# Posts

# Spring pics

… (Read More)

Photos of a Sunny afternoon at a nearby area

… (Read More)

# Painting mindmap (braindump)

I am a huge fan of mindmapping. Perhaps its just an excuse to keep the information unorganised in thoughts, not really. Thoughts even though they seem unorganised are related, and mindmap captures the relationships without digressing from the original queue. And looking back at a mindmap at later time can bring back the thought process in exact same way. Its a very powerful tool. So, whenever I come across any ideas I put 'em in a mindmap.
I want to write some blog post about painting I have created the following mindmap before putting that in words. Its not complete but hey its just a dump so far... :-)
click on image for larger size
(created using FtsMind app for windows phone8)

… (Read More)

# Genetic algorithms

Got very little time on this weekend. I was again interested in techniques for solving optimisation problems. Many of the problems in real-life are optimisation problems, where unknows are more than the information (equations) at hand. Examples: fitting of surface patches / curves on points, closest path - path planning computations, Nesting (manufacturing) for optimising the material cut from the raw material sheets, inverse kinematics, fluid mechanics, medicine, many other areas...
Optimisation techniques is a huge research area, and it would require considerable amount of time to study every technique. I have used basic optimization methods before, but I am starting to learn about advanced techniques. Many problems can be formulated and solved by calculus methods, but they become exceedingly complicated with more number of unknowns. A simple minima, maxima problem in single variable can be solved by calculus, where f`(x)=0... then solving by Newton Raphson method, other numerical methods, but when there are multiple variables we have to find partial derivatives which become complex with more variables. Some of the calculus approaches can only find out the local maximas for the functions, in that case optimization results heavily depend on the initial guesses. There are advanced numerical methods to solve non-linear optimization problems to get global roots.
GA uses a heuristic approach based on natural selection phenomenon. Got reading this book again about GA by David Goldberg, I am fascinated by the GA technique. Its a searching technique which narrows down to the answer in solution (search) space using different combination functions on the existing pool of intermediate solutions which they call as chromosomes. It can be used for problems that are NP-Hard as well. Most important part of this techniques is the problem formulation, we have to represent the problem in a particular form so that we can use GA. To get familiar with this technique I want to try and solve a problem with this, possibly Travelling salesman problem or best fit circle in points or both.
Useful links:
http://hyperphysics.phy-astr.gsu.edu/hbase/math/maxmin.html
http://en.wikipedia.org/wiki/Nonlinear_programming
http://www.ai-junkie.com/ga/intro/gat1.html
More later...

… (Read More)

# Quickhull algo in javascript - pt4

Finally got the hang of OOP in javascript.. the public functions for the classes are prototype methods.The quickhull works! Don't use the code for some serious stuff.. its come out of casual coding and might have some errors. Some escape sequences and optimizations remain, but as for the javascript experiment it looks good enough.. I had to put quite a bit of code in for even this simple algo, CVector, CPoint, CEdge and CHull classes. Perhaps a reusable good js geometry util can be written for general usage :-)
More later..

… (Read More)

subscribe via RSS