency rectifier ckt with a filtering capacitor is used to get an exact dc
waveform .The basic goal of this particular analysis is to observe the wave forms and characteristics of
converters with differently positioned passive elements in the form of tank circuits. The supported simulation
is done through PSIM 6.0 software tool
Amateurs Radio operator, also known as HAM communicates with other HAMs through Radio
waves. Wireless communication in which Moon is used as natural satellite is called Moon-bounce or EME
(Earth -Moon-Earth) technique. Long distance communication (DXing) using Very High Frequency (VHF)
operated amateur HAM radio was difficult. Even with the modest setup having good transceiver, power
amplifier and high gain antenna with high directivity, VHF DXing is possible. Generally 2X11 YAGI antenna
along with rotor to set horizontal and vertical angle is used. Moon tracking software gives exact location,
visibility of Moon at both the stations and other vital data to acquire real time position of moon.
“MS-Extractor: An Innovative Approach to Extract Microsatellites on „Y‟ Chrom...IJERD Editor
Simple Sequence Repeats (SSR), also known as Microsatellites, have been extensively used as
molecular markers due to their abundance and high degree of polymorphism. The nucleotide sequences of
polymorphic forms of the same gene should be 99.9% identical. So, Microsatellites extraction from the Gene is
crucial. However, Microsatellites repeat count is compared, if they differ largely, he has some disorder. The Y
chromosome likely contains 50 to 60 genes that provide instructions for making proteins. Because only males
have the Y chromosome, the genes on this chromosome tend to be involved in male sex determination and
development. Several Microsatellite Extractors exist and they fail to extract microsatellites on large data sets of
giga bytes and tera bytes in size. The proposed tool “MS-Extractor: An Innovative Approach to extract
Microsatellites on „Y‟ Chromosome” can extract both Perfect as well as Imperfect Microsatellites from large
data sets of human genome „Y‟. The proposed system uses string matching with sliding window approach to
locate Microsatellites and extracts them.
Importance of Measurements in Smart GridIJERD Editor
- The need to get reliable supply, independence from fossil fuels, and capability to provide clean
energy at a fixed and lower cost, the existing power grid structure is transforming into Smart Grid. The
development of a smart energy distribution grid is a current goal of many nations. A Smart Grid should have
new capabilities such as self-healing, high reliability, energy management, and real-time pricing. This new era
of smart future grid will lead to major changes in existing technologies at generation, transmission and
distribution levels. The incorporation of renewable energy resources and distribution generators in the existing
grid will increase the complexity, optimization problems and instability of the system. This will lead to a
paradigm shift in the instrumentation and control requirements for Smart Grids for high quality, stable and
reliable electricity supply of power. The monitoring of the grid system state and stability relies on the
availability of reliable measurement of data. In this paper the measurement areas that highlight new
measurement challenges, development of the Smart Meters and the critical parameters of electric energy to be
monitored for improving the reliability of power systems has been discussed.
Study of Macro level Properties of SCC using GGBS and Lime stone powderIJERD Editor
One of the major environmental concerns is the disposal of the waste materials and utilization of
industrial by products. Lime stone quarries will produce millions of tons waste dust powder every year. Having
considerable high degree of fineness in comparision to cement this material may be utilized as a partial
replacement to cement. For this purpose an experiment is conducted to investigate the possibility of using lime
stone powder in the production of SCC with combined use GGBS and how it affects the fresh and mechanical
properties of SCC. First SCC is made by replacing cement with GGBS in percentages like 10, 20, 30, 40, 50 and
by taking the optimum mix with GGBS lime stone powder is blended to mix in percentages like 5, 10, 15, 20 as
a partial replacement to cement. Test results shows that the SCC mix with combination of 30% GGBS and 15%
limestone powder gives maximum compressive strength and fresh properties are also in the limits prescribed by
the EFNARC.
