Housing-Market-CNN

I explore multimodal learning by using a Convolutional Neural Network (CNN) in order to predict housing prices based on the basic information on the house (such as the number of bedrooms, bathroom, square footage,zipcode, etc) and images of the house. The architecture used (in initial networks ) includes ReLU activation functions in all layers (including output), SGD optimizer, learning rate of 0.01, mean square error loss function, two hidden layers with 1024 neurons each, and one hidden output layer with an adequate number of neurons in the output layer for the regression problem. Architecture components are altered in a grid search in an exploration on their influence on network performance.