This is a collection of projects built utilizing Python during personal study. Many projects have been designed as templatized versions within Google Laboratory notebooks so you can clone them and easily utilize them in your own projects.
Web application built with Flask(Python) backend, featuring full user authorization leveraging JSON Web Tokens. User profiles demonstrate custom text generated from custom pre-trained GPT-2 model. Social interactivity between message boards and content sharing/editing added. Automated email integrating added utilizing SendGrid.
Coursework created through multiple machine learning courses from Perian Data, Jose Portilla, and 365 Careers. Including linear and logistic regressions to perform predictions as well as time series forecasting based on real world data. Integration of various statistical libraries with Python like SkLearn, Pandas, NumPy and Seaborn.
Courswork from multiple machine learning, deep learning, and AI development with John Harper, Mark Winterbottom, and Jose Portilla. Django development centered integration for machine learning and neural netowrks for time series forecasting, Natural Language Processing, and Computer Vision.
- Image classification projects created to identify pedestrians vs. roads in basic road safety automation.
- Generative Adversarial Networks utilized to create realistic images of dogs from scraped collection of images. Notebook also repurposed to design basic logos.
- Fake News Classifier - created utilziing LSTM neural network, 2,000,000 news articles from All the News dataset, Vaex lazy loading framework, and Vader sentiment analysis
- Working and fine-tuned integration of GPT-2 Transformer Neural Network - Pre-trained on scraped data from E-book collections. Easily extensible to be custom trained.
- Long Short Term Memory Neural Network - created utilizing 2 channel audio from youtube downloads. Creates predicted tones best on train/test split of data.
- Negative Matrix Factorization - machine learning model to analyze topics of focus for news articles, with Vader sentiment analysis integration.
- News Text Generator - GPT-2 custom-trained model on subsections of All the News dataset selected utilizing Negative Matrix Factorization.
- SQUAD GPT-2 - Conversational Q&A GPT-2 Bot trained on SQUAD question and answer dataset.
- Vader Sentiment Analysis - Text analysis project created utilizing Vader library to determine tone, and sentiment of text.
- Text Generation LSTM - Custom created Long Short Term Memory neural network designed to generate text. Gated neural network retains memory beyond the capabilities of Recurrent Neural Networks but still has severe memory/text length limitations in comparison to Transformer verions.
- Text To Speech - Text to speech script created utilziign Pystixx library and text generated from GPT-2 text model.
- Bitcoin Prediction LSTM - Long Short Term Memory Gated Neural Network featuring price data scraper to analyze and predict the future price of Bitcoin Crypto Asset.
- Bitcoin Prophet Model - Basic integration of Prophet AutoRegression Neural Network to predict the price of Bitcoin.
- Bitcoin TA/Prophet Bokeh - Statiscal analysis of Bitcoin price data in comparison with other financial assets including other leading crypto assets. Features analysis of Plan B's Stock to Flow Model, Marcel Berger's argument of cointegration between Bitcoin and other financial assets. Interactive Bokeh chart to perform technical analysis as well as improved Prophet model to predict future price expectations.
- First TensorTrade Bot - Automated trading bot created utilizing TensorTrade and Tensorflow frameworks. Integrated historical price data for leading crypto assets to model market conditions, reinforcement learning applied to historical data to create a profitable trading bot.