Showing papers by "Anupam Shukla published in 2014"

Discrete Krill Herd Algorithm --- A Bio-Inspired Meta-Heuristics for Graph Based Network Route Optimization

Abstract: Krill Herd Algorithm KHA is creature inspired meta-heuristic search algorithm, inspired by the tiny sea creature krill and its style of living, which can be utilized in optimization solution foundation of NP --- Hard problems. In this paper we have adopted the various activities of the creature and described a discrete version of the Krill Herd Algorithm for the first time which is favorable for graph network based search and optimization problems. KHA is operated on a multi-parametric road graph for search of optimized path with respect to some parameters and evaluation function and the convergence rate is compared with Ant Colony Optimization ACO and Intelligent Water Drop IWD algorithms. The proposed KHA works well when it comes to decision making and path planning for graph based networks and other discrete event based optimization problems and works on the principle of various random exploration schemes following some parameters which decides whether to include a node/edge or not. Due to the dynamicity of the road network with several dynamic parameters, the optimized path tends to change with intervals, the optimized path changes and will bring about a near fair distribution of vehicles in the road network and withdraw the excessive pressure on the busy roads and pave the way for proper exploitation of the underutilized.

Human Aided Text Summarizer "SAAR" Using Reinforcement Learning

Abstract: Due to information revolution, huge amount of data is available over internet but retrieving correct and relevant data is not an easy task. The information retrieval from search engines is still far greater than that a user can handle and manage. Thus there is need of presenting the information in an abstract way so that one can easily infer the meaning without reading the whole document. In this paper, Human aided text summarizer "SAAR" is being proposed for single document. From the document, a term-sentence matrix is generated. The entries in the matrix are weight from Reinforcement Learning. Thus generated summary is shown to the user and if the user approve it then it is the final summary, otherwise new summary is generated as per the user feedback in form of keywords. Results of experiments on DUC2006 documents indicate that the performance of the proposed approach compares very favorably with other approaches in terms of precision, recall, and F-score.

Evolutionary Ensemble Model for Breast Cancer Classification

Abstract: A major problem in medical science is attaining the correct diagnosis of disease in precedence of its treatment. For the ultimate diagnosis, many tests are generally involved. Too many tests could complicate the main diagnosis process so that even the medical experts might have difficulty in obtaining the end results from those tests. A well-designed computerized diagnosis system could be used to directly attain the ultimate diagnosis with the aid of artificial intelligent algorithms and hybrid system which perform roles as classifiers. In this paper, we describe a Ensemble model which uses MLP, RBF, LVQ models that could be efficiently solve the above stated problem. The use of the approach has fast learning time, smaller requirement for storage space during classification and faster classification with added possibility of incremental learning. The system was comparatively evaluated using different ensemble integration methods for breast cancer diagnosis namely weighted averaging, product, minimum and maximum integration techniques which integrate the results obtained by modules of ensemble, in this case MLP, RBF and LVQ. These models run in parallel and results obtained will be integrated to give final output. The best accuracy, sensitivity and specificity measures are achieved while using minimum integration technique.

Discrete Cuckoo Search Optimization Algorithm for Combinatorial Optimization of Vehicle Route in Graph Based Road Network

Abstract: Cuckoo Search (CS) Algorithm is a well-known and successful nature inspired meta-heuristics which mimicries the salient life-style feature of cuckoo bird and has been widely applied in various continuous domain problems, search analysis and optimization. The algorithm mostly depends on the random placement of the constrained value(s) of variable at the solution set and is being evaluated by the fitness function. There is also provision for slow increment of the solution variable for local search. But here in this paper we have concentrated on the development and application of a modified version of the algorithm called Discrete Cuckoo Search Optimization Algorithm (DCSO) for discrete problem domain like that of the graph based problem and other combinatorial optimization problems like traveling salesman problem etc. The algorithm is first tested on the Travelling Salesman Problem benchmark datasets and then it is applied on a road based graph network for optimization with respect to a non-weighted fitness function of travel time and waiting time and is compared with Ant Colony Optimization (ACO) and Intelligent Water Drop (IWD).

Multi Robot Path Planning for Known and Unknown Target Using Bacteria Foraging Algorithm

Abstract: This paper discusses a new method for multi robot path planning using bacteria foraging algorithm for known and unknown target. Here direction based movement is used to classify unknown and unknown target. The directional is representing by divide the area virtually by clustering based method. In which each cluster point represents the direction. When the target is known robot has idea for direction of movement to reach target. But when the target is unknown robot have no idea related to existence of target in which direction. After decide the direction robot will move according to the bacteria foraging algorithm that modified according to the robotics problem. The algorithm is tested for both simple and complex environments. Four parameters move, time, coverage and energy are calculated for comparison. The results show that proposed method work well for both known and unknown target path planning problem.

