- Gain experience in scientific visualization and data exploration
- Improve presentation skills and scientific redaction
Scientific visualization is the art of scientific data analysis. Its main purpose is to display two or tree-dimensional phenomena in a way the user will be able to easily explore and analyse information contained in data. Linked with computer graphics and other branches of science it uses, inter alia, topological analysis to help scientist undestand the structure of their data.
This kind of analysis require interactivity. As today computers grow horizontally (distributed / many-core architectures), we need to rethink algorithms to be efficient in parallel. Moreover in-situ visualization become increasingly employed, causing actual post-processing visualization algorithms to be executed on supercomputers.
I am currently working at Kitware (Lyon) under the supervision of Julien Tierny and Pierre Fortin, my two PhD advisors. We focus on the contour tree creating a new algorithm for its computation in parallel on shared memory workstations:Kitware and LIP6.
This internship was supervised by
The purpose was to develop a new algorithm to compute the contour tree in parallel. The contour tree is a topological abstraction vastely used for large data exploration, segmentation, isosurfaces extraction,...
The main purpose of this algorithm was (parallel) efficiency. This internship involved high performances computing, parallel programming and topology.
I have completed last year a 5 month internship in this corporation in order to assist the development team in various realms.
Among assignment I had, I had to set up an automatic completion on the main search bar aimed to help users and improve communication between buyers and sellers, but also to avoid mistakes by correcting input text on the fly. This correction was based on machine learning. (9 weeks)
I will present you here three projects among those I did in school.
You can try out this project: here.
The aim of this project was to manage students allocation to rooms during examinations and so minimize cheating chances. This application is able to take into account several rooms and several exams at the same time.
For performance reasons the main algorithm is written in C++, while data are stored and transmitted in JSON. The result is optimized in terms of distance: student of a same class are away from each other, but students from various classes are near.
You can try out this project: here.
Click on "voire analise" or here if you want to see the analysis. This application called MinilightX is a web statistics tool. It displays information about clicks and clicked objects and allows the webmaster to optimize his benefits showing the best position for each advertisement. The webmaster can see a Data Space Analysis of the clicks per object on the "Deduction" section. It shows groups of positions clicked by users for this object and so best positions for this object. To help him decide which one is the best solution, the last part make proposals of advertising positions.
MinilightX uses its own library for curves (in SVG) that we have made from scratch and GoogleChart for diagrams.
It's written in Java (EE), using Google Apps Engine a NoSQL database.
This project consists of a virtual ecosystem based on the individual centred approach in order to
several prey-predators models.
The environment is composed of water, grass and soil, which is eaten grass. Species are herbivorous,
carnivorous or omnivorous
and it is possible to create as many varieties as you want.
Many parameters are configurable for each species like the digestion speed or the threshold of hunger, thirst and so forth. Once species are configured you just have to launch the software and observe species evolving on the generated map.
With good parameters we can observe the Lotka-Volterra equations on map with herbivorous and carnivorous as prey-predator model is respected. When they are only herbivorous the Verhulst model is respected.
This project was written in C (SDL, SDL_images) and creates cvs files with life statistics for the differents species.