A tale of scale & speed: How the US Navy is enabling software delivery from l...sonjaschweigert1
Rapid and secure feature delivery is a goal across every application team and every branch of the DoD. The Navy’s DevSecOps platform, Party Barge, has achieved:
- Reduction in onboarding time from 5 weeks to 1 day
- Improved developer experience and productivity through actionable findings and reduction of false positives
- Maintenance of superior security standards and inherent policy enforcement with Authorization to Operate (ATO)
Development teams can ship efficiently and ensure applications are cyber ready for Navy Authorizing Officials (AOs). In this webinar, Sigma Defense and Anchore will give attendees a look behind the scenes and demo secure pipeline automation and security artifacts that speed up application ATO and time to production.
We will cover:
- How to remove silos in DevSecOps
- How to build efficient development pipeline roles and component templates
- How to deliver security artifacts that matter for ATO’s (SBOMs, vulnerability reports, and policy evidence)
- How to streamline operations with automated policy checks on container images
Key Trends Shaping the Future of Infrastructure.pdfCheryl Hung
Keynote at DIGIT West Expo, Glasgow on 29 May 2024.
Cheryl Hung, ochery.com
Sr Director, Infrastructure Ecosystem, Arm.
The key trends across hardware, cloud and open-source; exploring how these areas are likely to mature and develop over the short and long-term, and then considering how organisations can position themselves to adapt and thrive.
UiPath Test Automation using UiPath Test Suite series, part 3DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 3. In this session, we will cover desktop automation along with UI automation.
Topics covered:
UI automation Introduction,
UI automation Sample
Desktop automation flow
Pradeep Chinnala, Senior Consultant Automation Developer @WonderBotz and UiPath MVP
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...James Anderson
Effective Application Security in Software Delivery lifecycle using Deployment Firewall and DBOM
The modern software delivery process (or the CI/CD process) includes many tools, distributed teams, open-source code, and cloud platforms. Constant focus on speed to release software to market, along with the traditional slow and manual security checks has caused gaps in continuous security as an important piece in the software supply chain. Today organizations feel more susceptible to external and internal cyber threats due to the vast attack surface in their applications supply chain and the lack of end-to-end governance and risk management.
The software team must secure its software delivery process to avoid vulnerability and security breaches. This needs to be achieved with existing tool chains and without extensive rework of the delivery processes. This talk will present strategies and techniques for providing visibility into the true risk of the existing vulnerabilities, preventing the introduction of security issues in the software, resolving vulnerabilities in production environments quickly, and capturing the deployment bill of materials (DBOM).
Speakers:
Bob Boule
Robert Boule is a technology enthusiast with PASSION for technology and making things work along with a knack for helping others understand how things work. He comes with around 20 years of solution engineering experience in application security, software continuous delivery, and SaaS platforms. He is known for his dynamic presentations in CI/CD and application security integrated in software delivery lifecycle.
Gopinath Rebala
Gopinath Rebala is the CTO of OpsMx, where he has overall responsibility for the machine learning and data processing architectures for Secure Software Delivery. Gopi also has a strong connection with our customers, leading design and architecture for strategic implementations. Gopi is a frequent speaker and well-known leader in continuous delivery and integrating security into software delivery.
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...DanBrown980551
Do you want to learn how to model and simulate an electrical network from scratch in under an hour?
Then welcome to this PowSyBl workshop, hosted by Rte, the French Transmission System Operator (TSO)!
During the webinar, you will discover the PowSyBl ecosystem as well as handle and study an electrical network through an interactive Python notebook.
PowSyBl is an open source project hosted by LF Energy, which offers a comprehensive set of features for electrical grid modelling and simulation. Among other advanced features, PowSyBl provides:
- A fully editable and extendable library for grid component modelling;
- Visualization tools to display your network;
- Grid simulation tools, such as power flows, security analyses (with or without remedial actions) and sensitivity analyses;
The framework is mostly written in Java, with a Python binding so that Python developers can access PowSyBl functionalities as well.
What you will learn during the webinar:
- For beginners: discover PowSyBl's functionalities through a quick general presentation and the notebook, without needing any expert coding skills;
- For advanced developers: master the skills to efficiently apply PowSyBl functionalities to your real-world scenarios.
