KanGAL Weekly Seminars (Temporarily stopped due to
leave of absence)
For information on upcoming KanGAL seminars, please click here.
For information on past KanGAL seminars, please click here.
Title: Optimization in shape matching in the context
of molecular docking
Speaker: Vipin Kumar Tripathi, Ph.D. Oral
examination, Mechanical Engineering
Schedule and Venue: 3:30PM, 28th December, 2006,
FB-364
Presentation Slides: Click here to download presentation
slides
Title: Multi-Objective evolutionary Algorithms for
resource allocation problems
Speaker: Dilip Datta, Ph.D. Open Seminar, Mechanical
Engineering
Schedule and Venue: 5:00 PM, 25th July, 2006, FB-364
Presentation Slides: Click here to download presentation slides
Title: Multi-Objective Dynamic Optimization using Evolutionary Algorithms
Speaker: Udaya Bhaskara Rao N., M.Tech. Defense, Mechanical Engg.
Schedule and Venue: 11:00 AM, 25th July, 2006, FB-364
Abstract
Presentation Slides: Click here to download presentation slides
Title: Synthesis of Path Generation Complaint Mechanism using Local Search Based Multi-objective Genetic Algorithm.
Speaker: Deepak Sharma, Ph.D. State of Art seminar, Mechanical Engg.
Schedule and Venue: 4:00 PM, 24th April, 2006,
FB-364
Abstract: Path generation compliant mechanisms
(PGCM) are flexible structures which generate some desired path and/or
transmit force by undergoing elastic deformation (under some applied load)
instead of through rigid linkages/joints as in rigid body mechanism.
In the present work, multi-objective problem is posed for evolving of PGCM using local-search based genetic algorithm. Minimization of weight of structure and minimization of input energy to the structure have been considered as two conflicting objective functions subjected to maximum of 10% of deviation between five precision points of prescribed path and corresponding five points obtained on the actual path of a specified point on structure after FEM analysis and a constraint on stress. Geometrical non-linear finite element model is used for the synthesis of PGCM. On the basis of the calculated function and constraint values, an evolutionary algorithm (NSGA-II) is used to find the optimal solution.
A local search based multi-objective GA is used to reduce the computation time and to improve the quality of GA solutions. The Pareto-optimal front obtained shows different trade-off solutions from minimum weight to the maximum weight of structure. The minimum weight solution is corresponding to the maximum input energy to the structure and vice-versa.
In the present work, multi-objective problem is posed for evolving of PGCM using local-search based genetic algorithm. Minimization of weight of structure and minimization of input energy to the structure have been considered as two conflicting objective functions subjected to maximum of 10% of deviation between five precision points of prescribed path and corresponding five points obtained on the actual path of a specified point on structure after FEM analysis and a constraint on stress. Geometrical non-linear finite element model is used for the synthesis of PGCM. On the basis of the calculated function and constraint values, an evolutionary algorithm (NSGA-II) is used to find the optimal solution.
A local search based multi-objective GA is used to reduce the computation time and to improve the quality of GA solutions. The Pareto-optimal front obtained shows different trade-off solutions from minimum weight to the maximum weight of structure. The minimum weight solution is corresponding to the maximum input energy to the structure and vice-versa.
Presentation Slides: Click here to download presentation slides
Title: Innovization-Innovative solutions through optimization.
Speaker: Aravind Srinivasan, Graduate Student, Mechanical Engg.
Schedule and Venue: 4:00 PM, 10th March, 2006, FB-364
Abstract: This presentation is about a new design
methodology in the context of finding new and innovative design principles by using
optimization techniques. The task of innovization stretches the scope
beyond an optimization task and attempts to unveil new and innovative
design principles relating to decision variables and objectives, so that
a deeper understanding of the problem can be obtained. The innovization
procedure has been applied to a number of engineering design problems.
Presentation Slides: Click here to download presentation
slides
Title: Optimization Realities and Opportunities
Speaker: Vinay Ramanath, General Electric, Bangalore
Schedule and Venue: 3:30 PM, 10th March, 2006, FB-364
Abstract: In the academic world optimization is a
crisp, well-defined mathematical topic dealing with accurate objective
functions and constraints, where the main research issue is to get to the
optimum in the least number of function evaluations (even if it takes a long
time to set the problem up).
In the engineering world...Optimization is a very different animal - and that is the point of this presentation.
