ninaboord Goto Github PK

2024-exampleapplication

CS342 2024 Example Application

av-particle-filter

A particle filter for autonomous vehicle handoff

bayesian-structure-learning

Takes in a dataset of multiple variables and their instantiations. Outputs the optimal bayesian network (via networkX) using a scoring algorithm and K2 search.

clock

I implemented a simple two-button clock from scratch. I also wired up the display circuit connected to my raspberry pi.

cold-emailer

I made this program when I was 16 and still learning how to code. It partially automates cold emailing. The algorithm takes in a name and company URL and spits out every possible email address of that individual. It's not the most beautiful, but it increased my cold emailing efficiency by 10x, helping me raise $38k+!

dj-copilot

modified-thompson-sampling

The Thompson Algorithm is the current "best" algorithm for the multi-armed bandit problem, but this is an area of ongoing research. Through a lot of trial and error, I found a unique, simple modification and in all simulations, my algorithm was more optimal than the Thompson Algorithm. I received a distinction for this project.

ninaboord.github.io

notion-assignment

printf

I implemented printf from scratch in C. I did not sleep this entire week. Enjoy!

q-learning

Implementation of Q-Learning, a reinforcement learning technique that finds the Q-function Q(s,a) or the best action we can take given our state to obtain a maximal reward. Implemented on a large dataset to test runtime.

rocketberry-pi

Programmed an original game reminiscent of "space invaders" from scratch on bare metal (Raspberry Pi) using my own graphics, memory allocator, and interrupts libraries. The game has exemplary graphics and performance and uses an accelerometer-enabled external controller to move a rocket ship and buttons to shoot lasers. This was a pair project.

spezitemplateapplication

Template application demonstrating the usage of the Stanford Spezi framework.

startup-predictor

Predicts whether a startup will be successful or not based on a list of features. Compares methods of logistic regression, neural net with 2 hidden layers, a wider neural net with 2 hidden layers, a neural net with 3 hidden layers, a wider neural net with 3 hidden layers.

wildfire-predictor

I implemented Logistic Regression from scratch as well as LASSO and Random Forests to do an analysis on wildfire data, predicting wildfires with up to 99% accuracy. A full project description can be found here: https://fan-roadway-bca.notion.site/Wildfire-Machine-Learning-Project-3f9b9c80a95f457491bfb57a0c86efe5

wordle-solver

My algorithm takes in a Wordle board (popular word game by the New York Times) and spits out all of the possible words in the English language that can solve the Wordle.