Time filter

Source Type

Chaudhari P.,Symbiosis Institute of Technology | Agarwal H.,Symbiosis Institute of Technology
Advances in Intelligent Systems and Computing | Year: 2017

Deep learning is a thing of tomorrow which is causing a complete drift from shallow architecture to deep architecture and an estimate shows that by 2017 about 10% of computers will be learning rather than processing. Deep learning has fast growing effects in the area of pattern recognition, computer vision, speech recognition, feature extraction, language processing, bioinformatics, and statistical classification. To make a system learn, deep learning makes use of a wide horizon of machine learning algorithms. Gene expression data is uncertain and imprecise. In this paper, we discuss supervised and unsupervised algorithms applied to gene expression dataset. There are intermediate algorithms classified as semi-supervised and self taught which also play an important role to improve the prediction accuracy in diagnosis of cancer. We discuss deep learning algorithms which provide better analysis of hidden patterns in the dataset, thus improving the prediction accuracy. © Springer Science+Business Media Singapore 2017.

Guha S.,Symbiosis Institute of Technology | Sarmah D.K.,Symbiosis Institute of Technology
International Journal of Applied Engineering Research | Year: 2016

In this paper, we propose a dynamic approach to image steganography in the frequency domain in order to increase the hiding potential of the cover image. Concealing a secret file in another file for the sake of its safe transmission via the communication channel is the main aim of a steganography technique. Embedding small size messages in a cover image has less possibilities of hampering the quality of the cover image. However, hiding a bigger size file into a cover image is quite challenging as it may distort the cover image quality thereby revealing the existence of the hidden file. Many researchers who have worked in this domain have followed similar steps to process the cover image before embedding the secret message in it. The cover image is first splitted into blocks of 8X8 or 4X4 pixels per bock and the secret message bits are inserted into the cover blocks uniformly. In our work the cover image blocks are converted from the time domain to frequency domain using Discrete Cosine Transformation (DCT). The blocks are further quantized using the standard JPEG (Joint Photographic Expert Group) quantization matrix. The count of secret message bits that are to be hidden in each block is decided based on the number of coefficients remaining in the blocks after quantizing the cover image. Researchers have used Jpeg Quantization Modification Table (JQTM) to increase the capacity of the stego object. One drawback of using modified quantization table is that it might increase the cover file size than any standard JPEG file. Least Significant Bit substitution method is used for embedding purpose. A text file is taken as the secret message in our work. The quality of the stego object is assessed using peak-signal-to-noise ratio (PSNR) and mean squared error (MSE). © Research India Publications.

Mulay P.,Symbiosis Institute of Technology | Puri K.,Symbiosis Institute of Technology
Advances in Intelligent Systems and Computing | Year: 2016

Plagiarism has become a major cause of concern that has even spread its root across academic area. Universities are becoming more concerned about it because of growing development of internet especially socio media and thereby increasing opportunity among students to copy and paste the electronic content. Students in today’s digital era follow the trend of exchange and copying of information in order to maintain their socio integrity among their circle without considering its long term negative social impact especially from their career perspective. ‘They feel The more you exchange, the more social you are’. To avoid this kind of plagiarism especially in Universities labs, Hawk Eye an innovative mobile plagiarism detection system was an initiative in this regard. Hawk Eye combination with Cohort Intelligence (CI) represents higher state of vision to see things even with more clarity in ordinary experiences, by using Hawk’s keen and observant eyesight, and CI self-supervising nature. This would also help to take appropriate preventive measures to avoid plagiarism from its root among students. ‘Hawk Eye for Cohort (HEC)’ based on comparative analysis of various algorithms like CI and Genetic Search GA can play an important role in formulation of behavioral distribution patterns of students. CI algorithm deploys its self-supervising mechanism to improvise an individual behavior in a cohort and by observing these behavioral patterns, decisions can be taken by teachers in regard of re-design of appropriate evaluation systems to check and stop plagiarism among students. The final outcome of HEC would be an incrementally learning evaluation systems which would iteratively grow with evolving cohort behavioral patterns with every upcoming batch of students. This evolving behavioral patterns search process can be optimized using GA. HEC really would be a concrete evaluation system for analyzing percentage of plagiarism among students, understanding real time reasons behind the growing percentage and coming up with suitable prevention measures in order to cure plagiarism. The concept of study of cohort behavioral distribution pattern using algorithms like GA and CI for plagiarism detection based on student’s socio thinking using different Cohort Analysis Tools is indeed an entirely new idea which is being discussed in this paper in detail. © Springer International Publishing Switzerland 2016.

Puri K.,Symbiosis Institute of Technology | Mulay P.,Symbiosis Institute of Technology
Advances in Intelligent Systems and Computing | Year: 2016

College students today use their smartphones or electronic media smartly to capture software code that are part of their curriculum and then circulate same code among the entire batch, leading to plagiarism. To avoid this kind of plagiarism an innovative plagiarism detection mobile system is needed. The proposed system will eventually help students to develop their coding skills and avoid cloning. The system uses Multi-Language OCR-Compiler Engine to convert the clicked snapshot into relevant text file and triggers appropriate compiler to compile the code. Then the system uses plagiarism detection algorithms that use the concept of tokenizing source code. Tokenizing makes difficult to replace the system by unnecessary comments, punctuations, or changing variables–methods names. Mobile applications like Viper, Plagiarisma.Net, Plagiarism CheckerX can be deployed to check the cloned code. Future extension includes applying modified concepts related to (cloned) flowcharts and also have desktop compatible version of this mobile system. © Springer India 2016.

