Kaggle foot-in-the-door competition, see https://www.kaggle.com/c/titanic for more information.

nn.py - Using a 3-layer neural network to tackle the problem. More layers led to diminishing returns, but more nodes per layer led to better performance, up to a certain point.

avg.py - Using multiple models and average the output. Logistic Regression and Stochastic Gradient Descent consistently perform worse. By excluding those models, the script generates a performance on-par with the output produced by the neural network.

Features that have high correlation (|correlation| > 0.1)to survival are selected, below is the table showing correlation,