Pinned Repositories
Adv_ML_DL_NLP_GM
Repository for Stat 72556 Advanced Machine Learning, Deep Learning, NLP, and Graphical Models
An_Analysis_of_Pakistani_Homes
Different machine learning models to predict home prices in Pakistan. We use Linear Regression Methods, K-Nearest Neighbors Methods, and Decision Tree Methods. We look at many differnt predictors such as: area, location, number of bathrooms, number of bedrooms, and other predictors. Dataset found on Kaggle at: https://www.kaggle.com/ebrahimhaquebhatti/pakistan-house-price-prediction
Credit-Card-Fraud-Classification
CSCI133
Repository for CSCI 133 code
DigitClassifier
Model created and trained using Tensorflow. The dataset used is the MNIST handwritten digits dataset. The GUI was created using tkinter. User is able to “handwrite” a digit and then the model will return its best guess at which digit it is. Since this is a categorical problem, I decided to use a sigmoid activation function and the sparse categorical cross entropy loss function. Achieved an accuracy of 99.9% among the test data.
FaceIDWithSiameseNetworks
One-shot learning approach to create a model trained off thousands of pairs of images that can determine whether 2 images contain the face of the same person or not. Used a Siamese network architecture to create embedding vectors in representation space for 2 images at a time. Minimized contrastive loss using SGD. Optimized for precision (85%).
Gasoline-Demand-Forecasting
Explored patterns, trends, and seasonality of US gasoline demand data. Compared models based on the mean squared error of the testing set. Employed ARIMA, XGBoost, Facebook Prophet, and Recurrent Neural Network to forecast future gasoline demand
HAL9001
Personal assistant and chatbot created using TensorFlow. The model utilizes the “bag-of-words” NLP technique to decipher the command of the user. Model created using TensorFlow. Data manipulation done through the Natural Language Toolkit (NLTK). Incorporates the Spotify API to allow Spotify control and the OpenWeather API for real time weather retrieval. HAL9001 is able to search Google, YouTube, and Wikipedia. Other tasks include retrieving the time and date. Light conversation is also possible.
PlanetariumArcadium
A full (PERN) stack planetary system visualizer web app created by the "AstroNerds", a team of 4 people in their CS minor capstone class. The data was obtained from the NASA Exoplanet Archive and stored in a PostgreSQL database served through an Express server with more than 3000 items. Front end built using JavaScript, React.js, and Node.js. The project is hosted in an AWS EC2 instance on a Ubuntu server. Users are able to search up, visualize, and learn about distant stars and orbiting planets.
PortfolioOptimization
Worked with a team of 4 people total to find methods to optimize a portfolio with an arbitrary number of assets. Used a gradient based constrained optimization algorithm to find the optimal portfolio allocation given n stocks. Descent direction determined using “generalized gradients”.
Raafi101's Repositories
Raafi101/DigitClassifier
Model created and trained using Tensorflow. The dataset used is the MNIST handwritten digits dataset. The GUI was created using tkinter. User is able to “handwrite” a digit and then the model will return its best guess at which digit it is. Since this is a categorical problem, I decided to use a sigmoid activation function and the sparse categorical cross entropy loss function. Achieved an accuracy of 99.9% among the test data.
Raafi101/Adv_ML_DL_NLP_GM
Repository for Stat 72556 Advanced Machine Learning, Deep Learning, NLP, and Graphical Models
Raafi101/An_Analysis_of_Pakistani_Homes
Different machine learning models to predict home prices in Pakistan. We use Linear Regression Methods, K-Nearest Neighbors Methods, and Decision Tree Methods. We look at many differnt predictors such as: area, location, number of bathrooms, number of bedrooms, and other predictors. Dataset found on Kaggle at: https://www.kaggle.com/ebrahimhaquebhatti/pakistan-house-price-prediction
Raafi101/Credit-Card-Fraud-Classification
Raafi101/CSCI133
Repository for CSCI 133 code
Raafi101/FaceIDWithSiameseNetworks
One-shot learning approach to create a model trained off thousands of pairs of images that can determine whether 2 images contain the face of the same person or not. Used a Siamese network architecture to create embedding vectors in representation space for 2 images at a time. Minimized contrastive loss using SGD. Optimized for precision (85%).