Dev Dives: Train smarter, not harder – active learning and UiPath LLMs for do...UiPathCommunity
💥 Speed, accuracy, and scaling – discover the superpowers of GenAI in action with UiPath Document Understanding and Communications Mining™:
See how to accelerate model training and optimize model performance with active learning
Learn about the latest enhancements to out-of-the-box document processing – with little to no training required
Get an exclusive demo of the new family of UiPath LLMs – GenAI models specialized for processing different types of documents and messages
This is a hands-on session specifically designed for automation developers and AI enthusiasts seeking to enhance their knowledge in leveraging the latest intelligent document processing capabilities offered by UiPath.
Speakers:
👨🏫 Andras Palfi, Senior Product Manager, UiPath
👩🏫 Lenka Dulovicova, Product Program Manager, UiPath
Le nuove frontiere dell'AI nell'RPA con UiPath Autopilot™UiPathCommunity
In questo evento online gratuito, organizzato dalla Community Italiana di UiPath, potrai esplorare le nuove funzionalità di Autopilot, il tool che integra l'Intelligenza Artificiale nei processi di sviluppo e utilizzo delle Automazioni.
📕 Vedremo insieme alcuni esempi dell'utilizzo di Autopilot in diversi tool della Suite UiPath:
Autopilot per Studio Web
Autopilot per Studio
Autopilot per Apps
Clipboard AI
GenAI applicata alla Document Understanding
👨🏫👨💻 Speakers:
Stefano Negro, UiPath MVPx3, RPA Tech Lead @ BSP Consultant
Flavio Martinelli, UiPath MVP 2023, Technical Account Manager @UiPath
Andrei Tasca, RPA Solutions Team Lead @NTT Data
Builder.ai Founder Sachin Dev Duggal's Strategic Approach to Create an Innova...Ramesh Iyer
In today's fast-changing business world, Companies that adapt and embrace new ideas often need help to keep up with the competition. However, fostering a culture of innovation takes much work. It takes vision, leadership and willingness to take risks in the right proportion. Sachin Dev Duggal, co-founder of Builder.ai, has perfected the art of this balance, creating a company culture where creativity and growth are nurtured at each stage.
Accelerate your Kubernetes clusters with Varnish CachingThijs Feryn
A presentation about the usage and availability of Varnish on Kubernetes. This talk explores the capabilities of Varnish caching and shows how to use the Varnish Helm chart to deploy it to Kubernetes.
This presentation was delivered at K8SUG Singapore. See https://feryn.eu/presentations/accelerate-your-kubernetes-clusters-with-varnish-caching-k8sug-singapore-28-2024 for more details.
Securing your Kubernetes cluster_ a step-by-step guide to success !KatiaHIMEUR1
Today, after several years of existence, an extremely active community and an ultra-dynamic ecosystem, Kubernetes has established itself as the de facto standard in container orchestration. Thanks to a wide range of managed services, it has never been so easy to set up a ready-to-use Kubernetes cluster.
However, this ease of use means that the subject of security in Kubernetes is often left for later, or even neglected. This exposes companies to significant risks.
In this talk, I'll show you step-by-step how to secure your Kubernetes cluster for greater peace of mind and reliability.
Securing your Kubernetes cluster_ a step-by-step guide to success !
International Journal of Engineering Research and Development (IJERD)
1. International Journal of Engineering Research and Development
e-ISSN: 2278-067X, p-ISSN: 2278-800X, www.ijerd.com
Volume 7, Issue 3 (May 2013), PP. 44-55
44
A Comparison of Sofm With K-Mean Clustering And Ann
Pattern Reorganization:A Review
1,
Karishmatyagi,Viet Dadri 2,
Gauravsachan ,Viet Dadri
Abstract:- This paper mainly focus on how the input vectors organize themselves into different patterns,
that is, self-organizing maps, based on various algorithms like incremental batch algorithm, k-mean
clustering methods and linear vector quantization when applied in SOFM algorithm itself. It also throws
light on how SOFM algorithm works and defines various topologies. In this paper we have tried to show
how the map looks like after the number of iterations. Moreover we have compared the three known
strategies of self-organizing feature map and discussed the advantage of one technique over other.
