|
Research papers
Following are the research papers published from the
department:
1) Evolution Induced Secondary Immunity: An Artificial
Immune System based Intrusion Detection System
Authors:
Divyata Dal,
Siby Abraham, Ajith Abraham, Sugata Sanyal
and Mukund Sanglikar
Abstract:
The
analogy between Immune Systems and Intrusion Detection
Systems encourage the use of Artificial Immune Systems for
anomaly detection in computer networks. This paper describes
a technique of applying Artificial Immune System along with
Genetic algorithm to develop an Intrusion Detection System.
Far from developing Primary Immune Response, as most of the
related works do, it attempts to evolve this Primary Immune
Response to a Secondary Immune Response using the concept of
memory cells prevalent in Natural Immune Systems. A Genetic
Algorithm using genetic operators- selection, cloning,
crossover and mutation- facilitates this. Memory cells
formed enable faster detection of already encountered
attacks. These memory cells, being highly random in nature,
are dependent on the evolution of the detectors and
guarantee greater immunity from anomalies and attacks. The
fact that the whole procedure is enveloped in the concepts
of Approximate Binding and Memory Cells of lightweight of
Natural Immune Systems makes this system reliable, robust
and quick responding
2) Finding numerical solution to a Diophantine
Equation:simulated annealing as a viable search strategy
Authors:
Siby Abraham , Mukund Sanglikar
Abstract:
This paper introduces Simulated Annealing as a viable
informed search strategy to find numerical solutions to
Diophantine Equations, for which there exists no general
method of finding solutions. The methodology proposed uses a
tried and tested heuristic as the energy function to guide
the search process. It behaves like a typical steepest
valley descending till it reaches states facing only worse
states as possible successors. Then the difference in
heuristic function values of the current state and the
candidate successor state is taken as the energy difference
(ΔE). The heuristic function value of
the initial state is taken as the time-varying parameter
‘Temperature’ (T), which is reduced gradually using a domain
dependent annealing schedule. Here, the transition from a
current state to a successor state depends on the
satisfaction of the probability condition e - (ΔE / T). Experimental results
performed on Diophantine Equations of varied types validate
the effectiveness of the procedure.
3) Solution to Diophantine Equation: A Genetic Algorithm
Application
Authors:
Siby Abraham , Mukund Sanglikar
Abstract:
This
paper attempts to develop a Diophantine Equation Solver
using Genetic Algorithm. It shows how the directed random
search procedure of Genetic Algorithm (GA) reduces the
search procedure substantially and comes to a solution. The
GA developed in this paper obtains quick solutions to the
general diophantine equation given by f(a1, a2, …….,
an,x1,x2,…….,xk)=N, where ai and N are integers. A
chromosome is represented by integers within the supplied
range of values. The evolution process is carried out by
generating off-springs using genetic operators - mutation
and crossover - on a population of five chromosomes. The
pure random selection of these operators and subsequent
random exchange of genes, in case of mutation or swapping of
genes from the randomly generated crossover point onwards,
in case of crossover results in diverse nature of
chromosomes in the evolution process. In order to overcome
plateau and reduce premature convergence in the fitness
landscape, we apply biological concept of host-parasite
co-evolution. The evaluation function constructed acts as an
automatically defined function, which restricts the
occurrence of unwanted and already tested chromosomes. The
GA developed offers solutions to a diophantine equation
within a given range. The methodology also offers an option
to switch over to a new initial population of chromosomes
when the existing population becomes obsolete.
4) Nature’s way of Avoiding
Premature Convergence: A case study of Diophantine Equations
Authors:
Siby Abraham , Mukund Sanglikar
Abstract:
This paper proposes natural phenomenon of Host Parasite
Co-Evolution as a strategy to avoid premature convergence in
the fitness landscape of a Genetic Algorithm developed to
find numerical solutions to diophantine equations given by
f(a1,
a2,
a3,
…, an,
x1,
x2,
x3,…,
xn)
= N where a
i
and N are integers.
The procedure described avoids the occurrence of local
hilltops as premature converging points in the search space
of the GA developed. It uses selection, inversion and
crossover as genetic operators. As the genetic process is
evolved, the procedure gets converged prematurely. These
premature converging points are taken as attacks by
parasites. The candidate solutions are trained to be
defensive to these attacks using a composite evaluation
function. The parasites, in turn, develop new attacks in the
form of creating new converging points. This continuous
process of avoidance by hosts and attacks by parasites
reciprocally induce evolution. The pressure of this evolving
attacks and counter attacks works effectively as a powerful
optimization tool and hence directs the procedure to
discover a solution by generating untested, diverse and
fresh candidates. The paper tries to establish statistically
the diversity of attacks of parasites as it happens in
nature. It also attempts to demonstrate the inherent
relation between number of attacks by parasites and number
of generations produced irrespective of the degree of the
equation and the number of variables used.
