Latest Projects
...Always under construction...
Stochastic modeling for efficient, Light-Weight, machine learning models
This project aims to achieve light-weight, embedded, Machine Learning models through mathematical modeling, probabilistic, and symbolic approaches. Through data mined from different cases in games and other stochastic scenarios; the goal is to create accurate, light-weight, models that can achieve better performance, compared with ordinary neural-networks or other pre-packaged models.
Data Mining Analogic Games (DMAG).
We are collecting data from analogic games and mining these data in order to extract interesting patterns. These games, unlike Chess or Go, do not have an AI capable to compete with human professionals, therefore we are mining data to create information that can be further used in intelligent bots. Furthermore, these data can be used as challenges for machine learning and data mining, since most of the public available datasets have solutions that can be easily found on the internet, new datasets are always useful for education and practice.
The project aims to develop a set of techniques and methods for efficient solutions of Stochastic Automata Networks (SAN). SAN formalism has many applications ranging from processing production lines, through natural language, computer networks, protocols, parallel machines, software engineering and even fields related to the earth sciences, such as geology and meteorology.
Paleoprospec project aims to optimize the fossil fuels prospection on the Atlantic coast through paleogeographic and paleoclimate models. Two research teams participate of this project: Petrobras/CEPAC and
FACIN/PUCRS. The main topics of interest are concentrated in:
- Database systems and algorithms;
- Development of numerical software tools;
- Development of modelling tools for parallel and synchronized systems;
- Parallel programming.
For more details visit Paleoprospec Website.
PEG
