Gend Lal Prajapati

On Performing Classification Using SVM with Radial Basis and Polynomial Kernel Functions

Support Vector Machines, a new generation learning system based on recent advances in statistical learning theory deliver state-of-the-art performance in real-world applications such as text categorization, hand-written character recognition, image classification, bio-sequence analysis etc for the classification and regression. This paper emphasizes the classification task with Support Vector Machine. It has several kernel functions including linear, polynomial and radial basis for performing classification. Our comparison between polynomial and radial basis kernel functions for selected feature conclude that radial basis function is preferable than polynomial for large datasets.

An extended approach for SMS security using authentication functions

Nowadays, security of SMS is a crucial aspect because it plays an important role in value added services and mobile commerce. Asymmetric algorithm like Diffie-Hellman can be used to encrypt the SMS message in M-commerce or mobile banking system. We use authentication functions to maintain the integrity of data. Password key exchange protocol based on Diffie-Hellman algorithm generates a secret shared key which can be used in message encryption and in MAC function. MAC (message authentication code) or hash functions are used maintain the integrity of message and can be used with the encryption. These functions also act as an error detecting code or checksum. This paper discusses the comparative analysis of both the authentication functions separately for password key exchange protocol by analyzing some of the security issues. The discussion of this paper concludes that MAC functions are more secure than hash function, but having greater complexity and take more to execute. So, its better to use hash function for maintaining the integrity of message over a network where the transmitted amount of message is very small (SMS). Here, digital signature is generated with RSA to show the functionality of MD5 and SHA1, which prevents SMS from message modification and non-repudiation attack.

On the Inference of Automatic Generation of Software Tests

Automatic generation of software test cases is studied. Within software engineering the software testing phase aims to find errors in software. However, achieving a fully tested program is a hard problem. Moreover, automation of test generation seems to be useful in order to reduce the software development cost. A scheme for the test case generation using tabu search is presented by Srivastava et. al. in 2009. In this paper, we work on their scheme. The scheme is implemented by varying slightly and comprehended by taking example for the further improvement. The study confirms that for assigning priority among the generated test cases for complex problems dynamic clustering is more suitable as compared with k-means clustering.

On the Inference of Context-Free Grammars Based On Bottom-Up Parsing and Search

We consider the problem of incremental learning of context-free grammars, using inductive CYK (Cocke-YoungerKasami) algorithm [4], based on the non-deterministic learning scheme proposed by Nakamura and Matsumoto in 2005 [1]. We implement their learning scheme deterministically and illustrate several examples in order to understand the incremental learning process efficiently. On the basis of this study we also point out some lines of research for possible enhancements.

Extracting relation between brain region pairs from white text

Neuroscience researchers have a keen interest in finding the connection between various brain regions of an organism. Researchers all across the globe are finding new connections everyday and it is very difficult to keep track of all those, so it is important to create a centralized system which is able to give the relation between brain entities. Databases like PubMed contains abstracts and references from a large number of publications in the bio-medical domain. White text is the structured database (a part of PubMed) having structured collection of sentences from various documents containing several neuroscience entities (Brain Regions) and the neuroanatomical connections between them. Identifying such connectivity details automatically in such database is always a point of interest as it helps to centralize all the connectivity in brain regions from various literature. A word vector based machine learning solution is provided to find the relation between 2 entities present in a sentence having a binary label whether a connection is present or not.

Learning alignment profiles for structural similarity measure

Synthesis of context-free grammar based on alignment profile similarity from input sample texts of the unknown target language is studied. Among researches in artificial intelligence, learning context-free grammars from sample strings is a fundamental and important subject, since it is a basic means for defining natural language and for syntactic pattern recognition. Alignment profiles guide the induction of grammar, and hence better alignment profiles improve the accuracy of learned grammars. In this paper, we introduce a scheme for learning alignment profiles from input texts. Several examples are presented to illustrate the scheme and its behavior.

Efficient Learning of Linear Single Tree Grammars

The language class presented here is introduced by Subramanian et al. and includes the class inferred by Makinens algorithm. We present a modification of the work of Subramanian et al. in order to remove some limitations in the original algorithm.

On Learning Context-Free Grammars Using Skeletons

In 1992, Sakakibara introduced a well-known approach for learning context-free grammars from positive samples of structural descriptions (skeletons). In particular, Sakakibarapsilas approach uses reversible tree automata construction algorithm RT. Here, we introduce a modification of the learning algorithm RT for reversible tree automata. With respect to n, where n is the sum of the sizes of the input skeletons, our modification for RT, called e_RT, needs O(n3) operations and achieves the storage space saving by a factor of O(n) over RT. Using our e_RT, we give an algorithm e_RC to learn reversible context-free grammars from positive samples of their structural descriptions. Furthermore, we modify e_RC to learn extended reversible context-free grammars from positive-only examples. Finally, we present summary of our experiments carried out to see how our results compare with those of Sakakibara, which also confirms our approach as efficient and useful.