Discrete Bacteria Foraging Optimization Algorithm for Vehicle Distribution Optimization in Graph Based Road Network Management

Abstract: Bacteria Foraging Optimization (BFO) is a swarm intelligence optimization technique which has proven to be very effective in continuous search domain having several dimensions. In this paper a discrete and adaptive version of the Bacteria Foraging Optimization Algorithm is being introduced which will be useful in discrete search domain and all kind of multi-dimensional graph based problem. This Discrete Bacteria Foraging Optimization (DBFO) Algorithm is being analyzed and tested in the optimized route foundation phenomenon of a graph based road network and has been compared with the Ant Colony Optimization and Intelligent Water Drop with respect to global convergence. The road system is obsessed with multiple parameters which influence the management of the vehicles in the graph and needs to be analyzed and taken care of. Multiple parameters of the system demand multi-objective optimization using a weighted evaluation function which is carefully designed keeping in mind how the parameters behaves and how its variation dynamically changes the performance of the system. The new discrete version of BFO is being introduced for the first time and it readily suits all kind of graph based and combinatorial optimization problems.

Fusion of Face and Speech Features with Artificial Neural Network for Speaker Authentication

Abstract: Biometric person identity authentication is gaining more and more attention. The authentication task performed by an expert is a binary classification problem: reject or accept identity claim. Combining experts, each based on a different modality (speech, face, fingerprint, etc.), increases the performance and robustness of identity authentication systems. In this context, a key issue is the fusion of the different experts for taking a final decision (i.e., accept or reject identity claim). An automatic speaker authentication system based solely on speech or faces is often not able to meet the system performance requirements. We have developed a prototype biometric system, which integrates faces and speech utterances. The system overcomes the limitations of face recognition systems as well as speech based verification systems. The work here is broadly classified into three parts; firstly, extractions of speech parameters, secondly extractions of image parameters and finally the simulated Artificia...

Road Traffic Management Using Egyptian Vulture Optimization Algorithm: A New Graph Agent-Based Optimization Meta-Heuristic Algorithm

Abstract: In this paper we have continued the introduction and application of a new nature inspired meta-heuristics algorithm called Egyptian Vulture Optimization Algorithm (EVOA) which primarily favors combinatorial optimization problems and graph based problems. The algorithm is derived from the nature, behavior and key skills of the Egyptian Vultures for acquiring food for leading their livelihood. These spectacular, innovative and adaptive acts make Egyptian Vultures as one of the most intelligent of its kind among birds. The details of the bird’s habit and the mathematical modeling steps of the algorithm are illustrated demonstrating how the meta-heuristics can be applied on the route planning for a graph based road network depending on the multi-parametric optimization of distance (travel time) and waiting time. Due to the dynamically changing behavior of the waiting time for the various crossings of the network, the system is dynamic system and the best optimized path tend to change with time and will help in diverging the vehicle flow through the various routes of the road network. The road network problem is considered as a special case of Travelling Salesman Problem based combinatorial problem with changes and constraint imposed and also the steps of the algorithm is also changed to suit and quicken the solution finding process and imbibe the theory of chance and rejection subsequently. The results of application of the algorithm on the road network and its comparison with Ant Colony Optimization Algorithm & Intelligent Water Drops Algorithm show that the algorithm works well and provides the scope of utilization in similar kind of problems like path planning, scheduling, routing, and other constraint driven problems. EVOA is one of the very few algorithms which are readily applicable for discrete domain problems.

Smart Card Application for Attendance Management System

Abstract: This paper presents attendance management system based on the smart cards. This introduces the entities of the system and describes the set up of attendance management system and the role of these entities in the management system for managing attendance. This also represents the graphical representation of the working of entities of the system. This includes how card makes connection with the attendance reader and how verification authority verifies the smart card. And GUI interface for communicating with the card through this application is also shown with some description.

Gender Identification Using Gait Biometrics

Abstract: Soft biometrics-based gender classification is an interesting and a challenging area of neural networking and has potential application in visual surveillance as well as human–computer interaction. In this paper, we have investigated gender recognition from human gait in image sequence. For the above purpose, we have extracted silhouette of 15 males and 15 females from the database collected from CASIA Gait Database (Dataset B). The computer-vision-based gender classification is then carried out on the basis of standard deviation, center of mass, and height from head to toe. Experimental results demonstrate that the present gender recognition systems achieve superior recognition performance of 96.8 % on feed-forward back-propagation (FFBP) network. Data on different networks have also been trained and tested. The above study indicates that gait-based gender recognition is one of the best reliable biometric technologies that can be used to monitor people without their cooperation. Controlled environments such as banks, military installations, and even airports need to quickly detect threats and provide differing levels of access to different user groups.

A Diversity-Based Comparative Study for Advance Variants of Differential Evolution

Abstract: Differential evolution (DE) is a vector population-based stochastic search optimization algorithm. DE converges faster, finds the global optimum independent to initial parameters, and uses few control parameters. The exploration and exploitation are the two important diversity characteristics of population-based stochastic search optimization algorithms. Exploration and exploitation are compliment to each other, i.e., a better exploration results in worse exploitation and vice versa. The objective of an efficient algorithm is to maintain the proper balance between exploration and exploitation. This paper focuses on a comparative study based on diversity measures for DE and its prominent variants, namely JADE, jDE, OBDE, and SaDE.