federico-pizarrobejarano Goto Github PK

acados

Fast and embedded solvers for nonlinear optimal control

ccc

All the code I wrote for various Junior and Senior CCC contests

computer-vision-machine-learning

Abstract This investigation determines the extent that characters can be identified in images using the logistic regression and single-layer neural network algorithms. Optical character recognition (OCR) is a computer vision, supervised learning problem. The dependent variables were the optimal value of the regularization parameter lambda, the accuracy on the training, cross validation, and test sets, and the time needed to train each classifier. A dataset of 74,000 images composed of fonts, handwritten characters, and real images of letters and numbers was used. For the purposes of this investigation only a subset of the font dataset was used. Each image was resized to be 20x20 pixels and then converted to a 1x400 vector of pixel values. The logistic regression algorithm attempts to fit parameters to the 400 pixel values to form a hypothesis function. To optimize the parameters, the algorithm defines a cost function and then performs gradient descent on the parameters. The tunable parameters were additional features added in an attempt to create more complex, representative functions. A single-layer neural network passes the input data to a hidden layer where the data is partially processed. The partially processed data is then passed to the output layer where the final predictions are made. The tunable parameter was the number of hidden units in the hidden layer. The logistic regression algorithm achieved an accuracy of 85.14% with no added features and a lambda value of 1. The neural network achieved a significantly higher accuracy of 90.19% using 200 hidden units and no regularization. Logistic regression had a time complexity of O(n) while the neural network had a significantly better time complexity of O(√h). This paper investigates the properties of both algorithms as well as establishes the inability of both algorithms to identify characters to sufficiently high accuracies.

decentralized-multi-robot-exploration

Recreating a decentralized multi-robot exploration algorithm found here: https://www.researchgate.net/publication/235642260_Distributed_Value_Functions_for_Multi-Robot_Exploration_a_Position_Paper

dndtbs

A modified version of GubiD's GTBS for RPG Maker VX Ace that uses Dungeons and Dragons rules.

don-mills-online-judge

Here is TONS of code written for the Don Mills Online Judge. I know this code can be used by others to falsely pretend they have completed code for DMOJ but as it is a self-improvement site and gives no rewards, plagiarizing this seems unnecessary.

ecoo

All the code written for various ECOO competitions at various levels

federico-pizarrobejarano.github.io

Github Pages template for academic personal websites, forked from mmistakes/minimal-mistakes

image_denoising

Testing different image denoising techniques through the lens of convex optimization.

learning_django

microsoft-coding-competition

Code from 2017 Microsoft Coding Competition

online_dmpc

Code accompanying the RA-L / ICRA 2020 paper: "Online Trajectory Generation with Distributed Model Predictive Control for Multi-Robot Motion Planning"

safe-control-gym

PyBullet CartPole and Quadrotor environments—with CasADi symbolic a priori dynamics—for learning-based control and RL

utek

Completed Code from UTEK 2017 and UTEK 2018. UTEK 2017 is concerned with analyzing station locations

visual-odometry-cpet

Creating a visual odometry pipeline based on LSD-SLAM (https://vision.in.tum.de/research/vslam/lsdslam) to localize a rover using the CPET dataset (https://starslab.ca/enav-planetary-dataset/)