Alpha-Quoridor 🕹️

We build an agent that masters Quoridor Board Game. Quoridor is not so famous game and there is not a lot of research about it. It has a state-space complexity similar to Chess with a higher game-tree complexity. We code this game like AlphaGo Zero from Deepmind, which learns to play the game by self-playing without human games data.

🕒April 30, 2020 · 1 min · Kanvaly Fadiga

Credit default prediction 💰

In this work, we were trying to predict credit default risk. We were dealing with historical credit clients' information and were willing to predict whether or not a client will reimburse his credit. This is a standard supervised classification task. First of all, we analyze and clean the data. Then, extract the relevant pattern. Then train models and optimize the most promising. Finally, we use stacking to build a strong predictor....

🕒April 7, 2021 · 1 min · Kanvaly Fadiga

Link prediction 🔗

Our task was on predicting links between pages in a subgraph of the French web graph. The web graph is a directed graph G(V, E) whose vertices correspond to the pages of the French web. A directed edge connects page U to page V if there exists a hyperlink on page ​ U pointing to page ​ V . From the original subgraph, edges have been deleted at random​ . Given a set of candidate edges, our job was to predict which ones appeared in the original subgraph....

🕒January 1, 2020 · 1 min · Kanvaly Fadiga

Position Based dynamics 🤾‍♂️

One approach for the simulation of dynamic systems in computer graphics is force-based. Internal and external forces are accumulated from which accelerations are computed based on Newton’s second law of motion. A time integration method is then used to update the velocities and finally the positions of the object. A few simulation methods (most rigid body simulators) use impulse-based dynamics and directly manipulate velocities. The paper I studied present an approach that omits the velocity layer as well and immediately works on the positions....

🕒May 31, 2020 · 1 min · Kanvaly Fadiga

Object Tracking

Our goal was to firstly build a multiple object tracker by different methods of detection and tracking. Then choose an application domain. We chose to work on surveillance and security. The user can segment the region of interest in a video scene and any new object motion, position, and time spend there. Then after a certain amount of time, we start identification (if it is a human or a car) in an asynchronous way and launch an alarm given the result of identification....

🕒December 31, 2019 · 1 min · Kanvaly Fadiga