5) Finding Solution to a Hard Problem: An Evolutionary and
Coevolutionary Approach
Authors:
Siby Abraham, Mukund Sanglikar
Abstract:
This paper introduces a strategy, which uses Genetic
Algorithm as an evolutionary technique and Competitive
Two-population Co-evolution as a co-evolutionary technique
for finding a solution to a hard problem. The strategy is
applied to find numerical solution to a Diophantine
Equation, for which there does not exist a method to solve.
The GA uses selection, inversion and crossover as genetic
operators and helps to undergo evolution on a population of
twenty-five chromosomes. The local optimum points
encountered in the fitness landscape during the evolution
are tackled using Competitive Two-population Co-evolution by
creating populations of local optimum points and fitness
values of new untested and fit chromosomes. The paper shows
that the strategy is continuous, random, evolutionary and
adaptive in nature and is open to the concept of
introduction of a fresh population in the rare event that
the present population looses its diversity and hence
becomes obsolete
6) Particle swarm optimisation based Diophantine equation
solver
Authors:Siby
Abraham , Sugata Sanyal, Mukund Sanglikar
Abstract:
The paper introduces particle swarm optimization as a viable
strategy to find numerical solution of Diophantine equation,
for which there exists no general method of finding
solutions. The proposed methodology uses a population of
integer particles. The candidate solutions in the feasible
space are optimized to have better positions through
particle best and global best positions. The methodology,
which follows fully connected neighbourhood topology, can
offer many solutions of such equations.
7) Defect-based Reliability Analysis Model for Business
Systems Deployed on a large Scale
Authors:
Nita
Sarang, Mukund A. Sanglikar
Abstract:
Quality of software releases is paramount in business
systems deployed on a large-scale. As the deployed software
systems grow in functionality, the software maintenance task
becomes increasingly complex; maintaining quality control
over the life of the software is a challenge. We undertake
an analysis of defect data from two product-lines in a
specific business domain. We use hypothesis testing to
understand the underlying software maintenance process and
establish the defect models. The defect models are used to
define an optimal level of testing that meets the software
reliability requirements.
8) An Analysis of Effort Variance in Software Maintenance
Projects
Authors:
Nita
Sarang, Mukund A. Sanglikar
Abstract:
Quantitative project management, understanding process
variations and improving overall process capability, are
fundamental aspects of process improvements and are now
strongly propagated by all best-practice models of process
improvement. Organizations are moving to the next level of
quantitative management where empirical methods are used to
establish process predictability, thus enabling better
project planning and management. In this paper we use
empirical methods to analyze Effort Variance in software
maintenance projects. The Effort Variance model established
was used to identify process improvements and baseline
performance.
9) Computational Diagnosis of Learning Disability
Authors:
Kavita Jain, Pooja Manghirmalani, Jyotshna Dongardive and
Siby Abraham
Abstract:
The
paper proposes a perceptron based artificial neural
network model for diagnosing learning disability using
curriculum based test conducted by special educators in
medical environment. The model comprises of a single input
layer with eleven units which correspond to different
sections of a conventional test and one output unit. The
method is not only devoid of typical computational
complexity and sophistication associated with other methods
but gives comparable experimental results on detection
measures like accuracy, sensitivity and specificity.
10) Building Consensus of Human Papillomavirus using Genetic
Algorithm
Authors:
Aditya Bir, Jyotsna Dongerdive, Suruchi Jamkhedkar, Siby
Abraham.
Abstract:
The paper introduces novel three tier architecture to
build consensus of Human Papillomavirus (HPV). The proposed
procedure is based on simulation and uses all complete
genomic DNA sequences of registered HPV strains available in
NCBI GenBank. It uses the multiple sequence alignment tool
Clustal X to align these sequences. Genetic Algorithm (GA)
is used to evolve an optimized population of complete
genomic DNA sequences. The GA, which uses domain specific
genetic operators like migration, rank selection, mutation
and crossover, adopts a novel methodology in defining the
fitness function. A modified approach of the Weight Matrix
Algorithm is applied on the optimized and evolved population
to find a consensus of HPV. The effectiveness of the
procedure is validated with experimental results.
11) Finding Motifs Using Harmony Search
Authors:
Jyotsna Dongerdive, Aarti Patil, Aditya Bir, Suruchi
Jamkhedkar, Siby Abraham.
Abstract:
The paper proposes a novel methodology for finding motifs of
biological sequences. It uses music inspired meta-heuristic
optimization technique called harmony search to find Motif.
The
model is based on randomly generated l-mers as the
initial
harmony memory. Pitch adjustment and random selection are
used to generate new l-mers, which are adjudged by a
defined
objective function. The proposed method is experimentally
validated using sequences of Human Papillomavirus strains
obtained from accredited and authorized sources.