Keywords:- feature maps, linear vector quantization, neural network, clusters, classification, neurons,
competitive layer, neuron-neighbours, network design
1. INTRODUCTION
The self-organizing map methods are used for classifying the unknown samples into some categories
based on their features without having the prior knowledge about the input data.It provides us with a
resultant topological network that resembles the input onto the plane exactly as it resembles in high
dimensional space ,that is, the points or neurons that are near to each other in space are also near each other
onto the plane in topological design. The SOM can thus serve as a cluster analysing tool for very large
datasets. These feature maps can also understand and classify the inputs that they have never met before.
1.1 Multiple Layers of Neurons
An artificial neural network consists of several layers. The first layer is an input layer, the last
layer is known as output layer and all the intermediate or middle layers are referred as hidden layer. Each of
these layers has a weight vectorW, a biasor threshold valueb, and an output matrix or vector a. To show
difference between the weight, threshold andoutput vectors in each layer, the number of thelayer is added as
a superscript to every variable respectively.
The network shown below has R1 number of inputs with S1 neurons in the first layer, S2 neurons in the
second layer and so on.Each layer hasdifferent numbers of neurons. A constant value of 1 is given to
each neuron as a threshold value. The central layers are all hidden layers and the output of one first
hidden layer becomes the input for next hidden layer. Thus layer
2 can be analysed as a one-layer network with S1 inputs, S2 neurons, and an S2 × S1 weight matrix
W2.Second layer has input a1 but output is as a2.Since we know all the inputs and outputs of this layer, it
alone can be treated as single-layer network. Multiple-layer networks are very strong and powerful. If we
have a network of two layers where the first layer is sigmoid and
Fig1: Multiple layer neural network
2. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
45
the second layer is linear, then with the help of multilayer networks and backpropogation algorithms ,this can be
combination of any input that can be transformed into any function although with a finite number of
discontinuities arbitrarily.
1.2 Self-organizing in neural networks
Self-organizing neural networks are designed to detect regularities and correlationsin the input they receive
and adapt their future responses to that input accordingly. The strategy followed by self- organizing maps to
recognize the clusters of similar type of inputvectors is that the way neurons physically near each other in the
neuronlayer respond to similar input vectors. Self-organizing maps do not havetarget vectors. They simply
divide the input vectors into clustersof similar vectors. There is no desired output for these types of networks.
1.3Important Self-Organizing functions
We can create competitive layers and self- organizing maps with competlayer and selforgmap, respectively
in Matlab.
1.4 Architecture of competitive layer(a base for SOM layer)
The n dist box takes the input vector p and weight vectorIW1,1, and calculates the negative of the
distances between them and produces the result as a vector having S1 elements.
Fig2: Competitive layer
The net input n1 is calculated by adding the threshold value to the calculated S1 value. If the
value of p is equals to the neuron’s weight vector andall biases are zero, then a neuron will have
maximum net input 0.The competitive transfer C gives the output as 0 for all the neurons accept
for the neurons that have maximum positive value, that is n1.For this the neuron has output as 1
which is declared as winner.If all the threshold values are 0 ,but the weight vector is not equal to the
input vector p then the neuron whose weight vector and the input vector has the least negative
distance are declared as winner with output 1.
2. SELF-ORGANIZING FEATURE MAPS
Self-organizing feature maps (SOFM) classify the input vectors on the basis of their grouping in the input space
and their neighbouring sections. Thus, a self- organizing maps consists of both the information ,that is, about the
distribution of input vectors and their topology.The neurons in the layer of an SOFM are arranged according to
the following topological functions defined below in detail as gridtop, hextop and randtop that organize the
neurons in a grid, hexagonal, or random pattern. Distances between neurons are calculated from their positions
with a distancefunction. Self-organizing feature map network uses the same procedure as of competitive layer to
find the winning neuron i*but together with updating the winning neuron .It also updates the neighbourhood Ni*
(d) of the winning neuron using the Kohonenrule as: iw(q) = iw(q −1) +α(p(q) − iw(q −1)) Here the
neighbourhood Ni* (d) contains the indices for all of the neurons thatlie within a radius d of the winning neuron
i*. Ni (d) = {j,dij≤ d} So,the weights of the neuron that is winner and its nearby neighbours move towards the p
when p is presented.
