Here are some of my projects that I consider worth sharing with the community. Many include the GitHub link for the code and a downloadable PDF. Feel free to access and use those resources when applicable.
The idea surged thanks to a lab project in which we used ZigBee boards to implement the triangulation technique. In this project we used three of the boards as reference in order to calculate the unknown position of a third board based on the intensity of the received signal at each point. The idea isContinue reading “Thesis: Indoor Localization Techniques for Wireless Systems”
Developed with a Convolutional Neural Network and TensorFlow, this project surged because of an interview question which I considered really relevant in my area of knowledge and I generate a complete research explaining concept involved from the basis.
To follow with the research paper, I decided to develop the code using open source GitHub projects and tweets databases. My surprise was to discover that all the open source projects I found use TensorFlow v1, facing incompatibility problems with the last version of Python. Therefore, I decided to develop this project using TensorFlow v2 and make a positive contribution to the open source community.
Developed with a Recursive Neural Network and PyTorch, we used this open source GitHub code and modified it to predict the price of real stocks downloading any company information through Alpha Vantage API. We included technical indicators from the market and conducted a short study to compare what information outputs a better prediction, varying the neural network parameters or the most common indicators.
The results were interesting, as it seems that adding technical indicators used by human reasoning to make better predictions had almost no effect on the algorithm output. To be able to make the comparison we downloaded up to 5 years of the stock price data and used 70% of the data for training purposes and 30% to compare the prediction. As personal variation, I am modifying the algorithm to use all the data for training purposes and predict the future price of the stock.
Starting with a study about the available technologies to realize localization inside buildings, I continue explaining the different methods used to calculate the position of unknown objects. Choosing the Radio Signal Strength Indicator (RSSI) as the measured parameter, I test output errors of several algorithms widely used in the implementation of this systems.
After the initial study, I develop a self made algorithm that outputs minimum error in all simulated circumstances and I test its accuracy when implemented in a parking environment. To sum up, I calculate the implementation cost of the system.
Follow My Blog
Get new content delivered directly to your inbox.