In-Browser Demo

In this Demo, a Neural Network trains to predict house prices in San-Francisco.

The dataset consists of 9 parameters about residential city blocks. Green dots represent low prices, Red dots represent high prices. The moving dots are the model's predictions.

Here is the dannjs model used in this demo:

const nn = new Dann(9, 1);
nn.addHiddenLayer(8, 'leakyReLU');
nn.addHiddenLayer(8, 'leakyReLU');
nn.addHiddenLayer(6, 'leakyReLU');

nn.outputActivation('sigmoid');
nn.makeWeights();

nn.setLossFunction('mae');
nn.lr = 0.0001;

You can find the dataset here

Source code here