Shingote K.S.,Symbiosis Institute of Technology | Shahane P.,Symbiosis Institute of Technology
Proceedings - 6th International Advanced Computing Conference, IACC 2016 | Year: 2016

Modern growth in electronics systems, communication system, and information technology assist in designing canal automation system. The Canal irrigation is tremendously use source of water in irrigation. Microcontroller based system is very flexible for any modification required at site, while this system is cheaper than PLC can interface different modules easily. System involves intelligent Microcontroller based Remote Terminal Unit (RTU) which can communicate different sensors, communication modems, memory, ADC and different modules. In this paper we propose a microcontroller based design for flow control system for gate in canal automation. Flow control system consists of sub systems: RTU, Solar Power system, level measurement system, flow measurement system, gate actuator system, and communication system. In this paper more focus on flow control activities of Distributary, laterals and Direct Pipe Outlets (DPOs). Remote Terminal Unit monitors upstream level, downstream level, downstream flow, power status, gate opening, gate health and security. All system components designed to operates on solar power and battery backup. Conventional operational system has some drawbacks and inaccuracies. Proposed system helps to improve irrigation operational efficiency, power use, accuracy in measurement, water distribution and reactions to imbalance and controls the flow at gate location continuously. This system also helps to reduce water wastage and labour dependency. © 2016 IEEE.

Chaphalkar N.B.,College of Engineering, Pune | Sandbhor S.,Symbiosis Institute of Technology
Proceedings of 2016 SAI Computing Conference, SAI 2016 | Year: 2016

Multivariate analysis methods are used to reduce the dimensionality of the vast data available. In this process, the size of available sample happens to be a great deal of concern. Principle Component Analysis (PCA) method has been implemented in various fields including psychology, sociology, economics, engineering and so on. Recommendations made in the literature about the sample adequacy vary and are found to depend upon the study area under consideration. Very less reported work of implementation of PCA to real property valuation has been found so far. Naturally, sample sufficiency for applying PCA to real property valuation has not been studied in detail in the available literature. With no previous assessment of sample requirement, stability of outcomes of PCA applied to real property valuation is checked based on varying sample size. Comparison with the standards mentioned in literature is also done to validate the results of sample sufficiency for application of PCA to ascertain factors affecting real property valuation. © 2016 IEEE.

Guha S.,Symbiosis Institute of Technology | Sarmah D.K.,Symbiosis Institute of Technology
ACM International Conference Proceeding Series | Year: 2016

This paper proposes a steganography technique to enhance the capacity and quality of the stego object. The method is proposed for frequency domain grayscale images. It is assumed that hiding varied number of secret bits in the original image blocks depending upon the DCT coefficients remaining after quantizing the blocks might increase the capacity of the stego image. To enhance the quality PSO can be used to find the appropriate DCT coefficients to conceal the secret data. The secret message is to be encrypted using AES before embedding in the original image blocks to ensure security. © 2016 ACM.

Bhatt D.,Symbiosis Institute of Technology | Gite S.,Symbiosis Institute of Technology
ACM International Conference Proceeding Series | Year: 2016

The ability to categorize the driver behavior is very essentia for advance driver assistance system (ADAS). A metho for identifying driver's behavior is important to assist operativ mode transition between the driver and autonomou vehicles. We propose a driver behavior recognition metho in terms of steering and brake using Hidden Markov Mode (HMM). We are also proposing a HMM-based actuation behavio model for obstacle avoidance using collision avoidanc algorithm. We plan to analyze di-erent driver behavior based on HMM model and produce the results. Afte generating results, we intend to empower it further wit Dempster-Shafer Theory (DST) in order to improve system' performance and accuracy. © 2016 ACM.

Bansod G.V.,Symbiosis Institute of Technology
Proceedings - 2011 IEEE International Conference on Control System, Computing and Engineering, ICCSCE 2011 | Year: 2011

Low power applications are major concern area nowadays. One of the major aspects is cryptography which includes software cryptography in audio as well as video applications. Many algorithms are developed to achieve software cryptography. But in recent year's algorithm are proved to be more efficient. This paper aims to enhance characteristics of HDL based implementation of GRP algorithm. As GRP algorithm is most attractive in term of sorting and cryptographic contents [8]. In this paper low power HDL based design is implemented for audio based applications with much more less power without losing encryption standards as compared to previous algorithms like EMSN[1]. This paper proposes a new algorithm REMS which is rich in encryption standards and consumes much more less power. © 2011 IEEE.

Ahirrao S.,Symbiosis Institute of Technology | Ingle R.,Pune Institute of Technology
Proceedings of the 2013 3rd IEEE International Advance Computing Conference, IACC 2013 | Year: 2013

Cloud computing has become an promising and popular prototype, by presenting its different popular paradigm as infrastructure as a service, platform as a service, software as a service. Now a days there is a challenge of building scalable, available, consistent Cloud Data Stores. In this paper, data partitioning approach such as static and dynamic partitioning are discussed. The main objective of this paper is to bring out the issues in the area of database scalability and to discuss the opportunities open for research work. In this paper, taxonomy of data management in Cloud is proposed. This paper also introduces the design model for partitioning scheme. © 2013 IEEE.

Loading Symbiosis Institute of Technology collaborators
Loading Symbiosis Institute of Technology collaborators