2.1 Incremental Batch algorithm
In this algorithm, before making any updating ,the whole data set is presented to the network. The algorithm
then determines a winning neuron for each input vector. Then Each weight vector moves to the average position
of all of the input vectors for which it is a winner, or for which it is in the neighbourhood of a winner. In the
figure on right hand side, thefirst diagram shows a two-dimensional neighbourhood of radius d = 1 around
neuron 13. The second diagram shows a neighbourhood of radius d = 2.
3. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
46
Fig3: Neurons with their respective radius and neighbourhood
These neighbourhoods are:N13 (1) = {8,12, 13, 14, and 18} andN13 (2) = {3, 7, 8, 9, 11, 12, 13, 14, 15, 17, 18,
19, and 23}. The neurons in an SOFM can be arranged into one, two or more dimensionalpatterns.The
performance of the network is not sensitive to the exact shape of the neighbourhoods.
2.1.1 Topologies used in feature map
In Matlab ,the following functions are used to specify the locations of neurons in the network:
1. gridtop
pos = gridtop(2,3)
pos =
0 1 0 1 0 1
0 0 1 1 2 2
It gives u an array of neurons of 2 X 3 size.Here neuron 1 has the position (0,0), neuron 2 has the
position (1,0), andneuron
3 has the position (0,1), etc. Let us take an example:
If we write the following code on matlab window.It will create a gridtop topology of
8X10 size of neurons. pos = gridtop(8,10); plotsom(pos)
The graph will look as:
plotsom(pos) that gives the following graph.
Fig 4:Agridtop topology of 8 X10 size
2. HEXTOP
Similarly a graph of 8X10 set of neurons in a hextop topology is created withthe following code:
pos = hextop(8,10);
plotsom(pos)
that gives the following graph.
4. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
47
Fig 5: A hextop topology of 8 X10 size
3. RANDTOP
This function creates neurons in an
N-dimensional random pattern. The following code generates a random pattern of neurons.
pos = randtop(2,3)
pos =
00.7620 0.62601.4218 0.0663 0.7862
0.0925 00.49840.6007 1.1222 1.4228
The 8X10 set of neurons in a randtop topology has the following code:
pos = randtop(8,10);
Fig 6: A randtop topology of 8 X10 size
2.1.2 Architecture
Fig 7: The architecture for SOFM
This architecture works on the basic principle of competitive network, except the difference that no threshold
value is used in calculation here.The competitive transfer function produces 1 for output element a1i
corresponding to i*, the winning neuron. All other output elements in a1 are 0.Neurons close to the winning
neuronare updated along with the winning neuron. We can choose anytopology and distance expressions of
5. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
48
neurons. 2.1.3 Creating a Self-Organizing Map Neural Network The feature maps arecan be created with the
function selforgmap in Matlab window. Thisfunction defines variables used in two phases of learning: •
Ordering-phase learning rate • Ordering-phase steps • Tuning-phase learning rate • Tuning-phase neighbourhood
distance Training: In this phase, the network identifies the winning neuron for all inputs. The weight vectors
then move towards the location of inputs for which they are winner or closest to the winner. The distance or size
of the neighbourhood is changed or updated during trainingthrough two phases. Ordering Phase:The
neighbourhood distancestarts at a given initial distance, and decreases to the tuning neighbourhood distance
(1.0).As we know that distance among the neighbouring neurons decreases during this phase so the neurons
rearrange themselves in the input space in the same way as are arranged physically in the network. Tuning
Phase: This phase lasts for the rest of training or adaption. The neighbourhood sizehas decreased below 1 so
only the winning neuron learns for each sample. One-Dimensional Self-Organizing Map Let us take an example
of 1D map. If there are 100 two-element unit input vectors spread evenly between 0° and90°. angles =
0:0.5*pi/99:0.5*pi; so the data will be plotted with the following code : P = [sin(angles); cos(angles)];
Fig 8: middle stage of 1-D SOM
As shown below, initially all neurons are at the centre of the figure.
