Skip to content
This repository has been archived by the owner on May 22, 2021. It is now read-only.

Latest commit

 

History

History
41 lines (26 loc) · 2.12 KB

File metadata and controls

41 lines (26 loc) · 2.12 KB

AI Model Share Initiative at Columbia University

Mini-Hackathon 1 : Tabular Data - Predict country level happiness

Find the best model to predict a country's happiness ranking using various social variables (including those measuring perceptions of corruption, GDP, maintaining a healthy lifestyle, or social support, etc.). By running both DL and ML models, I could compare which ones showed stronger performance for this given task.

Data: 2019 World Happiness Survey Rankings + ISO 3166 Country Codes

Model : Convolutional neural network models using Keras + Machine Learning models

Code : https://github.com/jinokwon/AI-Model-Share-Projects-at-Columbia-Univ/blob/master/Project1/ML_DL_Project_1_Predicting_Happiness.ipynb


Mini-Hackathon 2 : Image Data - Predict brain tumor

Find the best model to predict brain tumor using MRI image data. I preprocessed (one-hot-encoded) image data and conducted object detection with deep learning models.

Data: 253 diagnositic brain MRI images

Model: Recurrent Neural Network models with Keras (w/ different numbers of hidden layers, epochs, and 2D Max Pooling, etc.)

Code : https://github.com/jinokwon/AI-Model-Share-Projects-at-Columbia-Univ/blob/master/Project2/ML_DL_Project_2_Predicting_Brain_Tumor_pynb.ipynb


Mini-Hackathon 3 : Text Data - Classify BBC news categories

Find the best model to predict the category of a given news article with based on a short text from that article. I ran various RNN (i.e., LSTM) models to conduct text classification.

Data: 2225 BBC News articles

Model: 5 Neural network models with Keras

  • Model 1 w/ an embedding layer and dense layers (but w/ no layers meant for sequential data)
  • Model 2 w/ an Embedding layer with Conv1d Layers
  • Model 3 w/ an Embedding layer with one sequential layer (LSTM or GRU)
  • Model 4 w/ an Embedding layer with stacked sequential layers (LSTM or GRU)
  • Model 5 w/ an Embedding layer with bidirectional sequential layers

Code: https://github.com/jinokwon/AI-Model-Share-Projects-at-Columbia-Univ/blob/master/Project3/ML_DL_Project_3_Classifying_BBC_News_Categories.ipynb