Raafi101/Gasoline-Demand-Forecasting
Explored patterns, trends, and seasonality of US gasoline demand data. Compared models based on the mean squared error of the testing set. Employed ARIMA, XGBoost, Facebook Prophet, and Recurrent Neural Network to forecast future gasoline demand
Raafi101/HAL9001
Personal assistant and chatbot created using TensorFlow. The model utilizes the “bag-of-words” NLP technique to decipher the command of the user. Model created using TensorFlow. Data manipulation done through the Natural Language Toolkit (NLTK). Incorporates the Spotify API to allow Spotify control and the OpenWeather API for real time weather retrieval. HAL9001 is able to search Google, YouTube, and Wikipedia. Other tasks include retrieving the time and date. Light conversation is also possible.
Raafi101/PlanetariumArcadium
A full (PERN) stack planetary system visualizer web app created by the "AstroNerds", a team of 4 people in their CS minor capstone class. The data was obtained from the NASA Exoplanet Archive and stored in a PostgreSQL database served through an Express server with more than 3000 items. Front end built using JavaScript, React.js, and Node.js. The project is hosted in an AWS EC2 instance on a Ubuntu server. Users are able to search up, visualize, and learn about distant stars and orbiting planets.
Raafi101/PortfolioOptimization
Worked with a team of 4 people total to find methods to optimize a portfolio with an arbitrary number of assets. Used a gradient based constrained optimization algorithm to find the optimal portfolio allocation given n stocks. Descent direction determined using “generalized gradients”.
Raafi101/DeepLearningMethodsforClassifyingAutomobiles
Used standard machine learning algorithms such as logistic regression and random forest, followed by many deep learning models such as a fully connected network, convolusional neural network, and transfer learning using ResNet50. Compared all of the models and concluded the ResNet50 model performed best with a test accuracy of 76.3%.
Raafi101/Economics-and-Machine-Learning
Raafi101/FaceIdApp
Web app for my recreation of FaceID with Siamese Neural Networks. Repo: https://github.com/Raafi101/FaceIDWithSiameseNetworks
Raafi101/FordPricePredictionBayesianLinReg
Used Bayesian Linear Regression to create a model that predicts Ford car prices. Utilized Hamiltonian Monte Carlo algorithm to accept or reject random samples. Used Confidence intervals to show accuracy of model.
Raafi101/GenerativeAdversarialNetworks
Raafi101/HAL9000
The latest installment of the HAL computer lineage
Raafi101/Intro_to_DS_and_ML
Repository for Statistics 72401 Introduction to Data Science and Machine Learning
Raafi101/Minesweeper
Remake of the original retro classic with a graphical twist
Raafi101/Modeling_and_Visualization
Repository for Stat 78600 Modeling & Visualization
Raafi101/OnlinePortfolio
OUTDATED! SEE ONLINEPORTFOLIOV2! My online resume and portfolio. Includes information about me and my projects. Video demonstrations of projects is also available.
Raafi101/OnlinePortfolioV2
UPDATED!!! My online resume and portfolio. Includes information about me and my projects. Video demonstrations of projects are also available.
Raafi101/Pakistani_Rupee_Note_Classification
Raafi101/Pong
A small adaptation of yet another retro classic, Pong! Grab a friend a try it out!
Raafi101/Raafi101
GitHub Header
Raafi101/ReinforcementLearning
Raafi101/ShortestPathFinder
A small path finding app that uses Dijkstra's path finding algorithm made in C++
Raafi101/Snake
A remake of a fan favorite game. Made using Pygame
Raafi101/Solve-Differential-Eqs-with-Neural-Nets
Raafi101/StockPricePredictor
Model created and trained using Tensorflow. The dataset used is Yahoo!'s stock price dataset. This is a very crude program that is Not accurate. Although the program is not very accurate, I do not consider this a failure. The main purpose of this project was to be an intro into Tensorflow and for me to learn the basics of Tensorflow. With this goal in mind, this project was a complete success!
Raafi101/streamlit-example
Example Streamlit app that you can fork to test out share.streamlit.io