12) Envelope Model: Data Access from any where
Authors:
Sushil Kulkarni
Abstract
In this paper, a definition of mobile database and mobile
transaction with illustrations from spatial and temporal
databases are given. Corresponding to mobile transaction
mobile query is defined. New model called Envelope model is
presented to execute a mobile query using various base
stations of the closed spheres. A mobile query is defined as
the union of jumping queries and contains only one unary or
binary operator operating on database. A mobile transaction
is the union of jumping transactions where each jumping
transaction is the unit of execution at the base station.
The jumping condition is checked prior to the starting of
each jumping query at the base station. This model is
implemented using developed environment called Data Invoking
Capsule. It is located at every base station and its main
job is to transfer the messages from one base station to
another, executing jumping queries and storing envelopes. A
mobile unit makes a request to execute mobile query to base
station called initializer, who creates a basket that store
jumping query number and the directed parse tree. This
basket is carried by the mobile unit till mobile query is
completely executed. Directed parse tree is created by an
initalizer using parse tree. It gives guidelines for
execution of jumping query at the respective base stations.
After the jumping query is executed an envelope is created,
that store jumping query number and the result obtained by
executing jumping query. Envelope flies from one base
station to another as required by another base station for
executing next jumping query under the guidelines of
directed parse tree. Every base station may not have the
required database on which jumping query is executed so the
method is proposed to execute it. Algorithms for creating
directed parse tree and an envelope model are provided.
Parallel execution of jumping transaction at a base station
is also discussed. Envelope model for mobile transaction is
introduced.
13) Spatio-Temporal Schema Model and Carrie and Courier
mobile query processor for Spatio-Temporal database
Authors:
Sushil Kulkarni, Dr. M.A. Sangalikar
Abstract:
Geographical Information Systems and their underlying
support, spatio-temporal database systems, deal with data
related to changes in geographical space according to time.
The space of interest is two dimensional abstraction of the
surface of the earth and the change of attribute instances
related to it. In this paper, we propose Y- schema, which
will enable to handle spatio-temporal data. Also we know
that in urban area, mobile users are increasing day by day
and there is a demand of retrieving, manipulating spatio-temporal
data. Here we are proposing a query processor called Carrier
and Courier processor to handle the spatio-temporal mobile
queries.
13) Envelope Transaction Mobile Model
Authors:
Sushil Kulkarni, Dr. M.A. Sangalikar
Abstract:
A long lived mobile transaction is divided into different
sub-transactions called jumping transactions. The proposed
model explains the commit of jumping transaction at the
server or at mobile unit. Model takes care of disconnection
because of jumping nature of mobile unit. This model relaxes
isolation property using intension commit of jumping
transaction.
14) Lock Mechanism for a cluster of peers
Authors:
Sushil Kulkarni, Dr. M.A. Sangalikar
Abstract:
A new approach for caching data objects and locks by a centroid
from a server in a cluster of peers using ad hoc
transactions is presented. A cluster of peers is formed by a
set of mobile hosts, which are clustered around a single
mobile host called a centroid. A set of mobile hosts are
called peers. A centroid is formed as a temporary work group
for processing and exchanging information.
In life cycle of a mobile transaction
processing system, data is cached from the server and a
centroid is disconnected from the server to create its own
cluster of peers. A dynamically configurable data
processing space, called an envelope repository, is created
to process various transactions at the centroid. The
processed data is kept in an envelope at the
private work space of the centroid and is brought to the
envelope repository whenever required.
15) Envelope Transaction Mechanism for a Cluster of Peers
Authors:
Sushil Kulkarni, Dr. M.A. Sangalikar
Abstract:
A new approach for mobile transaction processing is
presented in a cluster of peers. A cluster is a dynamic
collection of mobile hosts called peers and are clustered
around a single mobile host called a centroid to form a
temporary work group for processing and exchanging
information. All peers are connected to the centroid using a
short range wireless network. All peers and the centroid are
members of a peer connected set. The centroid is chosen in
such a way that it has a strong connection to the mobile
support station. After collecting data from a server, the
centroid is free to disconnect from the server and ready to
create its own cluster of peers. A dynamically configurable
data processing space, called envelope repository, is
created to process various transactions at the centroid. The
processed data is kept in an envelope at a private work
space of the centroid and is brought to the envelope
repository whenever required. Peer initiates a jumping
transaction to fetch data objects to the centroid. The
jumping transaction initiates an envelope transaction to
acquire the necessary locks on data objects at envelope
repository. After getting the locks, pseudo transaction
processes the data objects to give the result stored in an
envelope. Envelope pseudo write protocol is designed to make
jumping transactions globally serialized.
|
