AI Model Share Initiative at Columbia University
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
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.)
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