austin010209 Goto Github PK

2d-rigid-body-simulation

apply-dqn-on-atari-game-jellybean-world

best-ordered-prompts-in-few-shot

deepbind

Training and testing of DeepBind models on scATAC data

dna_bert

DNABERT: pre-trained BERT model for DNA-language in genome

edge-detection-and-hough-transformation

github_cmd_trial

homography-and-depth-estimation

impact-patterns-of-combining-model-pruning-and-continual-learning-on-model-performance

java_exercises

high school

multi-class-logistic-regression-and-gradient-descent

In this project, we implemented multi-class logistic regression from scratch and compared it with K-Nearest Neighbours (KNN) and Naive Bayes on the digits dataset and letters dataset from OpenML. We tracked performance indicators such as the accuracy and the running time of models with different hyper-parameters using a simple grid search. We have found that both logistic regression and KNN work well in classical multi-class classification problems. Logistic regression is fast during prediction while KNN struggles to scale with large datasets. Naive Bayes, on the other hand, has lower accuracy due to holding a strong and wrong assumption for our datasets.

multi-label-classification-of-mnist-image-data

particle-based-viscoelastic-fluid-simulation

Particle-based Viscoelastic Fluid Simulation

predicting-us-covid-19-hospitalizations-from-trends-in-google-symptom-searches

We use a US Google symptom search trends dataset[1] and a US COVID-19 hospitalization cases dataset[2] in order to compare how well K-Nearest Neighbour Regression and Decision Trees perform when predicting hospitalization cases from related symptom searches.

running-time-and-accuracy-comparison-between-miles-cplex-gurobi-and-mosek

sift-feature

spring-system-simulation

stable-eulerian-fluid-simulation

viprs_empirical_prior