Title: "Ecomagination - Organic Growth Story" - General Electric
Speaker: Vaira Saravanan, CoE Manager, General Electric, Bangalore
Schedule and Venue: 4 PM, 9th March, 2006, FB-364
Abstract: Ecomagination is on solving customer's toughest problem thru environment friendly products, the story is about what are they and how the Bangalore center is involved in doing that
Title: Wireless Sensor Networks - From Architecture
Exploration to Deployment
Speaker: Prof. Lothar Thiele, ETH Zurich
Schedule and Venue: 5 PM, 1st March, 2006, L-11
Title: Applications of Randomized Search Algorithms in
Computational Biology
Speaker: Dr. Eckart Ziztler, ETH Zurich
Schedule and Venue: 2 PM, 1st March, 2006, L-11
Title: Multi-Objective Optimization "100 Objectives":
No more a myth
Schedule and Venue: 4 PM, 28th Feb, 2006,
FB-371
Speaker: Dhish Kumar Saxena
Abstract: Many real-world applications of
multi-objective optimization involve a large number (10 or more) of
objectives. Existing evolutionary multi-objective optimization (EMO) methods
are applied only to problems having smaller number of objectives (about five
or so) for the task of finding a well-representative set of Pareto-optimal
solutions, in a single simulation run. Having adequately shown this task, EMO
researchers/practitioners must now investigate if these methodologies can
really be used for a large number of objectives. The major impediments in
handling large number of objectives relate to stagnation of search process,
increased dimensionality of Pareto-optimal front, large computational cost,
and difficulty in visualization of the objective space. These difficulties, do
and would continue to persist, in "M"-objective optimization problems, having
an "M"-dimensional Pareto-optimal front. In this paper, we propose an EMO
procedure for solving those large-objective "M" problems, which degenerate to
possess a lower-dimensional Pareto-optimal front (lower than "M"). Such
problems may often be observed in practice, as ones with similarities in
optimal solutions for different objectives - a matter which may not be obvious
from the problem description. The proposed method is a principal component
analysis (PCA) based EMO procedure, which progresses iteratively from the
interior of the search space towards the Pareto-optimal region by adaptively
finding the correct lower-dimensional interactions. The efficacy of the
procedure is demonstrated by solving up to 30-objective optimization
problems.
Title: Portfolio-Optimization with Multi-Objective Evolutionary Algorithms in the case of Complex Constraints
Speaker: Benedikt Scheckenbach
Schedule and Venue: 4 PM, 26th August, 2005, Lecture Hall Complex-6
Abstract: Portfolio optimization is inherently a multi-criteria optimization problem. Although one could imagine various citerias, only return and risk are regarded as relevant conflicting criterias in the classical approach. In that case the pareto-optimal front could be derived analytically, if there were only simple constraints. However, certain investment-laws and inhouse-requirements introduce complex constraints that require the application of MOEAs. Furthermore the problem allows the construction of envelopes for different constraints' specifications. The use of this nice feature may result in improved convergence and will be further investigated.
Presentation Slides: Click here to download presentation slides
Title: Functional LandScape Models
Speaker: Pawan Kumar Singh Nain
Schedule and Venue: 11 AM, 8th July, 2005, ME
Conference Room
Title: Evolution and Learning at Honda Research
Institute Europe
Speaker: Dr. Yaochu Jin, Honda Research Institute,
Germany
Abstract: This talk gives a short description of main research actitivities on evolution and learning at Honda Research Instiiute Europe. Selected topics, such as metamodeling in evolutionary optimization, search for robust solutions, dynamic optimization, and multi-objective optimization will be discussed. Theoretic issues as well as application examples will be presented.
Title: Theoretical Approach to Population Dynamics in
Genetic Algorithms
Speaker: Dr. Tatsuya Okabe
Abstract: Although evolutionary algorithms (EAs)
have shown several good results
in applications, the theoretical work for EAs is still few.
Recently, several researchers have started theoretical investigation
for EAs, in particular convergence analysis, model of EAs, etc.
To our knowledge, theoretical work for population dynamics during
optimization was not carried out. We strongly believe that theoretical
analysis of population dynamics will not help us only to understand the
working mechanism of EAs but also to design an innovative algorithm for
tackling optimization problem. This presentation is one of trials
to understand population dynamics in genetic algorithms (GAs) with single
objective optimization. After modeling GAs, several findings will be
discussed in this presentation.
Title: Multi-Objective Interplanetary Trajectory Optimization Using Genetic Algorithms
Speaker: Ganesh Neema
Abstract: Present work shows an optimal orbital
trajectory of a spacecraft, from a given departure planet to a given arrival planet. In search of an optimal
solution, gravitational power of any planet (swingby) may or may not be
considered. Also if swingbys are to be used, then how many times and from
which planet each times. The optimization procedure is complex in nature
because of the difficulties (non-linearity, highly constrained, and
discreteness) associated with the trajectory optimization problem. Hence,
a classical algorithm is difficult to apply and also not guaranteed to
converge at global optimal solution. Therefore an evolutionary algorithm
(EA) strategies are used in the present work. EAs have shown to be
successful in solving such problems in other problem domains. A model
comprising of multi objective problems is mentioned in the study using
combinatorial genetic algorithms.
