Concepts of Metaheuristics

The procedure used to carry out the learning process in a neural network is called the optimization algorithm (or optimizer). There are many different optimization algorithms. All have different characteristics and performance in terms of memory requirements, processing speed, and numerical precision. Four major parameters are estimated in the process of developing neural network-based models. The four significant parameters of neural networks include : 1)Activation function from Input to Hidden 2)Activation function from Hidden to Output 3)Number of hidden layers 4)The magnitude of weights of the connections (To know more about the above parameters see my tutorial on Artificial Neural Network ) This article is about the methods utilized to estimate the weights of the connections. The process of estimation of weights is similar to optimization problems. Here the weights are design variables. The transfer function prepared to transfer the information from input to output is the objectiv...

Training algorithms for Neural Networks from Mrinmoy Majumder A tutorial on Conjugate Gradient Descent and Newton's Method.Go through the PPT and see if you can understand the concept and apply the same.If not do reply me.

QNANN is an algorithm which are used for update of weights of the neural networks. These algorithms are also known as training algorithm and is known to be popular enough as a technique to optimize the accuracy of neural network. In this presentation the two important techniques for weight update of neural networks at the time of training. Quasi newton artificial neural network training algorithms from Mrinmoy Majumder

Auto and Cross Correlation Coefficient is used for approximation of the auto and cross correlation of the two part of same data series and two different data series. Their magnitude depicts the way they are related to each other..Such concepts are included in the basics of statistics.However the knowledge of these two metrics are important before a model is to be developed for prediction of real time case study. You can find the tutorial by going to my slide-share account .

Can you provide me a feedback on the following tutorial ? How to optimize with the help of the Particle Swarm Optimization(PSO) Technique and xlOptimizer ? This brief tutorial will help you to solve any optimization problem with the application of Particle Swarm Optimization Method and xl addin : xlOptimizer. After a brief introduction about PSO the tutorial show you the steps that you will need to follow for application of PSO in optimization even if you do not know any programming with the help of xlOptimizer.(Some basic knowledge of MS Excel 2010 and later is required). Introduction to particle swarm optimization from Mrinmoy Majumder

The most easy to use decision making techniques : Weighted Sum Method(WSM) and Weighted Product Method(WPM).  Visit this link  to access the WSM and WPM calculators which will help you to identify the best option from a set of available solutions for a given decision making problem(Developed by BP) Admin and Editor(hon) My Kudos Profile

Brief introduction on Weighted Product Method(WPM) and Weighted Sum Method(WSM) with an example. Weighted Product Method : A brief introduction from Mrinmoy Majumder Weighted Sum Method: An Introduction from Mrinmoy Majumder Admin and Editor(hon) My Kudos Profile

The most easy to use decision making techniques : Weighted Sum Method(WSM) and Weighted Product Method(WPM).   Visit this link  to access the WSM and WPM calculators which will help you to identify the best option from a set of available solutions for a given decision making problem. Admin and Editor(hon) My Kudos Profile

In recent years,different objective methods are applied to identify the best option from the available set of alternative for solving a given decision making problem.Analytical Hierarchical Process or AHP is one of such method which is mostly applied in decision making problems where multiple criteria is considered before selecting the best solution among the available set of solutions.This tool will help to identify the best solution with the help of AHP method from a set of feasible solutions to solve the given decision making problem. Suppose we want to select the best location for installation of a hydro power plant. Such kind of power plant require sufficient amount of flow at a regular frequency throughout the year.The interconnection length or the distance between the two banks of the river must also within the suitable limit. So Amount and Frequency of Flow including the Interconnection Length of the available locations can be selected as the Criteria by which the availab...

A brief introduction of Artificial neural network by example from Mrinmoy Majumder

As the name suggest, the characteristic and the trend shown in the applications of these algorithms has not often displayed under a one single paper ? A paper was recently published on the same topic where all the popular hierarchy algorithms were analysed and tested for accuracy. Below is the link to the document. https://arxiv.org/pdf/1307.4186 Popular techniques like Ant Colony,Particle Swarm Optimization,Cucko Search etc. were discussed along with the methodologies. Admin and Editor(hon) Methods of doing research Utilize Optimally : Learn about Optimization by Soft-computation Water based Renewable Energies Call for Papers Special Issue Subscribe to the newsletter / Contact me / Advertise / Call for Paper / Invitation to act as a contributor / My Author Profile / My Kudos Profile

There are some interesting videos availble from you tube which can help to learn the basics,working principle and area of application of the method/ A graphical and textual representation can be found at http://www.utilizeoptimally.com/articles.php?article_id=2 and http://www.utilizeoptimally.com/articles.php?article_id=1 . The concept of neural network was discussed in http://www.utilizeoptimally.com/articles.php?article_id=3 View the videos and learn about the Polynomial Neural Networks. Admin and Editor(hon) Utilize Optimally : Learn about Optimization by Soft-computation Water based Renewable Energies Call for Papers Special Issue Subscribe to the newsletter / Contact me / Advertise / Call for Paper / Invitation to act as a contributor / My Author Profile / My Kudos Profile