Fig 9: starting stage of 1-D SOM
Initially all the weight vectors are concentrated at the centre of the inputvector space.During the training phase
all the weight vectors simultaneously move towards the input vectors.They also become ordered as the
neighbourhood size decreases. Finally thelayer adjusts its weights in such a way that each neuron responds
effectively to a regionof the input space occupied by input vectors. The placement of neighbouring neuron
weight vectors also represents the network design of the input vectors.
6. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
49
Fig10: last stage of 1-D SOM
After 120 cycles, the self-organizing feature map starts to rearrange even further more. After 500 rounds or
cycles the map shows how the neurons are more cleanly n evenly distributed in input space. Two-Dimensional
Self-Organizing Map First of all, some random input data is created with the following code: P = rands(2,1000);
Fig 11: a plot of these 1000 input vectors.
A 5-by-6 two-dimensional map of 30 neurons is used to classify these inputvectors.
After 120 cycles the SOFM look like as
Fig12: the map after 120 cycles
Fig13:a plot after 5000 cycles
7. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
50
So after having near about 5000 cycles, the map shows further smooth distribution of neurons in input space.
So we can say a two-dimensional feature map finds the topology of itsinputs’ space even much better.
Although we know that these feature maps doesnot take very long time to arrange themselves so that
neighbouring neurons can also classify themselves, but the map itself takes long time to arrange according
to the distribution of input neurons vector.
2.1.4 Training with the Batch Algorithm
The batch training algorithm is much faster than the simple incremental algorithm, and it is the default
algorithm for SOFM training. The following code creates 6X6
2-D feature map of 36 neurons:
x = simplecluster_dataset net = selforgmap([6 6]); net = train(net,x);
Fig15: After 200 iterations of the batch algorithm, the map is well distributed through the input space.
If we click the SOM sample hit option in the command window, we will get the following
window shown below as output, which shows the data points of each neuron and also shows how
these data points are distributed in input space.
Fig14: Batch trainingalgorithm command window
8. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
51
The above figure shows the various options that can be performed using batch training
algorithm which is used to create more efficient feature maps. By clicking the various options different
operations can be performed.
Fig 16: data points with their respective neurons.
In a batch training method the process updates the map at every iteration with the help of the
following equation shown below. Some advantages of batch method over the sequential version are:
a) it produces a map faster and b) it does not need a learning rate to converge.
This algorithm runs in the following 4 steps:
1. Retrieval of historical knowledge from a map
2. Starting of the new map.
3. Searching the winning neurons.
4. Updating of weight vectors
All the above steps are performed for each input vector of neurons.
An incremental method to form maps employing the batch training under non- stationary environment
does not need the historical data chunks to form an updated map. Instead, it only
needs some knowledge from the historical maps and the new data in order to be performed. An
experiment using a data set representing different data distribution has been set in earlier research which
shows the topology of the map during training. The map is able to not forget the topology previously
learned from the past data chunks. Therefore, a batch training proposal for non-
stationary environments might be considered.
2.2 K-means algorithm for generating feature maps
It belongs to the category of partitioning algorithms. It minimizesthe reconstruction error by
assigning a nearest neighbour in that compresses the data vectors onto a smaller set of reference vectors.
In this algorithm, the total number of clusters is defined priorly. The new clusters are made by splitting on
the variable with the largest distance. Here the data set is splitted into the selected number of
clusters by maximizing between relative to within - cluster variation. This algorithm is an
iterative procedure that assigns the different data vectors to the predefined number of clusters
that do not overlap each other. In a first trial k-means was used with the specification to produce two
clusters. The data set is splitted into two subsets as:
This figure shows the resulting subsets of clusters.
9. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
52
Fig 17: k-means parameterized for two clusters
k-means splits the dataset in the same way as a hyperplane cutting through the rings would do. It seems
that k-means is unable to take the nonlinear structure of the data set into account. Since the number of
clusters is a parameter to the k-means algorithm, it is a common practice for exploratory data
analysis to run k-means with several different numbers of clusters. K-means with four pre specified
clusters produces the result given in the below figure :
Fig 18: k-means parameterized to 4 clusters
While two of the resulting clusters contain only data of one ring, the other two, marked with a dot
receptively a cross in
figure 18, contain a mixture of the two rings. In summary it can be said, the k- means algorithm is
unable to classify this dataset correctly.
2.3Artificial Neural-Network Pattern Recognition Technique for creating feature
maps
In ANN architecture the classification procedure is implemented in three phases. All the phases
are described in detail below
Learning phase1-Self-organizing feature map (SOFM): This phase provides the “approximate”
quantization of the input space by adapting the weight vectors of the neurons in the feature map.
A problem with SOFM is that if the weights are initialized at small random values it gives different
results at different runs which is not desired when we are trying to optimize the performance of the
algorithm even more. Let us take the case of physician. He wants to review the classification
result of neuromuscular disorder of the patient caught in the form of EMG signals by
motor unit action potentials. For this case, the motor unit is to be identified first that has EMG signals and
then it has to be classified using self- organizing map with linear vector quantization technique.
In order to avoid the problem in this particular case as an example, the weights of the output nodes
should not initialized at small random values but at 0.0001 using vector quantization method. The stepsof
algorithm are shown in the following figure below:
10. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
53
Fig 19: algorithm steps for SOFM Learning phase 2-Learning vector
quantization (LVQ):
It adapt the weights vectors slightly to optimize the classification performance. For this the vector
quantization demands the knowledge of correctly classified inputs. The adaptation carried out during
the first learning phase should be correct and the segmented inputs should be correctly classified.
Updatingthe weight and winning neuron selection is done same as in phase 1. The implementation steps are
given in the following figure below:
11. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
54
Fig 20: Implementation steps for SOFM applied with LVQ for the particular case.
correctly with different input patterns. But the neural networktechnique for developing feature
map supported by linear vector quantization method is fast and efficient mechanism, although
we have tried to explain with a particular case but it is so..The use of theLVQ algorithm with the SOFM
algorithm optimizes the classification boundaries.It makes the algorithm fast and suitable for real-time
applications.
Phase 3: Classification:
This is the last phase where all the input vectors are defined to one of the output nodes.
The use of the LVQ algorithm with the SOFM algorithm optimizes the classification
boundaries through slight adaptation of the weights vectors in ANN pattern reorganization method.
III.CONCLUSION:
Through this paper we have tried to compare the three mainself- organizing feature
mapsdeveloping techniques.We have studied the batch training method in detail together with the
topological functions itsupportsand shown how the vectors organize themselves into different patterns at
each iteration. The k-mean algorithm we studied also develops the feature maps but it is not as efficient as
incremental batch algorithm as it is unable to classify the dataset
REFERENCES:
[1]. International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 –8958,
Volume-1.
[2]. Ieee transactions on neural networks, vol. 11, no. 3, may 2000
[3]. <html><meta http-equiv="content- type”
[4]. content="text/html;charset=utf-8">
[5]. <CITE>genome.tugraz.at/MedicalI nformatics2/<B>SOM</B>.pdf</CI TE>
[6]. Tamkang Journal of Science and Engineering, Vol. 5, No. 1, pp. 35-48 (2002)
[7]. [Char92] Charalambous, C.,“Conjugate gradient algorithm forefficienttraining of
artificial neural networks,” IEEE Proceedings, Vol. 139, No. 3,1992, pp. 301–310.
[8]. [Koho97] Kohonen, T., Self- Organizing Maps, Second Edition, Berlin:Springer- Verlag, 1997.
[9]. [Lipp87] Lippman, R.P., “An Introduction to computing with
12. AComparison Of SOFM With K-Mean Clustering And ANN Pattern …
55
[10]. neural nets,”IEEE ASSP Magazine, 1987, pp. 4–22.
[11]. Neural network toolbox using Matlab
[12]. Internet explorer- google search.