Artificial Neural Network

When this page is loaded the neural network will be initialized with random weights for each neuron and will begin training itself on the dataset shown in the table. The neural network code is shown in the github gist at the bottom of the page.

After training finishes, use the "sleep" and "study" input boxes to input 2 values, click on "estimate" button to have the network make an educated guess about the output. The closest guess will be highlighted in the training data set table. If you choose two numbers that exist in the training data the guess should be accurate. Weights are randomly initialized on page load.

This demo makes use of WebWorkers to process the neural network training. Please use a modern web browser with support for web workers.

The network may take 5-10 seconds to converge. If the network does not finish learning within 20-30 seconds reload the page and try again.

• Learning rate: 0.8
• Maximum error margin: 0.0001

Below are 2 charts

• Chart 1:

  This chart represents the output of our cost function at each training iteration.

• Chart 2:

  This chart uses the network for regression to find an estimate of each possible combination of Sleep and Study hours (0-24).