There is a software which can help to develop neural network models by the implementation of Polynomial Neural Network(PNN) algorithms like Group Method of Data Handling(GMDH). The procedure for development of the model is simple and easy to use. Whenever data is entered the model will ask to identify input and output for the framework to be developed. Once the input and output is marked the model will utilize GMDH algorithm to identify the optimal characteristics of the model. When the error is equal or reduced to the desired level of accuracy te model is stopped for development. At the end,the model will generate an equation for estimation of the predicator along with relevant metrics like Root Mean Square Error,Correlation,Mean Absolute Error respectively which acts as the fitness function for selection of the optimal network. The description of the working principle is discussed in detail at UO Thanking you, Admin and Editor(hon) Utilize Optimally : Learn about Optimiza...

Neural Tools " NeuralTools can automatically update predictions when input data changes, so you don’t have to manually re-run your predictions when you get new data. Combine with Palisade’s Evolver or Excel’s Solver to optimize tough decisions and achieve your goals like no other Neural Networks    package can "...from Home Page. Link : http://www.palisade.com/neuraltools/ Advantages : This tool can be used along with the Risk Analysis and Evolver Optimization Engine.This tool comes as a part of the Decision Tools suite.Both Risk Analysis and Evolver is avaialble in the Decision Tool Suite along with the Neural model development toolkit. Suggestion : This tool is best suitable for a layman working in the field of stock trading or decision making problems often faced in business. This software is not suitable for anyone who is interested to modify the network architecture of the model.It does not allow the users to play with the basic parameters of the network model. ...

If you want to learn a new tool or technique you must collect a list of learning resources which will be easy to read and understand,easily available or downloadable and cheap or free of cost. Similarly if you want to learn about Artificial Neural Network you must know what are the learning resources you must have in your bookshelf or laptop. A list of : 1.Books 2.Presentations 3.Videos 4.Courses that you may attend were recently published in the Baipatra home page. While preparing this the following factors were considered : 1.The list must not be extensive but selective. 2.Only the best resources will be included. 3.Cost of the same will be minimal or nothing. 4.Links will be active and usable. After considering all the above factors and conducting a search through the net and based on experience finally the list is prepared and presented to those who are really interested to apply the advantage of Artificial Neural Network in their research.

Baipatra: How to predict time dependent variables with the h... : Objective : To predict the probability of rainfall based on monthly mean data of rainfall,runoff and evapotranspiration collected from a si...

The neural networks can be classified based on direction of signal flow,training algorithms,activation functions,network topology etc. All the neural networks can be classified into Supervised,Unsupervised and Reinforcement Class based on the learning method i.e.,training procedures it has adopted.With respect to direction of signal flow such kind of models can be grouped into feed-forward and feedback sub-groups. Based on number of layers neural networks can be subdivided into Single,Multi-layer and Recurrent Neural Networks.

In ANN modeling a graph having input nodes connected to one single output nod is developed.Each node will have its own weight assigned randomly. At the output node all the weighted input nodes are summed up and activated or magnified with the help of Activation Functions. The Activation Functions can be of many types like Step,Ramp,Hypertan,Sinusoidal,Sigmoidal etc. After being activated the output is compared with the desired output(Supervised NN) or the median/mean or any other measures of a set of attributes(for making clusters;Unsupervised NN). If the output from the model is satisfactory with the objective of the model then the model is said to be learning and if not the weightage of the connections are changed and the entire process is repeated. This changing of the weightage is known as Training of the network and is conducted by various methods but mainly by Conjugate Gradient Descent,Back Propagaton and/or Quick Propagation or by some special procedures like Levenberg Marqua...

Types of Neural Networks :  There are two different types of neural network based on the learning mechanism - Supervised and Unsupervised Neural Networks where in case of the former a desired dataset is provided("teacher") with which the model predicted data is compared to find level of accuracy of the model.The Unsupervised neural network will try to identify the inherent properties(data mining) of the input dataset with the help of different unsupervised learning methods. Based on the signal transmission through the neural networks,neural network models can be divided into three distinct classes : Feed forward - where the signal is transmitted only in the forward direction(Input-Hidden-Output); Feedback - a feedback or error correction signal is transmitted back to the source so that deviation from the desired data is actuated and Recurrent - signal is transmitted both ways.Recurrent Neural Networks can use their internal memory to process a...

Category Other Output Type Information/Fact TAG ANN Type of Message * Message Cover Feature ANN or Artificial Neural Network can be defined as the parallel processing units which can mimic the nervous signal of the human being for processing of information signals. It use the weights of the input processing units to create a weighted average of all the inputs.Then applies an activation function after the signal has passed through a Threshold function . Entire procedure is repeated when there is multiple layers within the input and output.Every time the signal creates an output it is compared with the observed one to differentiate between the two(error). Once the error is identified the weights are changed and again the same process is repeated( training ). When the difference between the predicted and actual/available output is within the desired limit or number of train...