Steven Newton's repositories
deep-learning
Python notebooks exploring topics in fully-connected deep neural networks with popular topologies like ResNet, MobileNet, YOLO, FaceNet, U-Net, and the Transformer architecture.
online-sports-betting
Most betting algorithms are competing with even better algorithms: the bookmakers'. Rather than design a competing model, Kaunitz, et al. (2017) show that it's possible to use aggregated bookmakers' odds to identify undervalued odds.
zillow-prices-linear-modeling-flask
Modeling the prices of single-unit homes using property features. The final model is deployed in Flask.
ace-versus-ace
MLB Statcast has performance data from every play at the major league level. I make a a head-to-head comparison of this data for Gerrit Cole and Jacob deGrom including hypothesis testing.
alien_invasion
Remake of a classic arcade game using Pygame
barrymores-wine-and-spirits
A Python script to clean inventory data. https://barrymoreswineandspirits.com/
brainpickings-nlp
An NLP analysis of Maria Popova's blog, brainpickings.org. I look in particular at the evolution of the themes of love, poetry, and science in her body of work.
intro-to-bert-model
Fine-tuning Google's SOTA, deep contextual language model on sentiment classification
montreal-nosql-openstreetmaps
"Mapping the globe is like eating a whale - you do it one bite at a time." I use MongoDB to explore Montreal's OpenStreetMaps data.
telco-churn-classification
Identifying the causes of customer churn and identifying likely churn with binary classification models.
time-series
Exercises in time series including a project modeling temperatures in Greenland.
youtube-eda
Exploration and hypothesis testing of YouTube data in various countries.
zillow-logerror-clustering
Zillow hosted a challenge to reduce the error on their estimates. I use k-means clustering and linear regression to model house valuation estimation errors.