Projects
Diff-Multi-ESM
This is an ongoing project aiming to emulate multiple Earth System Models (ESMs) simultaneously to predict extreme weather events.
We utilize diffusion models and deep learning techniques to achieve this goal.
So far, our results have been promising, and we are continually refining the model.
This project has been worked on through hutchresearch and builds upon the work detailed here: DIFFESM: Conditional Emulation of Earth System Models with Diffusion Models
Cribbinator
Cribbinator is a cribbage bot that I created to learn some more about game theory.
The bot uses minimax with alpha-beta pruning to determine the best move.
The bot is capable of playing against itself or a human player.
I had a lot of fun creating the art and animations for this project :)
Deep Neural Network from Scratch
This project was done for my machine learning class at WWU.
The goal was to create an arbitrary deep neural network from scratch using only numpy.
The model was capable of handling any number of neurons and layers desired.
It also had the ability switch between classification and regression tasks.
I implemented back propagation and gradient descent to train the model on various datasets.
Grapheme to Phoneme Model
This was a sequence to sequence model I created for my deep learning class at WWU.
The goal was to create a model that could take in any word and output the phonemes of that word.
Several different model approaches were implemented, including an RNN, LSTM(& bi-directional), GRU, and a Transformer model.
Variational Auto Encoder
This was a project that I did to understand the implementation and use case of creating and tuning Variational Auto Encoders.
I used the Fashion MNIST dataset to train and fine-tune the model.
I implemented several different architectures as well as techniques such as KL Annealing.
I also created several different visualizations to better understand the latent space.
This GIF shows the model learning how to reconstruct the images over a small number of steps.
ml_pipeline
This is a project that I created to streamline the proccess of creating and training machine learning models.
This pipeline is designed to be modular and easily adaptable depending on the desired usecase.
I use this pipeline for all my machine learning/deep learning needs.
A simple Fashion-MNIST classifier and the VAE are example projects where I used this.
(all art, graphics, and gifs, were made by me. feel free to use them!)
This is an ongoing project aiming to emulate multiple Earth System Models (ESMs) simultaneously to predict extreme weather events. We utilize diffusion models and deep learning techniques to achieve this goal. So far, our results have been promising, and we are continually refining the model. This project has been worked on through hutchresearch and builds upon the work detailed here: DIFFESM: Conditional Emulation of Earth System Models with Diffusion Models
Cribbinator is a cribbage bot that I created to learn some more about game theory. The bot uses minimax with alpha-beta pruning to determine the best move. The bot is capable of playing against itself or a human player. I had a lot of fun creating the art and animations for this project :)
This project was done for my machine learning class at WWU. The goal was to create an arbitrary deep neural network from scratch using only numpy. The model was capable of handling any number of neurons and layers desired. It also had the ability switch between classification and regression tasks. I implemented back propagation and gradient descent to train the model on various datasets.
This was a sequence to sequence model I created for my deep learning class at WWU. The goal was to create a model that could take in any word and output the phonemes of that word. Several different model approaches were implemented, including an RNN, LSTM(& bi-directional), GRU, and a Transformer model.
This was a project that I did to understand the implementation and use case of creating and tuning Variational Auto Encoders. I used the Fashion MNIST dataset to train and fine-tune the model. I implemented several different architectures as well as techniques such as KL Annealing. I also created several different visualizations to better understand the latent space. This GIF shows the model learning how to reconstruct the images over a small number of steps.
This is a project that I created to streamline the proccess of creating and training machine learning models. This pipeline is designed to be modular and easily adaptable depending on the desired usecase. I use this pipeline for all my machine learning/deep learning needs. A simple Fashion-MNIST classifier and the VAE are example projects where I used this.