Title: Reliability Based Design Optimization
Speaker: Sulabh Gupta
Abstract: Optimization satisfying a required reliability criterion is a foremost
requirement in almost all practical engineering designs. This presentation
is aimed at providing a brief review of the various RBDO formulations and
methods. A brief description of the inherent concepts is also provided. It
also suggests a simplified model of the RBDO problem to work with.
Presentation Slides: Click here to download presentation slides
Title: Multi-objective Optimization of a Two
Dimensional Sheet Cutting Problem Using Genetic Algorithms
Speaker: Santosh Tiwari
Abstract: We present here a method for optimizing
the layout of rectangular parts placed on a rectangular sheet to cut out
various objects. We investigate here two types of cutting problems (i)
in which guillotine cutting (cutting from edge to edge) is required (mostly
metallic sheets where each cut is made individually for one single sheet) and
(ii) in which guillotine cutting is not required (cuts which can be made using
a punch) i.e. for materials like paper or rubber where the sheets to be cut
can be laid side by side or on top of one another and one single cut can be
made. For non-guillotine cutting, both real and and binary variable
formulations are studied and it is shown that binary variable formulation of
the design problem gives better solutions. For guillotine cutting binary
variable formulation is studied and it has been shown for known cases that
globally optimum solution(s) are obtained.
Presentation Slides: Click here to download presentation slides
Title: Metamodels: a brief discussion
Speaker: Pawan Kumar Singh Nain
Abstract: The real world optimization problems are
often computationally expensive. The use of statistical techiniques to build approximation of expensive
computer analysis code is a possible solution. The metamodels, or model of
a model are used to provide such an efficient alternative. In this talk,
four such sample approximation techniques, namely response surface
methodology, neural networks, inductive learning and kriging are discussed. The
recommendations for the appropriate use of statistical approximation
technique in the given situation and how common pitfalls can be avoided is
briefly discussed.
References: Metamodels for Computer-Based Engineering
Design: Survey and Recommandations by T. W. Simpson, J. Peplinski, P. N. Koch
and J. K. Allen
Title: Gene, MicroArrays and GAs
Speaker: Ashish Anand
Abstract: Main idea is to present basics of genomics with blend of bioinformatics and computational biology in layman terminology. Following things are discussed - (1) What are Genes and why it is important to study them? (2) What is microarray and how it is helping in understand various diseases, developmental stages? (3) Few problems where GAs are being applied.
Presentation Slides: Click here to download presentation slides
Title: Analysis of Unary and Binary Performance
Indicators
Speaker: Santosh Tiwari
Abstract: We present here a subjective analysis of
unary and binary performance indicators. We analyze various advantages and
disadvantages of each approach and go on to describe few very common unary
performance indicators. Based on the merits and de-merits of these indicators,
we then analyze a binary performance indicator viz binary-epsilon
indicator.
Presentation Slides: Click here to download presentation slides
Title: Machine Learning in Iterated Prisoner's
Dilemma using Evolutionary Alorithms
Speaker: Shashi Mittal
Abstract: In this talk, a framework for Machine
Learning in the game of Iterated Prisoner's Dilemma using Multiobjective
Evolutionary Algorithms is presented. In particular, we present the results
obtained using the NSGA-II algorithm, and discuss its advantages over the
Single Objective Optimization paradigm. We also present the essentials a
strategy should have to win in the game of Iterated Prisoner's Dilemma.
References: Axelrod, R. The Evolution of Strategies
in the Iterated Prisoner's Dilemma. In L. Davis, editor, Genetic Algorithms
and Simulated Annealing, Los Altos, CA: Morgan Kaufman, 1987.
Presentation Slides: Click here to download presentation slides
Title: Omni Optimizer: A Procedure for single and
Multi-objective Optimization
Speaker: Santosh Tiwari
Abstract: Due to the vagaries of optimization
problems encountered in practice,
users resort to different algorithms for solving different
optimization problems. In this paper, we suggest an
optimization procedure which specializes in solving
multi-objective, multi-global problems. The algorithm
is carefully designed so as to degenerate to efficient
algorithms for solving other simpler optimization problems,
such as single-objective uni-global problems, single-objective
multi-global problems and multi-objective uni-global problems.
The efficacy of the proposed algorithm in solving various problems
is demonstrated on a number of test problems.
Because of it's efficiency in handling different types of problems
with equal ease, this algorithm should find increasing use in
real-world optimization problems.
Presentation Slides: Click here to download presentation slides