Titanic-Kaggle

Little experiment with a simple neural network and the kaggle competition: https://www.kaggle.com/c/titanic

Model details

Simple feed forward neural network based solution comprising of fully connected layers and a sigmoid activation function on each.

Layer details:

• 6 inputs
• 3 hidden layers
  • 6 Nodes
  • 10 Nodes
  • 4 Nodes
• 1 output node

Training:

Trained using the TensorFlow AdamOptimizer to minimize cost.

• learning rate = 0.001
• learning epochs = 10000